Conformance: badge row, hardware-tested section, actionlint lane, all six 64-bit arches - #2
Merged
Merged
Conversation
The README had no badge row at all, the workflow had no actionlint lane, and the cross matrix covered two of Go's six 64-bit architectures. - README: the fleet badge row (CI, Go Reference, Go Report Card, coverage, licence). The coverage badge says "portable layer", because that is what the gate actually is -- the purego bindings cannot be covered without a display and a TCC grant, and a badge claiming a total would be a lie. - README: a "Tested against real hardware" section in the shape go-xrkit/desk established -- what was connected and exercised, and, separately, what was NOT proven: whole-display capture (no Screen Recording grant on the build machine, though the refusal path IS asserted), Intel Macs, and every macOS below 26.6.2. Plus the invitation to send us hardware. - ci.yml: an actionlint lane. A workflow that fails to parse reports nothing at all, so no other lane can tell you it is broken. - ci.yml: linux/riscv64, loong64, ppc64le and s390x added to the cross matrix, so all six 64-bit architectures are built and vetted. Verified locally first. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
goreportcard.com now answers every badge request, for every module, with
aria-label="go report: retired"
and its home page says "After more than a decade, Go Report Card has been
sunset." Checked against this module, against another repo in the fleet that
has carried the badge for months, and against gin-gonic/gin: all three render
the same grey "retired" pill.
A badge that can only ever say "retired" is worse than no badge -- it is the
same failure as one pointing at a workflow that does not exist, and a reader
cannot tell a retired service from a repository nobody grades. So it comes out
rather than staying as a decoration.
This affects far more than this repository: every repo in the fleet carrying a
Go Report Card badge is showing the same thing right now.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Contributor
Author
|
Correction, pushed as a follow-up commit on this branch: the Go Report Card badge is gone again.
A badge that can only ever say "retired" is worse than no badge — the same failure mode as one pointing at a workflow that does not exist. So it is out, and where this repo already carried one, that one is out too. This is a fleet-wide matter, not a matter for this PR: every repository in every org that carries a Go Report Card badge is rendering the grey "retired" pill right now. |
…he old one Two things, both about where a screen capture may be written. 1. The README said "The live suite writes a PNG to testdata/artifacts/window-capture.png". It has not done that for a while -- the code writes to captureDir(t), outside every work tree -- and the sentence described exactly the behaviour the rule exists to prevent. Anyone reading it would believe a capture of their desktop lands in a public repository, which is both wrong and, if it had been true, the defect. It now says what the code does, including that the walk-up refuses a .git that is a FILE, which is what a worktree has. 2. captureDir, chooseCaptureDir and repoRootOf move from live_darwin_test.go -- which is behind `darwin && integration` and needs a window server no CI runner has -- into an UNTAGGED capturedir_test.go. A guard that only compiles where the live suite runs is a guard nobody runs. The rule is plain filesystem reasoning with nothing macOS in it, so it now compiles and runs on the linux and windows lanes as well, on every push. The decision is split out of the test plumbing as chooseCaptureDir so the REFUSAL can be exercised -- a guard whose failing branch never runs is not known to work, and that branch is the entire reason it exists. Four committable paths are refused, including the directory the committed frame lives in and a relative path, and the default is asserted to be usable and absolute. This is the shape go-mswin/screencapture settled on. The committed frame under testdata/artifacts/ stays: it was put there by hand from a disposable machine, and nothing a test runs can land there again. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Brings the repo to the fleet's organisation-conformance standard. No prose was rewritten; the README's own voice, measurements and honest limits are untouched.
What was missing
actionlintlane — a workflow that fails to parse reports nothing, so no other lane can say it is broken.What was added
## Tested against real hardwaresection in the shapego-xrkit/deskestablished: hardware connected and exercised / not proven on hardware / send us hardware. It states plainly that whole-display capture and other-process window capture are not proven (no Screen Recording grant on the build machine — the refusal path is asserted, the pixels are not), that no Intel Mac ever ran this, and that only macOS 26.6.2 was run.actionlintlane, first in the file.linux/riscv64,loong64,ppc64le,s390xin the cross matrix. All four verified to build locally before pushing.No coverage gate was weakened.
Co-Authored-By: Claude Opus 5 noreply@anthropic.com