test: delete low-value and redundant suite files by jackwener · Pull Request #2462 · apache/maka

jackwener · 2026-08-07T17:38:57Z

Summary

Delete 102 low-value or redundant source test files across Desktop, UI, Runtime, Headless, CLI, and Core, reducing the source test-file surface from 988 to 886 (-10.3%).

This branch is now based on the current main. The cleanup deliberately keeps tests that are the only guard for an active dependency patch or a security-sensitive boundary.

Removed

Desktop markup, copy, presentation, chrome, and overlapping geometry/workbar coverage
UI localization/copy panels, quiet chrome, empty states, and presentation pins
Runtime thin wire and duplicate provider helper coverage
Headless duplicate A/B, toolchain, and prompt-summary coverage
CLI layout/completion coverage and Core copy/locale helpers

Kept after review

Active dependency-patch guards:
- AI SDK tool-call index tracking
- Astryx List accessible-name forwarding
- Astryx new-message indicator behavior
Security-sensitive boundaries:
- split-chunk tool-output secret redaction
- permission escalation confirmation and persistence
- task-ledger prompt injection/redaction wiring
- Responses API store: false
- session-title internal-instruction stripping
Root authority, credentials, storage safety, primary runtime/session/tool suites, and fail-safe CI planning

Additional change

CLI process-contract tests ignore Node's ExperimentalWarning: SQLite stderr noise while continuing to reject other stderr output.

Validation

git diff --check origin/main...HEAD
npm run format:check
npm test — all workspace tests passed after synchronizing with current main

GitHub CI is running on the updated head.

Remove 122 source test files (~12% of packages/apps test files) that were markup/copy contracts, overlapping e2e chrome, duplicate UI presentation pins, and low-signal headless AB/toolchain matrix cases. Keep safety, authority, and primary product unit paths. Also tolerate Node SQLite ExperimentalWarning in CLI run process contracts so stderr-empty assertions are not flaky on modern Node.

Bring back e2e journeys, headless matrix/export, catalog contract, AiSdkFlow, and subscription non-leak coverage that are sole or primary gates. Keep the remaining ~11.6% file-level reduction of true low-value markup/chrome/matrix duplicates.

Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

This reverts commit 5ec1797.

Part of apache#2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until apache#2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds.

Part of #2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until #2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds. Co-authored-by: 陆逊 <211125649+UncertaintyDeterminesYou4ndMe@users.noreply.github.com>

apache#2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — apache#2161 pinned it against a containing block as tall as the conversation, apache#2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in apache#2462, and the multi-prompt fixtures it ran on in apache#2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine.

…2923) * fix(ui): give the prompt rail's tick bars a box again #2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — #2161 pinned it against a containing block as tall as the conversation, #2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in #2462, and the multi-prompt fixtures it ran on in #2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine. * fix(ui): make the prompt rail's hover and jump behave Four things the rail got wrong once it was visible again, found by using it: - A 4px gap between ticks was a band where the pointer was over the rail and over no tick, so the dock-style hover falloff dropped out and picked up again every few pixels of travel. The rail's `gap` moves into the ticks' own `padding-block`: same pitch, hit boxes now tile. - The hover preview waited 300ms before opening — Astryx's HoverCard default, meant for a pointer crossing a wide row on its way somewhere else. A tick is 22px of rail that nothing is on the way to, and the wait is the one part of this hover with no motion in it. Now 120ms. - The highlight glided 280ms to wherever a click landed, so crossing twenty prompts read as the bar flying off across the rail. A click now owns the highlight until its scroll settles: no glide, and the scroll no longer walks the highlight through every prompt it passes. - The first click into a session did nothing until the reader scrolled by hand. See below. That last one is a collision between Astryx's auto-follow lock and the progressive transcript mount, and neither side is wrong on its own. `useChatStreamScroll` unlocks on a scroll up, detected by comparing scrollTop between events — but it ignores any scroll event that arrives with a changed scrollHeight or offsetHeight, because Chrome fires those when content resizes and they are not the reader moving. A jump into an unmounted turn mounts it and the fill that follows changes scrollHeight for several frames, so the jump's own scroll is invisible to the lock: it stays on, and `scrollIfLocked` pulls the transcript back to the bottom. Only a wheel gesture broke it, which takes a separate path in Astryx. `holdJumpDestination` re-aims at the target on each height change until the fill stops. The last of those scrolls lands with a stable height, which is the one the lock finally reads as a scroll up. Measured on the 30-prompt fixture: clicking the first tick went to scrollTop 7042 (the bottom) and now goes to 24 and holds. The fixture grows from 8 prompts to 30 because the progressive mount's initial window is 10 — at 8 the head of the transcript is already mounted and the jump-into-unmounted-turns path never runs at all. Coverage note: the e2e case for the first click is an end-to-end check, not a guard. Whether the lock wins depends on which frame the fill lands on relative to a smooth scroll still in flight, and it goes green against the unfixed renderer often enough to be worthless as one. The guard is the `holdJumpDestination` unit test, which drives the frames itself. * fix(ui): own a rail jump through the mount instead of racing it Review of #2923 found the jump's ownership bound to a clock rather than to the navigation, and the e2e case that was supposed to guard it asserting almost nothing. Both hold. Jump ownership: - A second click during a jump only replaced the target; the first click's 700ms timer still governed, and could clear the second jump mid-flight. Each click now carries its own sequence and starts its own hold. - The fixed window is gone. A hold runs until the progressive mount reports the transcript filled AND nothing has moved for a few frames, so a long transcript is never released mid-fill, and it ends the moment the reader touches the transcript (wheel, touch, pointer, key) rather than outliving their interest in it. Chasing the "just release auto-follow" direction the review preferred found that ChatLayout publishes no such seam, so this adds one — `unlockAutoFollow` on `ChatLayoutContextValue`, exposing the scroll hook's existing `unlock` (patch hunk + patches/README entry). It is necessary and it is not sufficient, which the earlier framing got wrong: - Astryx re-locks on any `scrollend` that settles near the bottom, and a session that opens at the bottom produces exactly that while the mount is still catching up. Releasing once at the click is undone before the jump goes anywhere — traced: released at the click, landed at 154ms, dragged back to the bottom by 166ms. The release is now re-asserted for the life of the hold. - Auto-follow is not the only thing moving the transcript. The progressive mount's own scroll compensation holds the reader's position across each fill step, and mounting the turn a jump asked for IS a fill step, so it lands after the jump and restores the position the jump just left. That one no seam can fix; it is what the hold is for. Jumps also scroll instantly now, whatever the app's scroll-motion policy says. A jump is a teleport the reader asked for, and an animated one does not survive this surface: traced on the 30-prompt fixture, the smooth scroll was cancelled by the mount's compensation and by the follow spring and stalled two pixels from where it started. Coverage: - The first-click e2e case named the wrong turn (`[data-turn-id]` is the first MOUNTED turn, whose top is already negative at the opening scroll position, so an upper-bound-only check passed without the jump doing anything). It now names `turn-prompt-rail-1`, bounds it on both sides, and asserts that tick's `aria-current`. - `emulateMedia` could not put that case on the production scroll path: `resolveScrollMotionBehavior` collapses motion for ANY fixture, keyed on `data-maka-e2e-fixture` rather than on the media query. Fixtures can now ask for a behavior back (`scrollMotion`, per launch — it costs seconds of settling per window, so only the case that needs it pays), with unit coverage for the precedence: a fixture request never outranks a stated preference for less motion. - `holdJumpDestination`'s unit tests grew the two cases its rewrite is about: it must not settle while the transcript is still filling, and it must hand the transcript back the moment the reader touches it. Verified 5/5 on the smooth-scroll fixture, where the previous revision lost 1 in 4. `quote-selection.spec.ts` flakes on this machine (1 in 4) at upstream/main as well, unchanged by this branch.

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in claiming a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore the two assertions the comment is actually about, plus the ladder order #2348 exists to protect, in ONE file. Split across two they each stayed green against the half they could not see, which is how the stylesheet half got deleted without anything noticing: - the ladder is declared on the --space-* scale in strictly increasing order (retuning a rung is a design call and stays green; list gaps meeting block gaps fails) - every rung declared is actually spent by a rule, so the ladder cannot become decoration while literals do the spacing - the runtime hooks the selectors need are really emitted by MarkdownBody — data-maka-contract, data-density, astryx-markdown-heading, data-level, astryx-list-item. These names have a single upstream owner in Astryx and appear in Maka only inside selectors, so a rename kills every rule at once with nothing failing - compact markdown still has exactly one caller, which is the bet the comment names: the table carries heading TYPOGRAPHY on a density key, and Astryx's own RFC says density must not do that Why these are not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions and CSS-structure suites that asserted how a rule was written. Nothing here asserts an appearance or a rule's shape: three of the four are relations that must survive any retune, and the fourth renders the real component and checks the join between Astryx's output and Maka's selectors. Deliberately left unpinned, because they ARE how the table is written: the adjacent-sibling gap form, the `hr` rung, the ListItem padding reset, and the two heading size tiers. The stylesheet-reading half no longer revives the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read; it reads packages/ui/src/styles.css directly, in the package that owns it. Seven mutations, each verified to fail the assertion that describes it: inverting the ladder; taking a rung off the --space-* grid; hardcoding a rung's usage; hardcoding it while naming the rung in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Also state in the comment what the contract does and does not hold, so the next reader does not have to open the test to find out. Generated-by: Claude Code

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in #2462 and its stylesheet-side sibling in #2425, so the comment has been citing a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore only what cannot be noticed without a test. Both failure modes here are silent: no error, no failing check, just spacing that quietly stops being what the table declares, because a CSS selector that matches nothing never complains. - The table selects entirely on DOM Astryx generates at runtime — data-density, astryx-markdown-heading + data-level, astryx-list-item. Those names have one upstream owner and appear in Maka only inside selectors, so a rename kills every rule at once. Astryx is bumped regularly (0.4.0 in #2983, 0.4.3 in flight, plus the Dependabot minor group), so this is a recurring event, not a hypothetical. - Astryx's ListItem carries CONTROL row padding that `density` cannot reach from outside. That padding is what inverted the ladder in #2348 — list items ~10px apart against 4px paragraphs — and one rule neutralizes it. Lose the rule and the original defect returns, silently. - Plus the assumption the comment actually names: compact markdown still has exactly one caller. The table carries heading TYPOGRAPHY on a density key, which Astryx's own RFC says density must not do, so the key is honest only while `compact` and "transcript" are the same set. Deliberately NOT pinned: the ladder's declared values and their order. A first draft asserted them, and it was the wrong instinct — a reversed ladder has to be typed on purpose into four adjacent lines under a comment explaining the order, and it is visible the moment anyone looks at a transcript. It also would not have caught the defect it claimed to guard: #2348's inversion came from the ListItem padding above, which a test reading only the declared variables cannot see. Same for the adjacent-sibling gap form, the `hr` rung and the two heading size tiers — those are how the table is written, not what it promises. Why this is not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions, and CSS-structure suites that asserted how a rule was written. Two of the three assertions here render the real component and check the join between Astryx's output and Maka's selectors; the third asserts an outcome the table promises, not a syntax. The stylesheet-reading half does not revive the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read — it reads packages/ui/src/styles.css directly, in the package that owns it. Six mutations, each verified to fail the assertion that describes it: deleting the ListItem padding reset; un-zeroing it; un-zeroing it while naming the reset in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Generated-by: Claude Code

jackwener · 2026-08-07T17:38:57Z

Summary

Delete 102 low-value or redundant source test files across Desktop, UI, Runtime, Headless, CLI, and Core, reducing the source test-file surface from 988 to 886 (-10.3%).

This branch is now based on the current main. The cleanup deliberately keeps tests that are the only guard for an active dependency patch or a security-sensitive boundary.

