Skip to content

[CI] Flaky: Mac Catalyst NativeAOT exits 133 in CGImageMetadataTest #26391

Description

@dalexsoto

Flaky Test Report

Signature: maccatalyst-nativeaot-cgimagemetadata-exit-133
Test/configuration: monotouch-test/Mac Catalyst/Release (trimmable static registrar, NativeAOT) while entering MonoTouchFixtures.CoreGraphics.CGImageMetadataTest
Classification: Genuinely flaky test
First observed: 2026-07-21 12:22 UTC
Latest observed: 2026-08-03 14:27 UTC

Occurrences

PR / head Azure run / attempt Job Outcome
#26180 / 897e85a4b8e8e206c57ea59a7b0c3088abf18e15 14718322 / 1 T: monotouch_maccatalyst (73408371-a55e-5ff5-5372-c49d56cfe7c2) The NativeAOT process exited 133 while entering CGImageMetadataTest. The same head/configuration had passed in 14716050, and failed-job retry attempt 2 cleared this crash; unrelated jobs kept the overall run failed.
#26379 / d97396de11660c46503461cc61cc6aadc1c56dd4 14850072 / 1 T: monotouch_maccatalyst / Run tests, AcesShared 37 The same configuration exited 133 at CGImageMetadataTest; no crash report was produced.

Total: 2 confirmed occurrences across 2 distinct PRs.

Current diagnostic lines

MonoTouchFixtures.CoreGraphics.CGImageMetadataTest
Process monotouchtest exited with 133
Test run crashed (exit code: 133).
No crash reports

In run 14850072, the neighboring Release (NativeAOT) and Release (NativeAOT, .NET 11 defaults) configurations passed. This isolates the recurrence to the trimmable-static-registrar NativeAOT variation rather than a general NativeAOT build failure.

Impact and retry status

The crash fails the Mac Catalyst test job without an NUnit failure or native crash report. No retry was submitted for #26379 because the same run also has seven deterministic No valid ... code signing keys found in keychain failures; rerunning failed jobs would repeat those failures until the signing bundle is independently validated.

Investigation

Workaround

Retry the failed Mac Catalyst job only when the same run has no deterministic co-failures and retry accounting permits it. No source workaround is known.


Generated from recurring cross-PR CI monitoring evidence.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugIf an issue is a bug or a pull request a bug fixci-failureThe issue only affects CIci-postmortemCI post-mortem analysis: flaky tests, infrastructure failures, shared regressionscopilot

    Type

    No type

    Projects

    No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions