New issue
Advanced search Search tips

Issue 919944 link

Starred by 1 user

Issue metadata

Status: Untriaged
Owner: ----
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

False rejects caused by Linux TSAN builder due to inconsistent retry policy.

Project Member Reported by erikc...@chromium.org, Jan 8

Issue description

Example build: https://ci.chromium.org/p/chromium/builders/luci.chromium.try/linux_chromium_tsan_rel_ng/174770

If there are TSAN data races in 'with_patch', the failures are not retried.

However, we will run 'retry_with_patch' with 10 repeats, and if any of the repeats fails, then we will mark the build as a failure.

As per https://docs.google.com/document/d/1O9nzVMA6rEe2-rhjsni_8wS9Lw4OnFpK2NE0wWB5VaY/edit we should try to make our retries as fast as possible.

I think we should do one of the following:
1) Skip retries in 'with_patch'. Skip 'retry_with_patch' altogether.
2) Add retries to 'with_patch'. Skip 'retry_with_patch' altogether.
3) Add retries to 'with_patch'. Also run 'retry_with_patch'.

I think that (3) is the best option as it's most consistent with our other test suites. (1) would be consistent with the GPU-team style approach [but they have a dedicated triage rotation]. 
 

Sign in to add a comment