Removed

Desktop markup, copy, presentation, chrome, and overlapping geometry/workbar coverage
UI localization/copy panels, quiet chrome, empty states, and presentation pins
Runtime thin wire and duplicate provider helper coverage
Headless duplicate A/B, toolchain, and prompt-summary coverage
CLI layout/completion coverage and Core copy/locale helpers

Kept after review

Active dependency-patch guards:
- AI SDK tool-call index tracking
- Astryx List accessible-name forwarding
- Astryx new-message indicator behavior
Security-sensitive boundaries:
- split-chunk tool-output secret redaction
- permission escalation confirmation and persistence
- task-ledger prompt injection/redaction wiring
- Responses API store: false
- session-title internal-instruction stripping
Root authority, credentials, storage safety, primary runtime/session/tool suites, and fail-safe CI planning

Additional change

CLI process-contract tests ignore Node's ExperimentalWarning: SQLite stderr noise while continuing to reject other stderr output.

Validation

git diff --check origin/main...HEAD
npm run format:check
npm test — all workspace tests passed after synchronizing with current main

GitHub CI is running on the updated head.

Remove 122 source test files (~12% of packages/apps test files) that were markup/copy contracts, overlapping e2e chrome, duplicate UI presentation pins, and low-signal headless AB/toolchain matrix cases. Keep safety, authority, and primary product unit paths. Also tolerate Node SQLite ExperimentalWarning in CLI run process contracts so stderr-empty assertions are not flaky on modern Node.

Bring back e2e journeys, headless matrix/export, catalog contract, AiSdkFlow, and subscription non-leak coverage that are sole or primary gates. Keep the remaining ~11.6% file-level reduction of true low-value markup/chrome/matrix duplicates.

Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

This reverts commit 5ec1797.

Part of apache#2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until apache#2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds.

Part of #2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until #2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds. Co-authored-by: 陆逊 <211125649+UncertaintyDeterminesYou4ndMe@users.noreply.github.com>

apache#2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — apache#2161 pinned it against a containing block as tall as the conversation, apache#2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in apache#2462, and the multi-prompt fixtures it ran on in apache#2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine.

…2923) * fix(ui): give the prompt rail's tick bars a box again #2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — #2161 pinned it against a containing block as tall as the conversation, #2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in #2462, and the multi-prompt fixtures it ran on in #2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine. * fix(ui): make the prompt rail's hover and jump behave Four things the rail got wrong once it was visible again, found by using it: - A 4px gap between ticks was a band where the pointer was over the rail and over no tick, so the dock-style hover falloff dropped out and picked up again every few pixels of travel. The rail's `gap` moves into the ticks' own `padding-block`: same pitch, hit boxes now tile. - The hover preview waited 300ms before opening — Astryx's HoverCard default, meant for a pointer crossing a wide row on its way somewhere else. A tick is 22px of rail that nothing is on the way to, and the wait is the one part of this hover with no motion in it. Now 120ms. - The highlight glided 280ms to wherever a click landed, so crossing twenty prompts read as the bar flying off across the rail. A click now owns the highlight until its scroll settles: no glide, and the scroll no longer walks the highlight through every prompt it passes. - The first click into a session did nothing until the reader scrolled by hand. See below. That last one is a collision between Astryx's auto-follow lock and the progressive transcript mount, and neither side is wrong on its own. `useChatStreamScroll` unlocks on a scroll up, detected by comparing scrollTop between events — but it ignores any scroll event that arrives with a changed scrollHeight or offsetHeight, because Chrome fires those when content resizes and they are not the reader moving. A jump into an unmounted turn mounts it and the fill that follows changes scrollHeight for several frames, so the jump's own scroll is invisible to the lock: it stays on, and `scrollIfLocked` pulls the transcript back to the bottom. Only a wheel gesture broke it, which takes a separate path in Astryx. `holdJumpDestination` re-aims at the target on each height change until the fill stops. The last of those scrolls lands with a stable height, which is the one the lock finally reads as a scroll up. Measured on the 30-prompt fixture: clicking the first tick went to scrollTop 7042 (the bottom) and now goes to 24 and holds. The fixture grows from 8 prompts to 30 because the progressive mount's initial window is 10 — at 8 the head of the transcript is already mounted and the jump-into-unmounted-turns path never runs at all. Coverage note: the e2e case for the first click is an end-to-end check, not a guard. Whether the lock wins depends on which frame the fill lands on relative to a smooth scroll still in flight, and it goes green against the unfixed renderer often enough to be worthless as one. The guard is the `holdJumpDestination` unit test, which drives the frames itself. * fix(ui): own a rail jump through the mount instead of racing it Review of #2923 found the jump's ownership bound to a clock rather than to the navigation, and the e2e case that was supposed to guard it asserting almost nothing. Both hold. Jump ownership: - A second click during a jump only replaced the target; the first click's 700ms timer still governed, and could clear the second jump mid-flight. Each click now carries its own sequence and starts its own hold. - The fixed window is gone. A hold runs until the progressive mount reports the transcript filled AND nothing has moved for a few frames, so a long transcript is never released mid-fill, and it ends the moment the reader touches the transcript (wheel, touch, pointer, key) rather than outliving their interest in it. Chasing the "just release auto-follow" direction the review preferred found that ChatLayout publishes no such seam, so this adds one — `unlockAutoFollow` on `ChatLayoutContextValue`, exposing the scroll hook's existing `unlock` (patch hunk + patches/README entry). It is necessary and it is not sufficient, which the earlier framing got wrong: - Astryx re-locks on any `scrollend` that settles near the bottom, and a session that opens at the bottom produces exactly that while the mount is still catching up. Releasing once at the click is undone before the jump goes anywhere — traced: released at the click, landed at 154ms, dragged back to the bottom by 166ms. The release is now re-asserted for the life of the hold. - Auto-follow is not the only thing moving the transcript. The progressive mount's own scroll compensation holds the reader's position across each fill step, and mounting the turn a jump asked for IS a fill step, so it lands after the jump and restores the position the jump just left. That one no seam can fix; it is what the hold is for. Jumps also scroll instantly now, whatever the app's scroll-motion policy says. A jump is a teleport the reader asked for, and an animated one does not survive this surface: traced on the 30-prompt fixture, the smooth scroll was cancelled by the mount's compensation and by the follow spring and stalled two pixels from where it started. Coverage: - The first-click e2e case named the wrong turn (`[data-turn-id]` is the first MOUNTED turn, whose top is already negative at the opening scroll position, so an upper-bound-only check passed without the jump doing anything). It now names `turn-prompt-rail-1`, bounds it on both sides, and asserts that tick's `aria-current`. - `emulateMedia` could not put that case on the production scroll path: `resolveScrollMotionBehavior` collapses motion for ANY fixture, keyed on `data-maka-e2e-fixture` rather than on the media query. Fixtures can now ask for a behavior back (`scrollMotion`, per launch — it costs seconds of settling per window, so only the case that needs it pays), with unit coverage for the precedence: a fixture request never outranks a stated preference for less motion. - `holdJumpDestination`'s unit tests grew the two cases its rewrite is about: it must not settle while the transcript is still filling, and it must hand the transcript back the moment the reader touches it. Verified 5/5 on the smooth-scroll fixture, where the previous revision lost 1 in 4. `quote-selection.spec.ts` flakes on this machine (1 in 4) at upstream/main as well, unchanged by this branch.

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in claiming a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore the two assertions the comment is actually about, plus the ladder order #2348 exists to protect, in ONE file. Split across two they each stayed green against the half they could not see, which is how the stylesheet half got deleted without anything noticing: - the ladder is declared on the --space-* scale in strictly increasing order (retuning a rung is a design call and stays green; list gaps meeting block gaps fails) - every rung declared is actually spent by a rule, so the ladder cannot become decoration while literals do the spacing - the runtime hooks the selectors need are really emitted by MarkdownBody — data-maka-contract, data-density, astryx-markdown-heading, data-level, astryx-list-item. These names have a single upstream owner in Astryx and appear in Maka only inside selectors, so a rename kills every rule at once with nothing failing - compact markdown still has exactly one caller, which is the bet the comment names: the table carries heading TYPOGRAPHY on a density key, and Astryx's own RFC says density must not do that Why these are not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions and CSS-structure suites that asserted how a rule was written. Nothing here asserts an appearance or a rule's shape: three of the four are relations that must survive any retune, and the fourth renders the real component and checks the join between Astryx's output and Maka's selectors. Deliberately left unpinned, because they ARE how the table is written: the adjacent-sibling gap form, the `hr` rung, the ListItem padding reset, and the two heading size tiers. The stylesheet-reading half no longer revives the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read; it reads packages/ui/src/styles.css directly, in the package that owns it. Seven mutations, each verified to fail the assertion that describes it: inverting the ladder; taking a rung off the --space-* grid; hardcoding a rung's usage; hardcoding it while naming the rung in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Also state in the comment what the contract does and does not hold, so the next reader does not have to open the test to find out. Generated-by: Claude Code

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in #2462 and its stylesheet-side sibling in #2425, so the comment has been citing a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore only what cannot be noticed without a test. Both failure modes here are silent: no error, no failing check, just spacing that quietly stops being what the table declares, because a CSS selector that matches nothing never complains. - The table selects entirely on DOM Astryx generates at runtime — data-density, astryx-markdown-heading + data-level, astryx-list-item. Those names have one upstream owner and appear in Maka only inside selectors, so a rename kills every rule at once. Astryx is bumped regularly (0.4.0 in #2983, 0.4.3 in flight, plus the Dependabot minor group), so this is a recurring event, not a hypothetical. - Astryx's ListItem carries CONTROL row padding that `density` cannot reach from outside. That padding is what inverted the ladder in #2348 — list items ~10px apart against 4px paragraphs — and one rule neutralizes it. Lose the rule and the original defect returns, silently. - Plus the assumption the comment actually names: compact markdown still has exactly one caller. The table carries heading TYPOGRAPHY on a density key, which Astryx's own RFC says density must not do, so the key is honest only while `compact` and "transcript" are the same set. Deliberately NOT pinned: the ladder's declared values and their order. A first draft asserted them, and it was the wrong instinct — a reversed ladder has to be typed on purpose into four adjacent lines under a comment explaining the order, and it is visible the moment anyone looks at a transcript. It also would not have caught the defect it claimed to guard: #2348's inversion came from the ListItem padding above, which a test reading only the declared variables cannot see. Same for the adjacent-sibling gap form, the `hr` rung and the two heading size tiers — those are how the table is written, not what it promises. Why this is not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions, and CSS-structure suites that asserted how a rule was written. Two of the three assertions here render the real component and check the join between Astryx's output and Maka's selectors; the third asserts an outcome the table promises, not a syntax. The stylesheet-reading half does not revive the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read — it reads packages/ui/src/styles.css directly, in the package that owns it. Six mutations, each verified to fail the assertion that describes it: deleting the ListItem padding reset; un-zeroing it; un-zeroing it while naming the reset in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Generated-by: Claude Code

jackwener · 2026-08-07T17:38:57Z

Summary

Delete 102 low-value or redundant source test files across Desktop, UI, Runtime, Headless, CLI, and Core, reducing the source test-file surface from 988 to 886 (-10.3%).

This branch is now based on the current main. The cleanup deliberately keeps tests that are the only guard for an active dependency patch or a security-sensitive boundary.

Removed

Desktop markup, copy, presentation, chrome, and overlapping geometry/workbar coverage
UI localization/copy panels, quiet chrome, empty states, and presentation pins
Runtime thin wire and duplicate provider helper coverage
Headless duplicate A/B, toolchain, and prompt-summary coverage
CLI layout/completion coverage and Core copy/locale helpers

Kept after review

Active dependency-patch guards:
- AI SDK tool-call index tracking
- Astryx List accessible-name forwarding
- Astryx new-message indicator behavior
Security-sensitive boundaries:
- split-chunk tool-output secret redaction
- permission escalation confirmation and persistence
- task-ledger prompt injection/redaction wiring
- Responses API store: false
- session-title internal-instruction stripping
Root authority, credentials, storage safety, primary runtime/session/tool suites, and fail-safe CI planning

Additional change

CLI process-contract tests ignore Node's ExperimentalWarning: SQLite stderr noise while continuing to reject other stderr output.

Validation

git diff --check origin/main...HEAD
npm run format:check
npm test — all workspace tests passed after synchronizing with current main

GitHub CI is running on the updated head.

Remove 122 source test files (~12% of packages/apps test files) that were markup/copy contracts, overlapping e2e chrome, duplicate UI presentation pins, and low-signal headless AB/toolchain matrix cases. Keep safety, authority, and primary product unit paths. Also tolerate Node SQLite ExperimentalWarning in CLI run process contracts so stderr-empty assertions are not flaky on modern Node.

Bring back e2e journeys, headless matrix/export, catalog contract, AiSdkFlow, and subscription non-leak coverage that are sole or primary gates. Keep the remaining ~11.6% file-level reduction of true low-value markup/chrome/matrix duplicates.

Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

This reverts commit 5ec1797.

Part of apache#2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until apache#2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds.

Part of #2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until #2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds. Co-authored-by: 陆逊 <211125649+UncertaintyDeterminesYou4ndMe@users.noreply.github.com>

apache#2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — apache#2161 pinned it against a containing block as tall as the conversation, apache#2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in apache#2462, and the multi-prompt fixtures it ran on in apache#2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine.

…2923) * fix(ui): give the prompt rail's tick bars a box again #2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — #2161 pinned it against a containing block as tall as the conversation, #2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in #2462, and the multi-prompt fixtures it ran on in #2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine. * fix(ui): make the prompt rail's hover and jump behave Four things the rail got wrong once it was visible again, found by using it: - A 4px gap between ticks was a band where the pointer was over the rail and over no tick, so the dock-style hover falloff dropped out and picked up again every few pixels of travel. The rail's `gap` moves into the ticks' own `padding-block`: same pitch, hit boxes now tile. - The hover preview waited 300ms before opening — Astryx's HoverCard default, meant for a pointer crossing a wide row on its way somewhere else. A tick is 22px of rail that nothing is on the way to, and the wait is the one part of this hover with no motion in it. Now 120ms. - The highlight glided 280ms to wherever a click landed, so crossing twenty prompts read as the bar flying off across the rail. A click now owns the highlight until its scroll settles: no glide, and the scroll no longer walks the highlight through every prompt it passes. - The first click into a session did nothing until the reader scrolled by hand. See below. That last one is a collision between Astryx's auto-follow lock and the progressive transcript mount, and neither side is wrong on its own. `useChatStreamScroll` unlocks on a scroll up, detected by comparing scrollTop between events — but it ignores any scroll event that arrives with a changed scrollHeight or offsetHeight, because Chrome fires those when content resizes and they are not the reader moving. A jump into an unmounted turn mounts it and the fill that follows changes scrollHeight for several frames, so the jump's own scroll is invisible to the lock: it stays on, and `scrollIfLocked` pulls the transcript back to the bottom. Only a wheel gesture broke it, which takes a separate path in Astryx. `holdJumpDestination` re-aims at the target on each height change until the fill stops. The last of those scrolls lands with a stable height, which is the one the lock finally reads as a scroll up. Measured on the 30-prompt fixture: clicking the first tick went to scrollTop 7042 (the bottom) and now goes to 24 and holds. The fixture grows from 8 prompts to 30 because the progressive mount's initial window is 10 — at 8 the head of the transcript is already mounted and the jump-into-unmounted-turns path never runs at all. Coverage note: the e2e case for the first click is an end-to-end check, not a guard. Whether the lock wins depends on which frame the fill lands on relative to a smooth scroll still in flight, and it goes green against the unfixed renderer often enough to be worthless as one. The guard is the `holdJumpDestination` unit test, which drives the frames itself. * fix(ui): own a rail jump through the mount instead of racing it Review of #2923 found the jump's ownership bound to a clock rather than to the navigation, and the e2e case that was supposed to guard it asserting almost nothing. Both hold. Jump ownership: - A second click during a jump only replaced the target; the first click's 700ms timer still governed, and could clear the second jump mid-flight. Each click now carries its own sequence and starts its own hold. - The fixed window is gone. A hold runs until the progressive mount reports the transcript filled AND nothing has moved for a few frames, so a long transcript is never released mid-fill, and it ends the moment the reader touches the transcript (wheel, touch, pointer, key) rather than outliving their interest in it. Chasing the "just release auto-follow" direction the review preferred found that ChatLayout publishes no such seam, so this adds one — `unlockAutoFollow` on `ChatLayoutContextValue`, exposing the scroll hook's existing `unlock` (patch hunk + patches/README entry). It is necessary and it is not sufficient, which the earlier framing got wrong: - Astryx re-locks on any `scrollend` that settles near the bottom, and a session that opens at the bottom produces exactly that while the mount is still catching up. Releasing once at the click is undone before the jump goes anywhere — traced: released at the click, landed at 154ms, dragged back to the bottom by 166ms. The release is now re-asserted for the life of the hold. - Auto-follow is not the only thing moving the transcript. The progressive mount's own scroll compensation holds the reader's position across each fill step, and mounting the turn a jump asked for IS a fill step, so it lands after the jump and restores the position the jump just left. That one no seam can fix; it is what the hold is for. Jumps also scroll instantly now, whatever the app's scroll-motion policy says. A jump is a teleport the reader asked for, and an animated one does not survive this surface: traced on the 30-prompt fixture, the smooth scroll was cancelled by the mount's compensation and by the follow spring and stalled two pixels from where it started. Coverage: - The first-click e2e case named the wrong turn (`[data-turn-id]` is the first MOUNTED turn, whose top is already negative at the opening scroll position, so an upper-bound-only check passed without the jump doing anything). It now names `turn-prompt-rail-1`, bounds it on both sides, and asserts that tick's `aria-current`. - `emulateMedia` could not put that case on the production scroll path: `resolveScrollMotionBehavior` collapses motion for ANY fixture, keyed on `data-maka-e2e-fixture` rather than on the media query. Fixtures can now ask for a behavior back (`scrollMotion`, per launch — it costs seconds of settling per window, so only the case that needs it pays), with unit coverage for the precedence: a fixture request never outranks a stated preference for less motion. - `holdJumpDestination`'s unit tests grew the two cases its rewrite is about: it must not settle while the transcript is still filling, and it must hand the transcript back the moment the reader touches it. Verified 5/5 on the smooth-scroll fixture, where the previous revision lost 1 in 4. `quote-selection.spec.ts` flakes on this machine (1 in 4) at upstream/main as well, unchanged by this branch.

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in claiming a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore the two assertions the comment is actually about, plus the ladder order #2348 exists to protect, in ONE file. Split across two they each stayed green against the half they could not see, which is how the stylesheet half got deleted without anything noticing: - the ladder is declared on the --space-* scale in strictly increasing order (retuning a rung is a design call and stays green; list gaps meeting block gaps fails) - every rung declared is actually spent by a rule, so the ladder cannot become decoration while literals do the spacing - the runtime hooks the selectors need are really emitted by MarkdownBody — data-maka-contract, data-density, astryx-markdown-heading, data-level, astryx-list-item. These names have a single upstream owner in Astryx and appear in Maka only inside selectors, so a rename kills every rule at once with nothing failing - compact markdown still has exactly one caller, which is the bet the comment names: the table carries heading TYPOGRAPHY on a density key, and Astryx's own RFC says density must not do that Why these are not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions and CSS-structure suites that asserted how a rule was written. Nothing here asserts an appearance or a rule's shape: three of the four are relations that must survive any retune, and the fourth renders the real component and checks the join between Astryx's output and Maka's selectors. Deliberately left unpinned, because they ARE how the table is written: the adjacent-sibling gap form, the `hr` rung, the ListItem padding reset, and the two heading size tiers. The stylesheet-reading half no longer revives the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read; it reads packages/ui/src/styles.css directly, in the package that owns it. Seven mutations, each verified to fail the assertion that describes it: inverting the ladder; taking a rung off the --space-* grid; hardcoding a rung's usage; hardcoding it while naming the rung in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Also state in the comment what the contract does and does not hold, so the next reader does not have to open the test to find out. Generated-by: Claude Code

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in #2462 and its stylesheet-side sibling in #2425, so the comment has been citing a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore only what cannot be noticed without a test. Both failure modes here are silent: no error, no failing check, just spacing that quietly stops being what the table declares, because a CSS selector that matches nothing never complains. - The table selects entirely on DOM Astryx generates at runtime — data-density, astryx-markdown-heading + data-level, astryx-list-item. Those names have one upstream owner and appear in Maka only inside selectors, so a rename kills every rule at once. Astryx is bumped regularly (0.4.0 in #2983, 0.4.3 in flight, plus the Dependabot minor group), so this is a recurring event, not a hypothetical. - Astryx's ListItem carries CONTROL row padding that `density` cannot reach from outside. That padding is what inverted the ladder in #2348 — list items ~10px apart against 4px paragraphs — and one rule neutralizes it. Lose the rule and the original defect returns, silently. - Plus the assumption the comment actually names: compact markdown still has exactly one caller. The table carries heading TYPOGRAPHY on a density key, which Astryx's own RFC says density must not do, so the key is honest only while `compact` and "transcript" are the same set. Deliberately NOT pinned: the ladder's declared values and their order. A first draft asserted them, and it was the wrong instinct — a reversed ladder has to be typed on purpose into four adjacent lines under a comment explaining the order, and it is visible the moment anyone looks at a transcript. It also would not have caught the defect it claimed to guard: #2348's inversion came from the ListItem padding above, which a test reading only the declared variables cannot see. Same for the adjacent-sibling gap form, the `hr` rung and the two heading size tiers — those are how the table is written, not what it promises. Why this is not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions, and CSS-structure suites that asserted how a rule was written. Two of the three assertions here render the real component and check the join between Astryx's output and Maka's selectors; the third asserts an outcome the table promises, not a syntax. The stylesheet-reading half does not revive the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read — it reads packages/ui/src/styles.css directly, in the package that owns it. Six mutations, each verified to fail the assertion that describes it: deleting the ListItem padding reset; un-zeroing it; un-zeroing it while naming the reset in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Generated-by: Claude Code

jackwener · 2026-08-07T17:38:57Z

Summary

Delete 102 low-value or redundant source test files across Desktop, UI, Runtime, Headless, CLI, and Core, reducing the source test-file surface from 988 to 886 (-10.3%).

This branch is now based on the current main. The cleanup deliberately keeps tests that are the only guard for an active dependency patch or a security-sensitive boundary.

Removed

Desktop markup, copy, presentation, chrome, and overlapping geometry/workbar coverage
UI localization/copy panels, quiet chrome, empty states, and presentation pins
Runtime thin wire and duplicate provider helper coverage
Headless duplicate A/B, toolchain, and prompt-summary coverage
CLI layout/completion coverage and Core copy/locale helpers

Kept after review

Active dependency-patch guards:
- AI SDK tool-call index tracking
- Astryx List accessible-name forwarding
- Astryx new-message indicator behavior
Security-sensitive boundaries:
- split-chunk tool-output secret redaction
- permission escalation confirmation and persistence
- task-ledger prompt injection/redaction wiring
- Responses API store: false
- session-title internal-instruction stripping
Root authority, credentials, storage safety, primary runtime/session/tool suites, and fail-safe CI planning

Additional change

CLI process-contract tests ignore Node's ExperimentalWarning: SQLite stderr noise while continuing to reject other stderr output.

Validation

git diff --check origin/main...HEAD
npm run format:check
npm test — all workspace tests passed after synchronizing with current main

GitHub CI is running on the updated head.

Remove 122 source test files (~12% of packages/apps test files) that were markup/copy contracts, overlapping e2e chrome, duplicate UI presentation pins, and low-signal headless AB/toolchain matrix cases. Keep safety, authority, and primary product unit paths. Also tolerate Node SQLite ExperimentalWarning in CLI run process contracts so stderr-empty assertions are not flaky on modern Node.

