Bug 36911 - Add webkit-patch command to crawl buildbot history and find when tests started to fail
Summary: Add webkit-patch command to crawl buildbot history and find when tests starte...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: 528+ (Nightly build)
Hardware: Other OS X 10.5
: P2 Normal
Assignee: Nobody
URL:
Keywords:
Depends on:
Blocks: 37063
  Show dependency tree / graph
 
Reported: 2010-03-31 17:50 PDT by Eric Seidel (no email)
Modified: 2010-04-06 23:41 PDT (History)
5 users (show)

See Also:

Attachments
Patch (13.63 KB, patch)
2010-03-31 17:56 PDT, Eric Seidel (no email) 		no flags Details | Formatted Diff | Diff
First Try (16.80 KB, patch)
2010-04-06 20:26 PDT, Eric Seidel (no email) 		no flags Details | Formatted Diff | Diff
Patch (16.88 KB, patch)
2010-04-06 23:29 PDT, Eric Seidel (no email) 		abarth: review+
Details | Formatted Diff | Diff
Show Obsolete (2) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Eric Seidel (no email) 2010-03-31 17:50:55 PDT 
Add webkit-patch command to crawl buildbot history and find when tests started to fail
Comment 1 Eric Seidel (no email) 2010-03-31 17:56:34 PDT 
Created attachment 52228 [details]
Patch
Comment 2 Adam Barth 2010-03-31 18:06:46 PDT 
+ raise "Unhandled title: %s" % str(table_title)

We should only be raising Exception objects.  Better to raise an Exception().

+ def _print_blame_information_for_commit(tool, commit_info):

This should be in a base class?  Maybe on CommitInfo?  Having a random global function seems strange.

+ class FailureReason(AbstractDeclarativeCommand):

This class is kind of lame.  It makes it really hard to re-use this logic or to test it.  Maybe we need a FailureWalker class?
Comment 3 Eric Seidel (no email) 2010-04-06 20:26:26 PDT 
Created attachment 52696 [details]
First Try
Comment 4 Eric Seidel (no email) 2010-04-06 20:27:04 PDT 
Updated it so it applies to TOT.  Haven't bothered to abstract out the build-walker stuff.
Comment 5 Eric Seidel (no email) 2010-04-06 23:29:00 PDT 
Created attachment 52704 [details]
Patch
Comment 6 Adam Barth 2010-04-06 23:37:46 PDT 
Comment on attachment 52704 [details]
Patch

Ok.  Let's take this as a starting point.  As you noted in the other bug, there's a lot of overlap with the algorithm the sheriff uses to track recent regressions.  This code looks like it understands more cases (e.g., >= 20 failures).  We should find a way to unify and abstract out the different pieces,but I don't want to hold this patch hostage any longer.
Comment 7 Eric Seidel (no email) 2010-04-06 23:41:18 PDT 
Committed r57195: <http://trac.webkit.org/changeset/57195>