New issue
Advanced search Search tips

Issue 759113 link

Starred by 1 user

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 1
Type: Bug

Blocking:
issue 731558



Sign in to add a comment

A test infra-failed (purple), but build non-infra failed (red)

Project Member Reported by no...@chromium.org, Aug 25 2017

Issue description

https://luci-milo.appspot.com/swarming/task/3809848685195d10
has purple infra-failed context_lost_tests steps due to swarming capacity, but the build is marked as red. I expected the build to be marked as infra-failed (purple).

I do see some red steps there too. "webkit_python_tests (with patch)" failed, but I don't see "webkit_python_tests (without patch)", which makes me think that the build should not have decided that the cause of webkit_python_tests failure is the patch.

Please close the bug if my expectation is incorrect.

Fixing this bug is needed for LUCI correctness estimation ( bug 731558 )
 

Comment 2 by efoo@chromium.org, Sep 18 2017

Status: Available (was: Untriaged)
Owner: jbudorick@chromium.org
Status: Assigned (was: Available)
Cc: martiniss@chromium.org
Appears related to https://codesearch.chromium.org/chromium/build/scripts/slave/recipe_modules/swarming/api.py?rcl=5cfd9f3026fca2ef7d171755c0e795d2202295f1&l=1094. I don't quite understand the mechanics of the hack there, so +martiniss: what were you trying to avoid w/ those few lines?
That line is there because when you raise an InfraFailure, it asserts that the return code was 2. I'm not exactly sure where this is enforced in the recipe engine, but I know I've seen it before.

IIRC, I've seen things like this before, and usually the recipe itself is catching the infra failures and setting the status of the build itself, which is why the build isn't purple overall.

Comment 6 by no...@chromium.org, Sep 22 2017

so the recipe should be fixed not to replace an infra failure with a non-infra failure?
Something along those lines, yeah.
Collection is eating the exception raised by _handle_summary_json: https://codesearch.chromium.org/chromium/build/scripts/slave/recipe_modules/swarming/api.py?rcl=191fd1236a79db2f12c30d7c73edef77b417ab7c&l=860

Fixing this isn't as simple as letting InfraFailures pass through, though, as doing so would keep us from collecting all subsequent results due to https://codesearch.chromium.org/chromium/build/scripts/slave/recipe_modules/chromium_tests/api.py?rcl=191fd1236a79db2f12c30d7c73edef77b417ab7c&l=320

Need to think about this one a bit more.

Sign in to add a comment