Bring back e2e journeys, headless matrix/export, catalog contract, AiSdkFlow, and subscription non-leak coverage that are sole or primary gates. Keep the remaining ~11.6% file-level reduction of true low-value markup/chrome/matrix duplicates.

Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

This reverts commit 5ec1797.

Part of apache#2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until apache#2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds.

Part of #2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until #2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds. Co-authored-by: 陆逊 <211125649+UncertaintyDeterminesYou4ndMe@users.noreply.github.com>

apache#2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — apache#2161 pinned it against a containing block as tall as the conversation, apache#2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in apache#2462, and the multi-prompt fixtures it ran on in apache#2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine.

…2923) * fix(ui): give the prompt rail's tick bars a box again #2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — #2161 pinned it against a containing block as tall as the conversation, #2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in #2462, and the multi-prompt fixtures it ran on in #2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine. * fix(ui): make the prompt rail's hover and jump behave Four things the rail got wrong once it was visible again, found by using it: - A 4px gap between ticks was a band where the pointer was over the rail and over no tick, so the dock-style hover falloff dropped out and picked up again every few pixels of travel. The rail's `gap` moves into the ticks' own `padding-block`: same pitch, hit boxes now tile. - The hover preview waited 300ms before opening — Astryx's HoverCard default, meant for a pointer crossing a wide row on its way somewhere else. A tick is 22px of rail that nothing is on the way to, and the wait is the one part of this hover with no motion in it. Now 120ms. - The highlight glided 280ms to wherever a click landed, so crossing twenty prompts read as the bar flying off across the rail. A click now owns the highlight until its scroll settles: no glide, and the scroll no longer walks the highlight through every prompt it passes. - The first click into a session did nothing until the reader scrolled by hand. See below. That last one is a collision between Astryx's auto-follow lock and the progressive transcript mount, and neither side is wrong on its own. `useChatStreamScroll` unlocks on a scroll up, detected by comparing scrollTop between events — but it ignores any scroll event that arrives with a changed scrollHeight or offsetHeight, because Chrome fires those when content resizes and they are not the reader moving. A jump into an unmounted turn mounts it and the fill that follows changes scrollHeight for several frames, so the jump's own scroll is invisible to the lock: it stays on, and `scrollIfLocked` pulls the transcript back to the bottom. Only a wheel gesture broke it, which takes a separate path in Astryx. `holdJumpDestination` re-aims at the target on each height change until the fill stops. The last of those scrolls lands with a stable height, which is the one the lock finally reads as a scroll up. Measured on the 30-prompt fixture: clicking the first tick went to scrollTop 7042 (the bottom) and now goes to 24 and holds. The fixture grows from 8 prompts to 30 because the progressive mount's initial window is 10 — at 8 the head of the transcript is already mounted and the jump-into-unmounted-turns path never runs at all. Coverage note: the e2e case for the first click is an end-to-end check, not a guard. Whether the lock wins depends on which frame the fill lands on relative to a smooth scroll still in flight, and it goes green against the unfixed renderer often enough to be worthless as one. The guard is the `holdJumpDestination` unit test, which drives the frames itself. * fix(ui): own a rail jump through the mount instead of racing it Review of #2923 found the jump's ownership bound to a clock rather than to the navigation, and the e2e case that was supposed to guard it asserting almost nothing. Both hold. Jump ownership: - A second click during a jump only replaced the target; the first click's 700ms timer still governed, and could clear the second jump mid-flight. Each click now carries its own sequence and starts its own hold. - The fixed window is gone. A hold runs until the progressive mount reports the transcript filled AND nothing has moved for a few frames, so a long transcript is never released mid-fill, and it ends the moment the reader touches the transcript (wheel, touch, pointer, key) rather than outliving their interest in it. Chasing the "just release auto-follow" direction the review preferred found that ChatLayout publishes no such seam, so this adds one — `unlockAutoFollow` on `ChatLayoutContextValue`, exposing the scroll hook's existing `unlock` (patch hunk + patches/README entry). It is necessary and it is not sufficient, which the earlier framing got wrong: - Astryx re-locks on any `scrollend` that settles near the bottom, and a session that opens at the bottom produces exactly that while the mount is still catching up. Releasing once at the click is undone before the jump goes anywhere — traced: released at the click, landed at 154ms, dragged back to the bottom by 166ms. The release is now re-asserted for the life of the hold. - Auto-follow is not the only thing moving the transcript. The progressive mount's own scroll compensation holds the reader's position across each fill step, and mounting the turn a jump asked for IS a fill step, so it lands after the jump and restores the position the jump just left. That one no seam can fix; it is what the hold is for. Jumps also scroll instantly now, whatever the app's scroll-motion policy says. A jump is a teleport the reader asked for, and an animated one does not survive this surface: traced on the 30-prompt fixture, the smooth scroll was cancelled by the mount's compensation and by the follow spring and stalled two pixels from where it started. Coverage: - The first-click e2e case named the wrong turn (`[data-turn-id]` is the first MOUNTED turn, whose top is already negative at the opening scroll position, so an upper-bound-only check passed without the jump doing anything). It now names `turn-prompt-rail-1`, bounds it on both sides, and asserts that tick's `aria-current`. - `emulateMedia` could not put that case on the production scroll path: `resolveScrollMotionBehavior` collapses motion for ANY fixture, keyed on `data-maka-e2e-fixture` rather than on the media query. Fixtures can now ask for a behavior back (`scrollMotion`, per launch — it costs seconds of settling per window, so only the case that needs it pays), with unit coverage for the precedence: a fixture request never outranks a stated preference for less motion. - `holdJumpDestination`'s unit tests grew the two cases its rewrite is about: it must not settle while the transcript is still filling, and it must hand the transcript back the moment the reader touches it. Verified 5/5 on the smooth-scroll fixture, where the previous revision lost 1 in 4. `quote-selection.spec.ts` flakes on this machine (1 in 4) at upstream/main as well, unchanged by this branch.

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in claiming a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore the two assertions the comment is actually about, plus the ladder order #2348 exists to protect, in ONE file. Split across two they each stayed green against the half they could not see, which is how the stylesheet half got deleted without anything noticing: - the ladder is declared on the --space-* scale in strictly increasing order (retuning a rung is a design call and stays green; list gaps meeting block gaps fails) - every rung declared is actually spent by a rule, so the ladder cannot become decoration while literals do the spacing - the runtime hooks the selectors need are really emitted by MarkdownBody — data-maka-contract, data-density, astryx-markdown-heading, data-level, astryx-list-item. These names have a single upstream owner in Astryx and appear in Maka only inside selectors, so a rename kills every rule at once with nothing failing - compact markdown still has exactly one caller, which is the bet the comment names: the table carries heading TYPOGRAPHY on a density key, and Astryx's own RFC says density must not do that Why these are not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions and CSS-structure suites that asserted how a rule was written. Nothing here asserts an appearance or a rule's shape: three of the four are relations that must survive any retune, and the fourth renders the real component and checks the join between Astryx's output and Maka's selectors. Deliberately left unpinned, because they ARE how the table is written: the adjacent-sibling gap form, the `hr` rung, the ListItem padding reset, and the two heading size tiers. The stylesheet-reading half no longer revives the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read; it reads packages/ui/src/styles.css directly, in the package that owns it. Seven mutations, each verified to fail the assertion that describes it: inverting the ladder; taking a rung off the --space-* grid; hardcoding a rung's usage; hardcoding it while naming the rung in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Also state in the comment what the contract does and does not hold, so the next reader does not have to open the test to find out. Generated-by: Claude Code

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in #2462 and its stylesheet-side sibling in #2425, so the comment has been citing a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore only what cannot be noticed without a test. Both failure modes here are silent: no error, no failing check, just spacing that quietly stops being what the table declares, because a CSS selector that matches nothing never complains. - The table selects entirely on DOM Astryx generates at runtime — data-density, astryx-markdown-heading + data-level, astryx-list-item. Those names have one upstream owner and appear in Maka only inside selectors, so a rename kills every rule at once. Astryx is bumped regularly (0.4.0 in #2983, 0.4.3 in flight, plus the Dependabot minor group), so this is a recurring event, not a hypothetical. - Astryx's ListItem carries CONTROL row padding that `density` cannot reach from outside. That padding is what inverted the ladder in #2348 — list items ~10px apart against 4px paragraphs — and one rule neutralizes it. Lose the rule and the original defect returns, silently. - Plus the assumption the comment actually names: compact markdown still has exactly one caller. The table carries heading TYPOGRAPHY on a density key, which Astryx's own RFC says density must not do, so the key is honest only while `compact` and "transcript" are the same set. Deliberately NOT pinned: the ladder's declared values and their order. A first draft asserted them, and it was the wrong instinct — a reversed ladder has to be typed on purpose into four adjacent lines under a comment explaining the order, and it is visible the moment anyone looks at a transcript. It also would not have caught the defect it claimed to guard: #2348's inversion came from the ListItem padding above, which a test reading only the declared variables cannot see. Same for the adjacent-sibling gap form, the `hr` rung and the two heading size tiers — those are how the table is written, not what it promises. Why this is not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions, and CSS-structure suites that asserted how a rule was written. Two of the three assertions here render the real component and check the join between Astryx's output and Maka's selectors; the third asserts an outcome the table promises, not a syntax. The stylesheet-reading half does not revive the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read — it reads packages/ui/src/styles.css directly, in the package that owns it. Six mutations, each verified to fail the assertion that describes it: deleting the ListItem padding reset; un-zeroing it; un-zeroing it while naming the reset in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Generated-by: Claude Code

jackwener · 2026-08-07T17:38:57Z

Summary

Delete 102 low-value or redundant source test files across Desktop, UI, Runtime, Headless, CLI, and Core, reducing the source test-file surface from 988 to 886 (-10.3%).

This branch is now based on the current main. The cleanup deliberately keeps tests that are the only guard for an active dependency patch or a security-sensitive boundary.

Removed

Desktop markup, copy, presentation, chrome, and overlapping geometry/workbar coverage
UI localization/copy panels, quiet chrome, empty states, and presentation pins
Runtime thin wire and duplicate provider helper coverage
Headless duplicate A/B, toolchain, and prompt-summary coverage
CLI layout/completion coverage and Core copy/locale helpers

Kept after review

Active dependency-patch guards:
- AI SDK tool-call index tracking
- Astryx List accessible-name forwarding
- Astryx new-message indicator behavior
Security-sensitive boundaries:
- split-chunk tool-output secret redaction
- permission escalation confirmation and persistence
- task-ledger prompt injection/redaction wiring
- Responses API store: false
- session-title internal-instruction stripping
Root authority, credentials, storage safety, primary runtime/session/tool suites, and fail-safe CI planning

Additional change

CLI process-contract tests ignore Node's ExperimentalWarning: SQLite stderr noise while continuing to reject other stderr output.

Validation

git diff --check origin/main...HEAD
npm run format:check
npm test — all workspace tests passed after synchronizing with current main

GitHub CI is running on the updated head.

Remove 122 source test files (~12% of packages/apps test files) that were markup/copy contracts, overlapping e2e chrome, duplicate UI presentation pins, and low-signal headless AB/toolchain matrix cases. Keep safety, authority, and primary product unit paths. Also tolerate Node SQLite ExperimentalWarning in CLI run process contracts so stderr-empty assertions are not flaky on modern Node.

Bring back e2e journeys, headless matrix/export, catalog contract, AiSdkFlow, and subscription non-leak coverage that are sole or primary gates. Keep the remaining ~11.6% file-level reduction of true low-value markup/chrome/matrix duplicates.

Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

This reverts commit 5ec1797.

Part of apache#2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until apache#2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds.

Part of #2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until #2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds. Co-authored-by: 陆逊 <211125649+UncertaintyDeterminesYou4ndMe@users.noreply.github.com>

apache#2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — apache#2161 pinned it against a containing block as tall as the conversation, apache#2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in apache#2462, and the multi-prompt fixtures it ran on in apache#2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine.

