Several false positives found by stats-based bisect |
||||
Issue descriptionrefer to go/badbisects for a list of examples. General traits of these are: - Small regression found, - Bisect re-tests the reference range several times (> 25) - An unrelated culrpit is found with only a tiny regression We should tweak the constants we use to make bisect give up early in these cases.
,
Mar 14 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/tools/build.git/+/eba9f2047647330ebbd36bd565db4fcf8c1975f1 commit eba9f2047647330ebbd36bd565db4fcf8c1975f1 Author: robertocn@chromium.org <robertocn@chromium.org> Date: Mon Mar 14 21:08:33 2016 Further reducing the max required samples. Some cases have been observed where the bisect fails to give up after 18 samples. R=dtu BUG= 593470 Review URL: https://codereview.chromium.org/1778233002 git-svn-id: svn://svn.chromium.org/chrome/trunk/tools/build@299282 0039d316-1c4b-4281-b951-d872f2087c98 [modify] https://crrev.com/eba9f2047647330ebbd36bd565db4fcf8c1975f1/scripts/slave/recipe_modules/auto_bisect/bisector.py
,
Mar 22 2016
The following revision refers to this bug: http://goto.ext.google.com/viewvc/chrome-internal?view=rev&revision=85009 ------------------------------------------------------------------ r85009 | recipe-roller@chromium.org | 2016-03-10T00:49:17.301141Z -----------------------------------------------------------------
,
Mar 22 2016
Moving out of the queue for a while. Let's keep gathering samples.
,
Feb 3 2017
,
Feb 6 2017
|
||||
►
Sign in to add a comment |
||||
Comment 1 by bugdroid1@chromium.org
, Mar 9 2016