Skip to content

Re-audit the Modules/imports panel#96

Merged
tamnd merged 1 commit into
mainfrom
feat/v0.13.7-spec-1731-imports-reaudit
Jun 19, 2026
Merged

Re-audit the Modules/imports panel#96
tamnd merged 1 commit into
mainfrom
feat/v0.13.7-spec-1731-imports-reaudit

Conversation

@tamnd

@tamnd tamnd commented Jun 19, 2026

Copy link
Copy Markdown
Owner

Went back over the Modules/imports panel (the 12 flat test files plus the test_import, test_importlib and test_module suites) on the current build to recheck where it actually stands.

I ran each entry the way the gate runs it, one process at a time, and everything comes back clean:

  • test_module 39
  • test_import 118 (4 skipped)
  • test_importlib 1346 (10 skipped)
  • test_frozen 3
  • test_modulefinder 17
  • test_pkg 8
  • test_pkgutil 21
  • test_pyclbr 6
  • test_runpy 40
  • test_zipimport 91 (4 skipped)
  • test_zipimport_support 4
  • test_zipapp 35

No errors, no failures, and the MANIFEST counts already line up, so there was no port change to make here.

The one thing worth flagging: while re-running test_zipimport with several copies going at once in the same directory I'd occasionally see an EOFError come out of zipimport's _get_data. That's the suite reusing a single fixed archive name in the cwd, so two concurrent processes end up truncating each other's file mid-read. CPython behaves the same way in that setup. Sequentially, one process per entry, it passes every time. I left the code alone and wrote the finding into the spec rather than papering over it.

So this PR is just the re-audit note in spec 1731.

… artifact

Re-ran every entry in the panel one process at a time, the way the gate
drives them, to recheck ground truth on the current build. All twelve files
plus the three directory suites come back green with no errors and no
failures, and the MANIFEST counts already match, so nothing in the port
needed changing.

While re-running test_zipimport I hit the occasional EOFError out of
zipimport's _get_data when I had several copies going at once in the same
directory. That turns out to be the suite reusing one fixed archive name in
the cwd, so concurrent processes stomp on each other's file. CPython does
the same thing under that setup. Run sequentially, one process per entry,
it passes every time, so I left the code alone and wrote the finding into
the spec.
@tamnd

tamnd commented Jun 19, 2026

Copy link
Copy Markdown
Owner Author

CI's green across the board (linux/macos/windows test matrices, vet, lint, cfg-phase-parity). Nothing in the panel needed touching, so this is just the re-audit note landing.

@tamnd tamnd merged commit a885245 into main Jun 19, 2026
6 checks passed
@tamnd tamnd deleted the feat/v0.13.7-spec-1731-imports-reaudit branch June 19, 2026 09:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant