Project: chromium Issues People Development process History Sign in
New issue
Advanced search Search tips
Issue 158923 startup_tests failing on Windows perf bots
Starred by 1 user Project Member Reported by senorblanco@chromium.org, Nov 1, 2012 Back to list
Status: Fixed
Owner: zelidrag@chromium.org
Closed: Jan 2013
Cc: nsylvain@chromium.org, jam...@chromium.org, zelidrag@chromium.org
Components:
OS: Windows
Pri: 2
Type: Bug


Sign in to add a comment
startup_tests is failing on XP and Win7 perf bots.  It first seemed to occur in this range of Chrome revisions:

http://build.chromium.org/f/chromium/perf/dashboard/ui/changelog.html?url=%2Ftrunk%2Fsrc&range=165058%3A165064&mode=html

(from first known bad at http://build.chromium.org/p/chromium.perf/builders/Win7%20Interactive%20Perf/builds/5399).
 
Comment 1 by nsylvain@google.com, Nov 1, 2012
Owner: nsylvain@chromium.org
Status: Assigned
Thanks, I'll see if I can find the source.
Comment 2 by ojan@chromium.org, Nov 1, 2012
Relatedly something made a 25% improvement in shutdown and a 2% regression in virtual memory used (seems like a fine tradeoff!). Looking at the webkit and chromium blamelists I don't see any obvious culprits though and the reference build didn't change, so it shouldn't be a bot configuration change.

Looking at the stdio output that senorblanco linked to this may just be that we're crashing early. :(

http://build.chromium.org/f/chromium/perf/linux-release-webkit-latest/startup/report.html?rev=165309&graph=shutdown&history=150

http://build.chromium.org/f/chromium/perf/linux-release-webkit-latest/intl2/report.html?rev=165353&graph=vm_size_final_b&history=150

http://build.chromium.org/f/chromium/perf/linux-release-webkit-latest/intl2/report.html?rev=165309&graph=vm_size_final_t&trace=vm_size_f_t_extwr&history=150
Comment 3 by ojan@chromium.org, Nov 1, 2012
Actually, maybe not related. The test is still passing on the linux release webkit canary.

http://build.chromium.org/p/chromium.webkit/builders/Linux%20Perf/builds/17652/steps/startup_test/logs/stdio
Labels: OS-Windows
Yeah, this seems to be a Windows-only failure.  Will label that.
Comment 5 by nsylvain@google.com, Nov 1, 2012
I forced a build at revision 165061 (just before zelidrag's change). It will take a while to finish, but that will tell us with more certainty what change is causing the issue.

I'm also trying to build locally to see if I can get a stack in a debugger.

Those IPC errors look interesting too:

[236:976:1030/175515:1244950:ERROR:ipc_channel_win.cc(130)] pipe error: 109

Although I see it in a lot of other open bugs, so perhaps it's unrelated (I know nothing about IPC).
Comment 7 by nsylvain@google.com, Nov 1, 2012
Owner: zelidrag@chromium.org
Revision 165061 passed. So it has to be 165062 or 165063.  I'm going to guess 165062 for now.  Assigning to Zelidrag to take a look.

Zel: If you can revert that would be great, at least we would know for sure if your change is causing those crashes.
Comment 8 by nsylvain@google.com, Nov 1, 2012
Change has been reverted.  Let's see what happens...
Comment 9 by nsylvain@google.com, Nov 1, 2012
The revert seems to have worked. The first bot cycled and is now green.
Labels: Mstone-25 Iteration-68
Comment 11 by k...@google.com, Jan 7, 2013
Labels: -Mstone-25 Mstone-26 MovedFrom-25
Punting non-releaseblocking bugs in M25.  You can find the list via MovedFrom-25
Status: Fixed
I'm marking this as fixed.  If that was in error, feel free to reopen.
Project Member Comment 13 by bugdroid1@chromium.org, Mar 10, 2013
Labels: -Area-Internals -Mstone-26 Cr-Internals M-26
Sign in to add a comment