…2923) * fix(ui): give the prompt rail's tick bars a box again #2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — #2161 pinned it against a containing block as tall as the conversation, #2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in #2462, and the multi-prompt fixtures it ran on in #2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine. * fix(ui): make the prompt rail's hover and jump behave Four things the rail got wrong once it was visible again, found by using it: - A 4px gap between ticks was a band where the pointer was over the rail and over no tick, so the dock-style hover falloff dropped out and picked up again every few pixels of travel. The rail's `gap` moves into the ticks' own `padding-block`: same pitch, hit boxes now tile. - The hover preview waited 300ms before opening — Astryx's HoverCard default, meant for a pointer crossing a wide row on its way somewhere else. A tick is 22px of rail that nothing is on the way to, and the wait is the one part of this hover with no motion in it. Now 120ms. - The highlight glided 280ms to wherever a click landed, so crossing twenty prompts read as the bar flying off across the rail. A click now owns the highlight until its scroll settles: no glide, and the scroll no longer walks the highlight through every prompt it passes. - The first click into a session did nothing until the reader scrolled by hand. See below. That last one is a collision between Astryx's auto-follow lock and the progressive transcript mount, and neither side is wrong on its own. `useChatStreamScroll` unlocks on a scroll up, detected by comparing scrollTop between events — but it ignores any scroll event that arrives with a changed scrollHeight or offsetHeight, because Chrome fires those when content resizes and they are not the reader moving. A jump into an unmounted turn mounts it and the fill that follows changes scrollHeight for several frames, so the jump's own scroll is invisible to the lock: it stays on, and `scrollIfLocked` pulls the transcript back to the bottom. Only a wheel gesture broke it, which takes a separate path in Astryx. `holdJumpDestination` re-aims at the target on each height change until the fill stops. The last of those scrolls lands with a stable height, which is the one the lock finally reads as a scroll up. Measured on the 30-prompt fixture: clicking the first tick went to scrollTop 7042 (the bottom) and now goes to 24 and holds. The fixture grows from 8 prompts to 30 because the progressive mount's initial window is 10 — at 8 the head of the transcript is already mounted and the jump-into-unmounted-turns path never runs at all. Coverage note: the e2e case for the first click is an end-to-end check, not a guard. Whether the lock wins depends on which frame the fill lands on relative to a smooth scroll still in flight, and it goes green against the unfixed renderer often enough to be worthless as one. The guard is the `holdJumpDestination` unit test, which drives the frames itself. * fix(ui): own a rail jump through the mount instead of racing it Review of #2923 found the jump's ownership bound to a clock rather than to the navigation, and the e2e case that was supposed to guard it asserting almost nothing. Both hold. Jump ownership: - A second click during a jump only replaced the target; the first click's 700ms timer still governed, and could clear the second jump mid-flight. Each click now carries its own sequence and starts its own hold. - The fixed window is gone. A hold runs until the progressive mount reports the transcript filled AND nothing has moved for a few frames, so a long transcript is never released mid-fill, and it ends the moment the reader touches the transcript (wheel, touch, pointer, key) rather than outliving their interest in it. Chasing the "just release auto-follow" direction the review preferred found that ChatLayout publishes no such seam, so this adds one — `unlockAutoFollow` on `ChatLayoutContextValue`, exposing the scroll hook's existing `unlock` (patch hunk + patches/README entry). It is necessary and it is not sufficient, which the earlier framing got wrong: - Astryx re-locks on any `scrollend` that settles near the bottom, and a session that opens at the bottom produces exactly that while the mount is still catching up. Releasing once at the click is undone before the jump goes anywhere — traced: released at the click, landed at 154ms, dragged back to the bottom by 166ms. The release is now re-asserted for the life of the hold. - Auto-follow is not the only thing moving the transcript. The progressive mount's own scroll compensation holds the reader's position across each fill step, and mounting the turn a jump asked for IS a fill step, so it lands after the jump and restores the position the jump just left. That one no seam can fix; it is what the hold is for. Jumps also scroll instantly now, whatever the app's scroll-motion policy says. A jump is a teleport the reader asked for, and an animated one does not survive this surface: traced on the 30-prompt fixture, the smooth scroll was cancelled by the mount's compensation and by the follow spring and stalled two pixels from where it started. Coverage: - The first-click e2e case named the wrong turn (`[data-turn-id]` is the first MOUNTED turn, whose top is already negative at the opening scroll position, so an upper-bound-only check passed without the jump doing anything). It now names `turn-prompt-rail-1`, bounds it on both sides, and asserts that tick's `aria-current`. - `emulateMedia` could not put that case on the production scroll path: `resolveScrollMotionBehavior` collapses motion for ANY fixture, keyed on `data-maka-e2e-fixture` rather than on the media query. Fixtures can now ask for a behavior back (`scrollMotion`, per launch — it costs seconds of settling per window, so only the case that needs it pays), with unit coverage for the precedence: a fixture request never outranks a stated preference for less motion. - `holdJumpDestination`'s unit tests grew the two cases its rewrite is about: it must not settle while the transcript is still filling, and it must hand the transcript back the moment the reader touches it. Verified 5/5 on the smooth-scroll fixture, where the previous revision lost 1 in 4. `quote-selection.spec.ts` flakes on this machine (1 in 4) at upstream/main as well, unchanged by this branch.

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in claiming a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore the two assertions the comment is actually about, plus the ladder order #2348 exists to protect, in ONE file. Split across two they each stayed green against the half they could not see, which is how the stylesheet half got deleted without anything noticing: - the ladder is declared on the --space-* scale in strictly increasing order (retuning a rung is a design call and stays green; list gaps meeting block gaps fails) - every rung declared is actually spent by a rule, so the ladder cannot become decoration while literals do the spacing - the runtime hooks the selectors need are really emitted by MarkdownBody — data-maka-contract, data-density, astryx-markdown-heading, data-level, astryx-list-item. These names have a single upstream owner in Astryx and appear in Maka only inside selectors, so a rename kills every rule at once with nothing failing - compact markdown still has exactly one caller, which is the bet the comment names: the table carries heading TYPOGRAPHY on a density key, and Astryx's own RFC says density must not do that Why these are not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions and CSS-structure suites that asserted how a rule was written. Nothing here asserts an appearance or a rule's shape: three of the four are relations that must survive any retune, and the fourth renders the real component and checks the join between Astryx's output and Maka's selectors. Deliberately left unpinned, because they ARE how the table is written: the adjacent-sibling gap form, the `hr` rung, the ListItem padding reset, and the two heading size tiers. The stylesheet-reading half no longer revives the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read; it reads packages/ui/src/styles.css directly, in the package that owns it. Seven mutations, each verified to fail the assertion that describes it: inverting the ladder; taking a rung off the --space-* grid; hardcoding a rung's usage; hardcoding it while naming the rung in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Also state in the comment what the contract does and does not hold, so the next reader does not have to open the test to find out. Generated-by: Claude Code

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in #2462 and its stylesheet-side sibling in #2425, so the comment has been citing a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore only what cannot be noticed without a test. Both failure modes here are silent: no error, no failing check, just spacing that quietly stops being what the table declares, because a CSS selector that matches nothing never complains. - The table selects entirely on DOM Astryx generates at runtime — data-density, astryx-markdown-heading + data-level, astryx-list-item. Those names have one upstream owner and appear in Maka only inside selectors, so a rename kills every rule at once. Astryx is bumped regularly (0.4.0 in #2983, 0.4.3 in flight, plus the Dependabot minor group), so this is a recurring event, not a hypothetical. - Astryx's ListItem carries CONTROL row padding that `density` cannot reach from outside. That padding is what inverted the ladder in #2348 — list items ~10px apart against 4px paragraphs — and one rule neutralizes it. Lose the rule and the original defect returns, silently. - Plus the assumption the comment actually names: compact markdown still has exactly one caller. The table carries heading TYPOGRAPHY on a density key, which Astryx's own RFC says density must not do, so the key is honest only while `compact` and "transcript" are the same set. Deliberately NOT pinned: the ladder's declared values and their order. A first draft asserted them, and it was the wrong instinct — a reversed ladder has to be typed on purpose into four adjacent lines under a comment explaining the order, and it is visible the moment anyone looks at a transcript. It also would not have caught the defect it claimed to guard: #2348's inversion came from the ListItem padding above, which a test reading only the declared variables cannot see. Same for the adjacent-sibling gap form, the `hr` rung and the two heading size tiers — those are how the table is written, not what it promises. Why this is not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions, and CSS-structure suites that asserted how a rule was written. Two of the three assertions here render the real component and check the join between Astryx's output and Maka's selectors; the third asserts an outcome the table promises, not a syntax. The stylesheet-reading half does not revive the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read — it reads packages/ui/src/styles.css directly, in the package that owns it. Six mutations, each verified to fail the assertion that describes it: deleting the ListItem padding reset; un-zeroing it; un-zeroing it while naming the reset in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Generated-by: Claude Code

jackwener · 2026-08-07T17:38:57Z

Summary

Delete 102 low-value or redundant source test files across Desktop, UI, Runtime, Headless, CLI, and Core, reducing the source test-file surface from 988 to 886 (-10.3%).

This branch is now based on the current main. The cleanup deliberately keeps tests that are the only guard for an active dependency patch or a security-sensitive boundary.

Removed

Desktop markup, copy, presentation, chrome, and overlapping geometry/workbar coverage
UI localization/copy panels, quiet chrome, empty states, and presentation pins
Runtime thin wire and duplicate provider helper coverage
Headless duplicate A/B, toolchain, and prompt-summary coverage
CLI layout/completion coverage and Core copy/locale helpers

Kept after review

Active dependency-patch guards:
- AI SDK tool-call index tracking
- Astryx List accessible-name forwarding
- Astryx new-message indicator behavior
Security-sensitive boundaries:
- split-chunk tool-output secret redaction
- permission escalation confirmation and persistence
- task-ledger prompt injection/redaction wiring
- Responses API store: false
- session-title internal-instruction stripping
Root authority, credentials, storage safety, primary runtime/session/tool suites, and fail-safe CI planning

Additional change

CLI process-contract tests ignore Node's ExperimentalWarning: SQLite stderr noise while continuing to reject other stderr output.

Validation

git diff --check origin/main...HEAD
npm run format:check
npm test — all workspace tests passed after synchronizing with current main

GitHub CI is running on the updated head.

Remove 122 source test files (~12% of packages/apps test files) that were markup/copy contracts, overlapping e2e chrome, duplicate UI presentation pins, and low-signal headless AB/toolchain matrix cases. Keep safety, authority, and primary product unit paths. Also tolerate Node SQLite ExperimentalWarning in CLI run process contracts so stderr-empty assertions are not flaky on modern Node.

Bring back e2e journeys, headless matrix/export, catalog contract, AiSdkFlow, and subscription non-leak coverage that are sole or primary gates. Keep the remaining ~11.6% file-level reduction of true low-value markup/chrome/matrix duplicates.

Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

This reverts commit 5ec1797.

Part of apache#2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until apache#2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds.

Part of #2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until #2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds. Co-authored-by: 陆逊 <211125649+UncertaintyDeterminesYou4ndMe@users.noreply.github.com>

apache#2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — apache#2161 pinned it against a containing block as tall as the conversation, apache#2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in apache#2462, and the multi-prompt fixtures it ran on in apache#2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine.

