Uh oh!
There was an error while loading. Please reload this page.
Rewrite the backlog around what is left after the container works - #1241
Conversation
The goal these items were organised around is met, so the Todo said things that are no longer true - item 13 read as open when its Jass half landed, and the ordering still pointed at the feature rather than at what surrounds it. What is left is generality (items 6, 7, 13's Lua half, 15) and two gaps in what the suite can see: no standard library program is ever executed on Lua, and a bump of the pinned library is only checked for compiling. Both are recorded as items rather than as asides, because the first is how the empty-allocation bug survived. Items 6 and 13 both end at the same erasure decision, which is now item 18 rather than a paragraph repeated in each.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit:bb2f32d5d8
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| 20. **`luaOutputIsDeterministicForGenericOverrideSlots` is itself nondeterministic.** It failed once | ||
| on Windows CI and passed on a re-run of the same commit, having blocked an unrelated pull | ||
| request in between. A test asserting the compiler is deterministic which is not itself | ||
| deterministic is worse than none: every failure has to be investigated before it can be | ||
| dismissed, and the next one will be too. Either find what differs between the two compiles or |
There was a problem hiding this comment.
Do not replace this assertion merely because it fails intermittently: the test compiles the same repro twice and compares byte-identical output, so an intermittent mismatch is evidence of the Lua emission nondeterminism it is specifically required to detect unless harness interference is first proven. Keep the compile-twice equality check and diagnose the differing output or unstable compiler iteration instead of weakening the assertion.
AGENTS.md reference: AGENTS.md:L277-L280
Useful? React with 👍 / 👎.
| Jass, and compiles on Lua (#1239). What is left is generality around it rather than the feature | ||
| itself, and one gap in what the suite can see. | ||
| 16. **Nothing executes a standard library program on Lua.** Every test which puts the library on |
There was a problem hiding this comment.
The new Todo entries reuse 16, 17, and 18 even though completed items with those IDs remain under Done at lines 213–225. This defeats the stated stable-numbering scheme and makes references such as “item 18” ambiguous between the new erasure decision and the completed unresolved-default fix; allocate fresh numbers instead.
Useful? React with 👍 / 👎.
| 9. **Keep `CHANGELOG.md` current** as items land — a standing practice rather than a task to | ||
| finish. `WURST_LANGUAGE.md`, which this item used to name, is not in this repository: the | ||
| language documentation lives in the docs repo, so the changelog is the only dev-facing surface | ||
| reachable from here. It covers type class bounds with a worked example. |
There was a problem hiding this comment.
The repository does contain de.peeeq.wurstscript/src/main/resources/agent-docs/WURST_LANGUAGE.md, and it is a tracked compiler resource that already documents type-class bounds and closure behavior. Removing it from this standing item based on the false claim that it is external will let that shipped language reference drift as the remaining behavior lands.
Useful? React with 👍 / 👎.
…eference Three corrections to the rewrite. The new items reused 16, 17 and 18, which Done already holds, so a reference to "item 18" was ambiguous between the erasure decision and a finished one. They are 21, 22 and 23 now, and the determinism item is 24. That item said the assertion could be replaced if diagnosing it proved hard. It compiles one repro twice and compares byte for byte, so an intermittent mismatch is evidence of intermittent nondeterminism in Lua emission, which is what it exists to catch. Calling it flaky on the strength of a passing re-run was too quick, and the item says so and says to diagnose instead. WURST_LANGUAGE.md is in this repository, under the compiler's resources rather than at the root, which is where I looked. Item 9 names it again, and since the standing practice is to fold documentation into whichever item changes the behaviour, the reference now covers subclassing a bounded generic (and that it does not work on Lua) and strings being bytes.
Frotty
commented
Aug 16, 2026
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit:81159183e2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| which asserts the program compiles, runs, and never reaches `testSuccess` rather than leaving | ||
| the difference between the targets to be discovered. `transformGenericNewOnly` runs neither | ||
| `simplifyClasses` nor `addMemberTypeArguments`, so the type variables are never lifted there and | ||
| a super call has nothing to carry. Closing it is item 23. |
There was a problem hiding this comment.
After centralizing the Lua erasure work under item 23, TypeClassTests.java:351 still says that the remaining subclassing issue is tracked by item 5. Item 5 is now listed under Done, so anyone following that test's JavaDoc is directed to completed closure work rather than this decision item; update the JavaDoc as part of the renumbering.
Useful? React with 👍 / 👎.
Its JavaDoc named item 5, which is finished, so a reader following it landed on closure work rather than the decision the test is waiting on. The erasure question is item 23 now that it is one item instead of a paragraph repeated under two.
Frotty
commented
Aug 16, 2026
All three addressed, and two of them were errors of mine rather than differences of opinion. Keep the compile-twice determinism check — you are right, and I was too quick to call it flaky. It compiles one repro twice and compares byte for byte, so an intermittent mismatch is evidence of intermittent nondeterminism in Lua emission, which is the thing it exists to catch; a re-run passing says the nondeterminism is intermittent, not that the test is at fault. The item now says so in those terms, says not to weaken the assertion, and says to diagnose instead — capture both outputs on a failing run and diff them, and prove harness interference rather than assume it. One thing already ruled out: Assign unused IDs — correct, Keep the tracked language digest — also correct, and my claim that it was external was simply wrong: I searched two levels from the repository root and it sits under Since the standing practice is to fold documentation into whichever item changes the behaviour, and two items landed without doing so, the digest now also covers what shipped: subclassing a bounded generic class and that it does not work on Lua, and strings being byte sequences on both targets, including that a compiletime expression cannot return half a character. Also caught by the follow-up: @codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit:06b2626582
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| interference rather than assuming it. `WurstScriptTest` clears the global caches around every | ||
| test method, so the two compiles do start from the same cache state; whatever differs is |
There was a problem hiding this comment.
@BeforeMethod and @AfterMethod only clear caches before and after the entire test method, so they do not by themselves give the two internal compilations the same starting state. That is actually guaranteed by the explicit GlobalCaches.clearAll() between the compilations in LuaTranslationTests.luaOutputIsDeterministicForGenericOverrideSlots; naming that reset matters when diagnosing this failure, because removing or bypassing it would invalidate the stated conclusion about cache interference.
Useful? React with 👍 / 👎.
The item said clearing the caches around every test method is what gives the two compilations the same starting state. That is half of it: the before-method reset covers the first compile, and the explicit clearAll between them covers the second. Naming the wrong one matters when diagnosing, because removing the in-test reset would invalidate the conclusion rather than explain the failure.
Frotty
commented
Aug 16, 2026
Right, and the distinction matters for exactly the reason given: the conclusion depends on which reset is doing the work. The item now names both. @codex review |
Codex Review: Didn't find any major issues. Chef's kiss. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
Uh oh!
There was an error while loading. Please reload this page.
Documentation only. The goal these items were organised around is met, so the Todo said things that are no longer true — item 13 read as open when its Jass half landed in #1237, and the ordering still pointed at the feature rather than at what surrounds it.
What changed:
WURST_LANGUAGE.md, which is not in this repository.moveFunctionsOutOfClasscopies type variables deliberately and Give a super call the type arguments of the class it reaches #1237 depends on that copy, so what is being made canonical is per scope rather than per program.