New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 839031 link

Starred by 2 users

Issue metadata

Status: Fixed
Owner:
Closed: Aug 3
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 1
Type: Bug

Blocking:
issue 837855



Sign in to add a comment

Correctly handle without patch logic in new flake detection

Project Member Reported by liaoyuke@chromium.org, May 2 2018

Issue description

There is a bug in ChromiumTryFlake in how it handles without patch logic:
https://cs.chromium.org/chromium/infra/appengine/chromium_try_flakes/handlers/flake_issues.py?l=834

The way ChromiumTryFlake handles it is that for any given test step, if both with patch and without patch steps fail, then this test step is completely ignored, this is not correct, for example:

fail build:
  dummy test (with patch): fails t1, t2, t3
  dummy test (without patch): fails t3

success build:
  dummy test (with patch): success

In this case, it tells us t3 is unrelated to the patch (likely a consistent failure). But t1 and t2 should be deemed as flakes.

 

Comment 1 by st...@chromium.org, May 2 2018

If it is not an easy fix, I'm wondering whether it's better to focus on the new flake detection instead.
That said, your call on this.

https://bigquery.cloud.google.com/results/chrome-infra-events:US.bquijob_56bf6030_163223ba682

or the table chrome-infra-events:test_data.CTF_false_negatives

Comment 2 by st...@chromium.org, May 2 2018

It seems I misread. NVM.
Labels: flake-detection
Status: Fixed (was: Assigned)

Sign in to add a comment