…2923) * fix(ui): give the prompt rail's tick bars a box again #2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — #2161 pinned it against a containing block as tall as the conversation, #2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in #2462, and the multi-prompt fixtures it ran on in #2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine. * fix(ui): make the prompt rail's hover and jump behave Four things the rail got wrong once it was visible again, found by using it: - A 4px gap between ticks was a band where the pointer was over the rail and over no tick, so the dock-style hover falloff dropped out and picked up again every few pixels of travel. The rail's `gap` moves into the ticks' own `padding-block`: same pitch, hit boxes now tile. - The hover preview waited 300ms before opening — Astryx's HoverCard default, meant for a pointer crossing a wide row on its way somewhere else. A tick is 22px of rail that nothing is on the way to, and the wait is the one part of this hover with no motion in it. Now 120ms. - The highlight glided 280ms to wherever a click landed, so crossing twenty prompts read as the bar flying off across the rail. A click now owns the highlight until its scroll settles: no glide, and the scroll no longer walks the highlight through every prompt it passes. - The first click into a session did nothing until the reader scrolled by hand. See below. That last one is a collision between Astryx's auto-follow lock and the progressive transcript mount, and neither side is wrong on its own. `useChatStreamScroll` unlocks on a scroll up, detected by comparing scrollTop between events — but it ignores any scroll event that arrives with a changed scrollHeight or offsetHeight, because Chrome fires those when content resizes and they are not the reader moving. A jump into an unmounted turn mounts it and the fill that follows changes scrollHeight for several frames, so the jump's own scroll is invisible to the lock: it stays on, and `scrollIfLocked` pulls the transcript back to the bottom. Only a wheel gesture broke it, which takes a separate path in Astryx. `holdJumpDestination` re-aims at the target on each height change until the fill stops. The last of those scrolls lands with a stable height, which is the one the lock finally reads as a scroll up. Measured on the 30-prompt fixture: clicking the first tick went to scrollTop 7042 (the bottom) and now goes to 24 and holds. The fixture grows from 8 prompts to 30 because the progressive mount's initial window is 10 — at 8 the head of the transcript is already mounted and the jump-into-unmounted-turns path never runs at all. Coverage note: the e2e case for the first click is an end-to-end check, not a guard. Whether the lock wins depends on which frame the fill lands on relative to a smooth scroll still in flight, and it goes green against the unfixed renderer often enough to be worthless as one. The guard is the `holdJumpDestination` unit test, which drives the frames itself. * fix(ui): own a rail jump through the mount instead of racing it Review of #2923 found the jump's ownership bound to a clock rather than to the navigation, and the e2e case that was supposed to guard it asserting almost nothing. Both hold. Jump ownership: - A second click during a jump only replaced the target; the first click's 700ms timer still governed, and could clear the second jump mid-flight. Each click now carries its own sequence and starts its own hold. - The fixed window is gone. A hold runs until the progressive mount reports the transcript filled AND nothing has moved for a few frames, so a long transcript is never released mid-fill, and it ends the moment the reader touches the transcript (wheel, touch, pointer, key) rather than outliving their interest in it. Chasing the "just release auto-follow" direction the review preferred found that ChatLayout publishes no such seam, so this adds one — `unlockAutoFollow` on `ChatLayoutContextValue`, exposing the scroll hook's existing `unlock` (patch hunk + patches/README entry). It is necessary and it is not sufficient, which the earlier framing got wrong: - Astryx re-locks on any `scrollend` that settles near the bottom, and a session that opens at the bottom produces exactly that while the mount is still catching up. Releasing once at the click is undone before the jump goes anywhere — traced: released at the click, landed at 154ms, dragged back to the bottom by 166ms. The release is now re-asserted for the life of the hold. - Auto-follow is not the only thing moving the transcript. The progressive mount's own scroll compensation holds the reader's position across each fill step, and mounting the turn a jump asked for IS a fill step, so it lands after the jump and restores the position the jump just left. That one no seam can fix; it is what the hold is for. Jumps also scroll instantly now, whatever the app's scroll-motion policy says. A jump is a teleport the reader asked for, and an animated one does not survive this surface: traced on the 30-prompt fixture, the smooth scroll was cancelled by the mount's compensation and by the follow spring and stalled two pixels from where it started. Coverage: - The first-click e2e case named the wrong turn (`[data-turn-id]` is the first MOUNTED turn, whose top is already negative at the opening scroll position, so an upper-bound-only check passed without the jump doing anything). It now names `turn-prompt-rail-1`, bounds it on both sides, and asserts that tick's `aria-current`. - `emulateMedia` could not put that case on the production scroll path: `resolveScrollMotionBehavior` collapses motion for ANY fixture, keyed on `data-maka-e2e-fixture` rather than on the media query. Fixtures can now ask for a behavior back (`scrollMotion`, per launch — it costs seconds of settling per window, so only the case that needs it pays), with unit coverage for the precedence: a fixture request never outranks a stated preference for less motion. - `holdJumpDestination`'s unit tests grew the two cases its rewrite is about: it must not settle while the transcript is still filling, and it must hand the transcript back the moment the reader touches it. Verified 5/5 on the smooth-scroll fixture, where the previous revision lost 1 in 4. `quote-selection.spec.ts` flakes on this machine (1 in 4) at upstream/main as well, unchanged by this branch.

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in claiming a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore the two assertions the comment is actually about, plus the ladder order #2348 exists to protect, in ONE file. Split across two they each stayed green against the half they could not see, which is how the stylesheet half got deleted without anything noticing: - the ladder is declared on the --space-* scale in strictly increasing order (retuning a rung is a design call and stays green; list gaps meeting block gaps fails) - every rung declared is actually spent by a rule, so the ladder cannot become decoration while literals do the spacing - the runtime hooks the selectors need are really emitted by MarkdownBody — data-maka-contract, data-density, astryx-markdown-heading, data-level, astryx-list-item. These names have a single upstream owner in Astryx and appear in Maka only inside selectors, so a rename kills every rule at once with nothing failing - compact markdown still has exactly one caller, which is the bet the comment names: the table carries heading TYPOGRAPHY on a density key, and Astryx's own RFC says density must not do that Why these are not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions and CSS-structure suites that asserted how a rule was written. Nothing here asserts an appearance or a rule's shape: three of the four are relations that must survive any retune, and the fourth renders the real component and checks the join between Astryx's output and Maka's selectors. Deliberately left unpinned, because they ARE how the table is written: the adjacent-sibling gap form, the `hr` rung, the ListItem padding reset, and the two heading size tiers. The stylesheet-reading half no longer revives the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read; it reads packages/ui/src/styles.css directly, in the package that owns it. Seven mutations, each verified to fail the assertion that describes it: inverting the ladder; taking a rung off the --space-* grid; hardcoding a rung's usage; hardcoding it while naming the rung in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Also state in the comment what the contract does and does not hold, so the next reader does not have to open the test to find out. Generated-by: Claude Code

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in #2462 and its stylesheet-side sibling in #2425, so the comment has been citing a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore only what cannot be noticed without a test. Both failure modes here are silent: no error, no failing check, just spacing that quietly stops being what the table declares, because a CSS selector that matches nothing never complains. - The table selects entirely on DOM Astryx generates at runtime — data-density, astryx-markdown-heading + data-level, astryx-list-item. Those names have one upstream owner and appear in Maka only inside selectors, so a rename kills every rule at once. Astryx is bumped regularly (0.4.0 in #2983, 0.4.3 in flight, plus the Dependabot minor group), so this is a recurring event, not a hypothetical. - Astryx's ListItem carries CONTROL row padding that `density` cannot reach from outside. That padding is what inverted the ladder in #2348 — list items ~10px apart against 4px paragraphs — and one rule neutralizes it. Lose the rule and the original defect returns, silently. - Plus the assumption the comment actually names: compact markdown still has exactly one caller. The table carries heading TYPOGRAPHY on a density key, which Astryx's own RFC says density must not do, so the key is honest only while `compact` and "transcript" are the same set. Deliberately NOT pinned: the ladder's declared values and their order. A first draft asserted them, and it was the wrong instinct — a reversed ladder has to be typed on purpose into four adjacent lines under a comment explaining the order, and it is visible the moment anyone looks at a transcript. It also would not have caught the defect it claimed to guard: #2348's inversion came from the ListItem padding above, which a test reading only the declared variables cannot see. Same for the adjacent-sibling gap form, the `hr` rung and the two heading size tiers — those are how the table is written, not what it promises. Why this is not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions, and CSS-structure suites that asserted how a rule was written. Two of the three assertions here render the real component and check the join between Astryx's output and Maka's selectors; the third asserts an outcome the table promises, not a syntax. The stylesheet-reading half does not revive the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read — it reads packages/ui/src/styles.css directly, in the package that owns it. Six mutations, each verified to fail the assertion that describes it: deleting the ListItem padding reset; un-zeroing it; un-zeroing it while naming the reset in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Generated-by: Claude Code

jackwener · 2026-08-07T17:38:57Z

Summary

Delete 102 low-value or redundant source test files across Desktop, UI, Runtime, Headless, CLI, and Core, reducing the source test-file surface from 988 to 886 (-10.3%).

This branch is now based on the current main. The cleanup deliberately keeps tests that are the only guard for an active dependency patch or a security-sensitive boundary.

Removed

Desktop markup, copy, presentation, chrome, and overlapping geometry/workbar coverage
UI localization/copy panels, quiet chrome, empty states, and presentation pins
Runtime thin wire and duplicate provider helper coverage
Headless duplicate A/B, toolchain, and prompt-summary coverage
CLI layout/completion coverage and Core copy/locale helpers

Kept after review

Active dependency-patch guards:
- AI SDK tool-call index tracking
- Astryx List accessible-name forwarding
- Astryx new-message indicator behavior
Security-sensitive boundaries:
- split-chunk tool-output secret redaction
- permission escalation confirmation and persistence
- task-ledger prompt injection/redaction wiring
- Responses API store: false
- session-title internal-instruction stripping
Root authority, credentials, storage safety, primary runtime/session/tool suites, and fail-safe CI planning

Additional change

CLI process-contract tests ignore Node's ExperimentalWarning: SQLite stderr noise while continuing to reject other stderr output.

Validation

git diff --check origin/main...HEAD
npm run format:check
npm test — all workspace tests passed after synchronizing with current main

GitHub CI is running on the updated head.

Remove 122 source test files (~12% of packages/apps test files) that were markup/copy contracts, overlapping e2e chrome, duplicate UI presentation pins, and low-signal headless AB/toolchain matrix cases. Keep safety, authority, and primary product unit paths. Also tolerate Node SQLite ExperimentalWarning in CLI run process contracts so stderr-empty assertions are not flaky on modern Node.

Bring back e2e journeys, headless matrix/export, catalog contract, AiSdkFlow, and subscription non-leak coverage that are sole or primary gates. Keep the remaining ~11.6% file-level reduction of true low-value markup/chrome/matrix duplicates.

Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

This reverts commit 5ec1797.

Part of apache#2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until apache#2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds.

Part of #2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until #2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds. Co-authored-by: 陆逊 <211125649+UncertaintyDeterminesYou4ndMe@users.noreply.github.com>

apache#2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — apache#2161 pinned it against a containing block as tall as the conversation, apache#2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in apache#2462, and the multi-prompt fixtures it ran on in apache#2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine.

…2923) * fix(ui): give the prompt rail's tick bars a box again #2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — #2161 pinned it against a containing block as tall as the conversation, #2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in #2462, and the multi-prompt fixtures it ran on in #2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine. * fix(ui): make the prompt rail's hover and jump behave Four things the rail got wrong once it was visible again, found by using it: - A 4px gap between ticks was a band where the pointer was over the rail and over no tick, so the dock-style hover falloff dropped out and picked up again every few pixels of travel. The rail's `gap` moves into the ticks' own `padding-block`: same pitch, hit boxes now tile. - The hover preview waited 300ms before opening — Astryx's HoverCard default, meant for a pointer crossing a wide row on its way somewhere else. A tick is 22px of rail that nothing is on the way to, and the wait is the one part of this hover with no motion in it. Now 120ms. - The highlight glided 280ms to wherever a click landed, so crossing twenty prompts read as the bar flying off across the rail. A click now owns the highlight until its scroll settles: no glide, and the scroll no longer walks the highlight through every prompt it passes. - The first click into a session did nothing until the reader scrolled by hand. See below. That last one is a collision between Astryx's auto-follow lock and the progressive transcript mount, and neither side is wrong on its own. `useChatStreamScroll` unlocks on a scroll up, detected by comparing scrollTop between events — but it ignores any scroll event that arrives with a changed scrollHeight or offsetHeight, because Chrome fires those when content resizes and they are not the reader moving. A jump into an unmounted turn mounts it and the fill that follows changes scrollHeight for several frames, so the jump's own scroll is invisible to the lock: it stays on, and `scrollIfLocked` pulls the transcript back to the bottom. Only a wheel gesture broke it, which takes a separate path in Astryx. `holdJumpDestination` re-aims at the target on each height change until the fill stops. The last of those scrolls lands with a stable height, which is the one the lock finally reads as a scroll up. Measured on the 30-prompt fixture: clicking the first tick went to scrollTop 7042 (the bottom) and now goes to 24 and holds. The fixture grows from 8 prompts to 30 because the progressive mount's initial window is 10 — at 8 the head of the transcript is already mounted and the jump-into-unmounted-turns path never runs at all. Coverage note: the e2e case for the first click is an end-to-end check, not a guard. Whether the lock wins depends on which frame the fill lands on relative to a smooth scroll still in flight, and it goes green against the unfixed renderer often enough to be worthless as one. The guard is the `holdJumpDestination` unit test, which drives the frames itself. * fix(ui): own a rail jump through the mount instead of racing it Review of #2923 found the jump's ownership bound to a clock rather than to the navigation, and the e2e case that was supposed to guard it asserting almost nothing. Both hold. Jump ownership: - A second click during a jump only replaced the target; the first click's 700ms timer still governed, and could clear the second jump mid-flight. Each click now carries its own sequence and starts its own hold. - The fixed window is gone. A hold runs until the progressive mount reports the transcript filled AND nothing has moved for a few frames, so a long transcript is never released mid-fill, and it ends the moment the reader touches the transcript (wheel, touch, pointer, key) rather than outliving their interest in it. Chasing the "just release auto-follow" direction the review preferred found that ChatLayout publishes no such seam, so this adds one — `unlockAutoFollow` on `ChatLayoutContextValue`, exposing the scroll hook's existing `unlock` (patch hunk + patches/README entry). It is necessary and it is not sufficient, which the earlier framing got wrong: - Astryx re-locks on any `scrollend` that settles near the bottom, and a session that opens at the bottom produces exactly that while the mount is still catching up. Releasing once at the click is undone before the jump goes anywhere — traced: released at the click, landed at 154ms, dragged back to the bottom by 166ms. The release is now re-asserted for the life of the hold. - Auto-follow is not the only thing moving the transcript. The progressive mount's own scroll compensation holds the reader's position across each fill step, and mounting the turn a jump asked for IS a fill step, so it lands after the jump and restores the position the jump just left. That one no seam can fix; it is what the hold is for. Jumps also scroll instantly now, whatever the app's scroll-motion policy says. A jump is a teleport the reader asked for, and an animated one does not survive this surface: traced on the 30-prompt fixture, the smooth scroll was cancelled by the mount's compensation and by the follow spring and stalled two pixels from where it started. Coverage: - The first-click e2e case named the wrong turn (`[data-turn-id]` is the first MOUNTED turn, whose top is already negative at the opening scroll position, so an upper-bound-only check passed without the jump doing anything). It now names `turn-prompt-rail-1`, bounds it on both sides, and asserts that tick's `aria-current`. - `emulateMedia` could not put that case on the production scroll path: `resolveScrollMotionBehavior` collapses motion for ANY fixture, keyed on `data-maka-e2e-fixture` rather than on the media query. Fixtures can now ask for a behavior back (`scrollMotion`, per launch — it costs seconds of settling per window, so only the case that needs it pays), with unit coverage for the precedence: a fixture request never outranks a stated preference for less motion. - `holdJumpDestination`'s unit tests grew the two cases its rewrite is about: it must not settle while the transcript is still filling, and it must hand the transcript back the moment the reader touches it. Verified 5/5 on the smooth-scroll fixture, where the previous revision lost 1 in 4. `quote-selection.spec.ts` flakes on this machine (1 in 4) at upstream/main as well, unchanged by this branch.

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in claiming a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore the two assertions the comment is actually about, plus the ladder order #2348 exists to protect, in ONE file. Split across two they each stayed green against the half they could not see, which is how the stylesheet half got deleted without anything noticing: - the ladder is declared on the --space-* scale in strictly increasing order (retuning a rung is a design call and stays green; list gaps meeting block gaps fails) - every rung declared is actually spent by a rule, so the ladder cannot become decoration while literals do the spacing - the runtime hooks the selectors need are really emitted by MarkdownBody — data-maka-contract, data-density, astryx-markdown-heading, data-level, astryx-list-item. These names have a single upstream owner in Astryx and appear in Maka only inside selectors, so a rename kills every rule at once with nothing failing - compact markdown still has exactly one caller, which is the bet the comment names: the table carries heading TYPOGRAPHY on a density key, and Astryx's own RFC says density must not do that Why these are not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions and CSS-structure suites that asserted how a rule was written. Nothing here asserts an appearance or a rule's shape: three of the four are relations that must survive any retune, and the fourth renders the real component and checks the join between Astryx's output and Maka's selectors. Deliberately left unpinned, because they ARE how the table is written: the adjacent-sibling gap form, the `hr` rung, the ListItem padding reset, and the two heading size tiers. The stylesheet-reading half no longer revives the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read; it reads packages/ui/src/styles.css directly, in the package that owns it. Seven mutations, each verified to fail the assertion that describes it: inverting the ladder; taking a rung off the --space-* grid; hardcoding a rung's usage; hardcoding it while naming the rung in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Also state in the comment what the contract does and does not hold, so the next reader does not have to open the test to find out. Generated-by: Claude Code

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in #2462 and its stylesheet-side sibling in #2425, so the comment has been citing a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore only what cannot be noticed without a test. Both failure modes here are silent: no error, no failing check, just spacing that quietly stops being what the table declares, because a CSS selector that matches nothing never complains. - The table selects entirely on DOM Astryx generates at runtime — data-density, astryx-markdown-heading + data-level, astryx-list-item. Those names have one upstream owner and appear in Maka only inside selectors, so a rename kills every rule at once. Astryx is bumped regularly (0.4.0 in #2983, 0.4.3 in flight, plus the Dependabot minor group), so this is a recurring event, not a hypothetical. - Astryx's ListItem carries CONTROL row padding that `density` cannot reach from outside. That padding is what inverted the ladder in #2348 — list items ~10px apart against 4px paragraphs — and one rule neutralizes it. Lose the rule and the original defect returns, silently. - Plus the assumption the comment actually names: compact markdown still has exactly one caller. The table carries heading TYPOGRAPHY on a density key, which Astryx's own RFC says density must not do, so the key is honest only while `compact` and "transcript" are the same set. Deliberately NOT pinned: the ladder's declared values and their order. A first draft asserted them, and it was the wrong instinct — a reversed ladder has to be typed on purpose into four adjacent lines under a comment explaining the order, and it is visible the moment anyone looks at a transcript. It also would not have caught the defect it claimed to guard: #2348's inversion came from the ListItem padding above, which a test reading only the declared variables cannot see. Same for the adjacent-sibling gap form, the `hr` rung and the two heading size tiers — those are how the table is written, not what it promises. Why this is not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions, and CSS-structure suites that asserted how a rule was written. Two of the three assertions here render the real component and check the join between Astryx's output and Maka's selectors; the third asserts an outcome the table promises, not a syntax. The stylesheet-reading half does not revive the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read — it reads packages/ui/src/styles.css directly, in the package that owns it. Six mutations, each verified to fail the assertion that describes it: deleting the ListItem padding reset; un-zeroing it; un-zeroing it while naming the reset in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Generated-by: Claude Code

jackwener · 2026-08-07T17:38:57Z

Summary

Delete 102 low-value or redundant source test files across Desktop, UI, Runtime, Headless, CLI, and Core, reducing the source test-file surface from 988 to 886 (-10.3%).

This branch is now based on the current main. The cleanup deliberately keeps tests that are the only guard for an active dependency patch or a security-sensitive boundary.

Removed

Desktop markup, copy, presentation, chrome, and overlapping geometry/workbar coverage
UI localization/copy panels, quiet chrome, empty states, and presentation pins
Runtime thin wire and duplicate provider helper coverage
Headless duplicate A/B, toolchain, and prompt-summary coverage
CLI layout/completion coverage and Core copy/locale helpers

Kept after review

Active dependency-patch guards:
- AI SDK tool-call index tracking
- Astryx List accessible-name forwarding
- Astryx new-message indicator behavior
Security-sensitive boundaries:
- split-chunk tool-output secret redaction
- permission escalation confirmation and persistence
- task-ledger prompt injection/redaction wiring
- Responses API store: false
- session-title internal-instruction stripping
Root authority, credentials, storage safety, primary runtime/session/tool suites, and fail-safe CI planning

Additional change

CLI process-contract tests ignore Node's ExperimentalWarning: SQLite stderr noise while continuing to reject other stderr output.

Validation

git diff --check origin/main...HEAD
npm run format:check
npm test — all workspace tests passed after synchronizing with current main

GitHub CI is running on the updated head.

Remove 122 source test files (~12% of packages/apps test files) that were markup/copy contracts, overlapping e2e chrome, duplicate UI presentation pins, and low-signal headless AB/toolchain matrix cases. Keep safety, authority, and primary product unit paths. Also tolerate Node SQLite ExperimentalWarning in CLI run process contracts so stderr-empty assertions are not flaky on modern Node.

Bring back e2e journeys, headless matrix/export, catalog contract, AiSdkFlow, and subscription non-leak coverage that are sole or primary gates. Keep the remaining ~11.6% file-level reduction of true low-value markup/chrome/matrix duplicates.

Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

This reverts commit 5ec1797.

Part of apache#2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until apache#2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds.

Part of #2387. The Headless CLI and contamination-scan suites started a Node subprocess per assertion for scenarios that only exercise argument validation, task-run business logic, and report verdicts — semantics the exported entry points already expose. - cli.test.ts: 10 of 11 tests now run through mapLegacyMakaHeadlessArgs + runMakaEvalCli with stdout/stderr captured at the process-stream seam and env overrides applied and restored around the call. The non-Headless-root test keeps the real bin route as the representative wiring contract (real exit code, stack-free stderr). - contamination-scan-cli.test.ts: 13 of 14 tests call the script's exported main(argv), mirroring the executable footer exactly (thrown error -> stderr + exit 2). The no-argument rejection keeps the real subprocess as representative coverage of that footer itself, including its realpath main-module guard. - New shared helper withCapturedProcessIo swaps and restores the process-wide stream writers; safe for the sequential node:test runs these files use. runtime-policy-ab-cli.test.ts is intentionally untouched: its single test is the representative subprocess for run-runtime-policy-ab.mjs, whose main() is not exported, and adding an export to shave 0.6s is not warranted. (harness-ab-cli.test.ts was part of this change until #2462 deleted that suite on main.) Timing (node --test, local, warm build): cli.test.js 21.8s -> 13.1s (spawns 20 -> 2) contamination-scan-cli.test.js 1.05s -> 0.63s (spawns 14 -> 1) Tests pass across 3 consecutive rounds. Co-authored-by: 陆逊 <211125649+UncertaintyDeterminesYou4ndMe@users.noreply.github.com>

apache#2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — apache#2161 pinned it against a containing block as tall as the conversation, apache#2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in apache#2462, and the multi-prompt fixtures it ran on in apache#2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine.

