KarpelesLab / libwallet / 25899011415
24%

DEFAULT BRANCH: master
Ran
Jobs 1
Files 118
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 21.864%. Remained the same
25899011415

push

github

MagicalTux 
ci/dart-test: retry the iOS job on any pre-test infra failure

The test-ios retry classifier only recognised two flake signatures —
the watchdog timeout and the "Waiting for VM Service port" hang.
A third recurring flake, "Application ... is unknown to FrontBoard"
(the simulator's install database losing the app bundle), exits
plain 1 and matched neither, so the job hit "real failure — not
retrying" and went red. It bit 0.4.28, 0.4.30 and 0.4.32, each time
clearing on a manual rerun.

Invert the classifier. Instead of an allowlist of known-flaky
signatures, key off the one reliable GENUINE-failure signal:
flutter's canonical `Some tests failed.` end-of-run line, printed
iff tests actually ran and an assertion failed. Any non-zero exit
that doesn't reach that line never got a test to run — watchdog
kill, VM Service hang, FrontBoard launch failure, pre-setUpAll
crash — all macos-runner / simulator infra, all retryable.

Retrying a genuine compile error or deterministic failure just
burns one cycle then fails identically — no bug is masked — so the
worst case of the broader retry is marginally slower red CI, never
a falsely-green one.

Between attempts: `simctl erase` (full wipe) instead of just a
reboot. The FrontBoard failure is a corrupt install DB; only erase
clears it. The retry then gets a pristine simulator.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

3608 of 16502 relevant lines covered (21.86%)

115.16 hits per line

Jobs
ID Job ID Ran Files Coverage
1 25899011415.1 118
21.86
 GitHub Action Run
Source Files on build 25899011415