…2923) * fix(ui): give the prompt rail's tick bars a box again #2580 moved the rail's tick onto Astryx's Button. The bar the tick draws was a direct child of the flex tick and got blockified; the Button wraps its children in a label span, so the bar went back to normal flow as an inline box. An inline box takes no width or height, so every bar computed to 0x0 and the rail shipped invisible in 0.1.9 and 0.1.10 — present in the DOM, painting nothing. `display: block` on the bar restores it. Measured on the new fixture at 1280x800: the rail's box goes from 8px wide (its own padding, ticks contributing nothing) back to the designed 22px. This is the third time the rail has failed by rendering and not painting — #2161 pinned it against a containing block as tall as the conversation, #2338 parked it under macOS's overlay scrollbar — and the second time it reached a release. The e2e coverage that would have caught all three was deleted in #2462, and the multi-prompt fixtures it ran on in #2656, so this adds back the smallest thing that closes the gap: - `chat-prompt-rail`, a plain 8-prompt conversation. The rail hides itself below three prompts, so the shipped single-prompt fixture cannot show it at all. - `prompt-rail.spec.ts` with one test per past failure: bars have a real box, the rail stays inside the scrollport at both scroll extremes, and a tick is what the pointer lands on. Three tests where the deleted suite had nine. Verified the first test fails on the unfixed renderer with "Expected: > 0, Received: 0" and passes with the fix. Neither a static CSS read nor a jsdom unit test can see any of this: jsdom has no layout engine. * fix(ui): make the prompt rail's hover and jump behave Four things the rail got wrong once it was visible again, found by using it: - A 4px gap between ticks was a band where the pointer was over the rail and over no tick, so the dock-style hover falloff dropped out and picked up again every few pixels of travel. The rail's `gap` moves into the ticks' own `padding-block`: same pitch, hit boxes now tile. - The hover preview waited 300ms before opening — Astryx's HoverCard default, meant for a pointer crossing a wide row on its way somewhere else. A tick is 22px of rail that nothing is on the way to, and the wait is the one part of this hover with no motion in it. Now 120ms. - The highlight glided 280ms to wherever a click landed, so crossing twenty prompts read as the bar flying off across the rail. A click now owns the highlight until its scroll settles: no glide, and the scroll no longer walks the highlight through every prompt it passes. - The first click into a session did nothing until the reader scrolled by hand. See below. That last one is a collision between Astryx's auto-follow lock and the progressive transcript mount, and neither side is wrong on its own. `useChatStreamScroll` unlocks on a scroll up, detected by comparing scrollTop between events — but it ignores any scroll event that arrives with a changed scrollHeight or offsetHeight, because Chrome fires those when content resizes and they are not the reader moving. A jump into an unmounted turn mounts it and the fill that follows changes scrollHeight for several frames, so the jump's own scroll is invisible to the lock: it stays on, and `scrollIfLocked` pulls the transcript back to the bottom. Only a wheel gesture broke it, which takes a separate path in Astryx. `holdJumpDestination` re-aims at the target on each height change until the fill stops. The last of those scrolls lands with a stable height, which is the one the lock finally reads as a scroll up. Measured on the 30-prompt fixture: clicking the first tick went to scrollTop 7042 (the bottom) and now goes to 24 and holds. The fixture grows from 8 prompts to 30 because the progressive mount's initial window is 10 — at 8 the head of the transcript is already mounted and the jump-into-unmounted-turns path never runs at all. Coverage note: the e2e case for the first click is an end-to-end check, not a guard. Whether the lock wins depends on which frame the fill lands on relative to a smooth scroll still in flight, and it goes green against the unfixed renderer often enough to be worthless as one. The guard is the `holdJumpDestination` unit test, which drives the frames itself. * fix(ui): own a rail jump through the mount instead of racing it Review of #2923 found the jump's ownership bound to a clock rather than to the navigation, and the e2e case that was supposed to guard it asserting almost nothing. Both hold. Jump ownership: - A second click during a jump only replaced the target; the first click's 700ms timer still governed, and could clear the second jump mid-flight. Each click now carries its own sequence and starts its own hold. - The fixed window is gone. A hold runs until the progressive mount reports the transcript filled AND nothing has moved for a few frames, so a long transcript is never released mid-fill, and it ends the moment the reader touches the transcript (wheel, touch, pointer, key) rather than outliving their interest in it. Chasing the "just release auto-follow" direction the review preferred found that ChatLayout publishes no such seam, so this adds one — `unlockAutoFollow` on `ChatLayoutContextValue`, exposing the scroll hook's existing `unlock` (patch hunk + patches/README entry). It is necessary and it is not sufficient, which the earlier framing got wrong: - Astryx re-locks on any `scrollend` that settles near the bottom, and a session that opens at the bottom produces exactly that while the mount is still catching up. Releasing once at the click is undone before the jump goes anywhere — traced: released at the click, landed at 154ms, dragged back to the bottom by 166ms. The release is now re-asserted for the life of the hold. - Auto-follow is not the only thing moving the transcript. The progressive mount's own scroll compensation holds the reader's position across each fill step, and mounting the turn a jump asked for IS a fill step, so it lands after the jump and restores the position the jump just left. That one no seam can fix; it is what the hold is for. Jumps also scroll instantly now, whatever the app's scroll-motion policy says. A jump is a teleport the reader asked for, and an animated one does not survive this surface: traced on the 30-prompt fixture, the smooth scroll was cancelled by the mount's compensation and by the follow spring and stalled two pixels from where it started. Coverage: - The first-click e2e case named the wrong turn (`[data-turn-id]` is the first MOUNTED turn, whose top is already negative at the opening scroll position, so an upper-bound-only check passed without the jump doing anything). It now names `turn-prompt-rail-1`, bounds it on both sides, and asserts that tick's `aria-current`. - `emulateMedia` could not put that case on the production scroll path: `resolveScrollMotionBehavior` collapses motion for ANY fixture, keyed on `data-maka-e2e-fixture` rather than on the media query. Fixtures can now ask for a behavior back (`scrollMotion`, per launch — it costs seconds of settling per window, so only the case that needs it pays), with unit coverage for the precedence: a fixture request never outranks a stated preference for less motion. - `holdJumpDestination`'s unit tests grew the two cases its rewrite is about: it must not settle while the transcript is still filling, and it must hand the transcript back the moment the reader touches it. Verified 5/5 on the smooth-scroll fixture, where the previous revision lost 1 in 4. `quote-selection.spec.ts` flakes on this machine (1 in 4) at upstream/main as well, unchanged by this branch.

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in claiming a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore the two assertions the comment is actually about, plus the ladder order #2348 exists to protect, in ONE file. Split across two they each stayed green against the half they could not see, which is how the stylesheet half got deleted without anything noticing: - the ladder is declared on the --space-* scale in strictly increasing order (retuning a rung is a design call and stays green; list gaps meeting block gaps fails) - every rung declared is actually spent by a rule, so the ladder cannot become decoration while literals do the spacing - the runtime hooks the selectors need are really emitted by MarkdownBody — data-maka-contract, data-density, astryx-markdown-heading, data-level, astryx-list-item. These names have a single upstream owner in Astryx and appear in Maka only inside selectors, so a rename kills every rule at once with nothing failing - compact markdown still has exactly one caller, which is the bet the comment names: the table carries heading TYPOGRAPHY on a density key, and Astryx's own RFC says density must not do that Why these are not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions and CSS-structure suites that asserted how a rule was written. Nothing here asserts an appearance or a rule's shape: three of the four are relations that must survive any retune, and the fourth renders the real component and checks the join between Astryx's output and Maka's selectors. Deliberately left unpinned, because they ARE how the table is written: the adjacent-sibling gap form, the `hr` rung, the ListItem padding reset, and the two heading size tiers. The stylesheet-reading half no longer revives the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read; it reads packages/ui/src/styles.css directly, in the package that owns it. Seven mutations, each verified to fail the assertion that describes it: inverting the ladder; taking a rung off the --space-* grid; hardcoding a rung's usage; hardcoding it while naming the rung in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Also state in the comment what the contract does and does not hold, so the next reader does not have to open the test to find out. Generated-by: Claude Code

styles.css told readers the "compact == transcript" assumption was held by `__tests__/markdown-rhythm-dom-contract.test.tsx`. That file was deleted in #2462 and its stylesheet-side sibling in #2425, so the comment has been citing a guard that does not exist — the worst state to leave an assumption in, since the next person reads the citation and stops looking. Restore only what cannot be noticed without a test. Both failure modes here are silent: no error, no failing check, just spacing that quietly stops being what the table declares, because a CSS selector that matches nothing never complains. - The table selects entirely on DOM Astryx generates at runtime — data-density, astryx-markdown-heading + data-level, astryx-list-item. Those names have one upstream owner and appear in Maka only inside selectors, so a rename kills every rule at once. Astryx is bumped regularly (0.4.0 in #2983, 0.4.3 in flight, plus the Dependabot minor group), so this is a recurring event, not a hypothetical. - Astryx's ListItem carries CONTROL row padding that `density` cannot reach from outside. That padding is what inverted the ladder in #2348 — list items ~10px apart against 4px paragraphs — and one rule neutralizes it. Lose the rule and the original defect returns, silently. - Plus the assumption the comment actually names: compact markdown still has exactly one caller. The table carries heading TYPOGRAPHY on a density key, which Astryx's own RFC says density must not do, so the key is honest only while `compact` and "transcript" are the same set. Deliberately NOT pinned: the ladder's declared values and their order. A first draft asserted them, and it was the wrong instinct — a reversed ladder has to be typed on purpose into four adjacent lines under a comment explaining the order, and it is visible the moment anyone looks at a transcript. It also would not have caught the defect it claimed to guard: #2348's inversion came from the ListItem padding above, which a test reading only the declared variables cannot see. Same for the adjacent-sibling gap form, the `hr` rung and the two heading size tiers — those are how the table is written, not what it promises. Why this is not what #2425/#2462 retired. Those PRs removed markup/copy pins, duplicate presentation assertions, and CSS-structure suites that asserted how a rule was written. Two of the three assertions here render the real component and check the join between Astryx's output and Maka's selectors; the third asserts an outcome the table promises, not a syntax. The stylesheet-reading half does not revive the deleted desktop css-test-helpers module or the renderer-wide CSS aggregate it read — it reads packages/ui/src/styles.css directly, in the package that owns it. Six mutations, each verified to fail the assertion that describes it: deleting the ListItem padding reset; un-zeroing it; un-zeroing it while naming the reset in a comment (this is why comments are stripped first); renaming the contract wrapper; cutting density off before it reaches Astryx; and adding a second compact caller. Generated-by: Claude Code

jackwenerand others added 8 commits August 8, 2026 01:38

test: restore models.dev and provider-catalog contract gates
a91e190
Keep the sole PROVIDER_REGISTRY×models.dev sync sweep and the provider catalog tab/OAuth/slug contract. These are not markup-copy pins and have no thinner remaining suite.

Merge remote-tracking branch 'origin/main' into cleanup/delete-low-va…
4665170
…lue-tests # Conflicts: # apps/desktop/e2e/new-messages-indicator.spec.ts # apps/desktop/e2e/prompt-rail.spec.ts # packages/ui/src/__tests__/conversation-localization.test.tsx

test: restore unique safety and behavior coverage
5ec1797

Revert "test: restore unique safety and behavior coverage"
a95af4e
This reverts commit 5ec1797.

test: retain essential safety and patch guards
120b063

refactor(desktop): remove test-only footer helper
de609e5

Astro-Han merged commit 2f5c75b into mainAug 8, 2026
21 of 23 checks passed

Astro-Han deleted the cleanup/delete-low-value-tests branch August 8, 2026 03:28

Astro-Han mentioned this pull request Aug 8, 2026
test: delete obsolete presentation and helper suites #2478
Merged

This was referenced Aug 8, 2026
test(headless): run ordinary CLI command semantics in process #2480
Merged
test(cli): run ordinary run-command semantics in process #2476
Closed

AidenNovak mentioned this pull request Aug 9, 2026
fix(desktop): fuse the collapsed rail with the plate under the traffic lights #2396
Closed

ARE404 mentioned this pull request Aug 12, 2026
fix(ui): restore the prompt anchor rail and fix its hover and jump #2923
Merged

This was referenced Aug 21, 2026
fix(ui): give transcript paragraphs a full block step #3436
Merged
test(ui): re-pin the transcript rhythm contract in one file #3437
Closed