Issue metadata
Sign in to add a comment
|
win_chromium_rel_ng interrupted exception |
||||||||||||||||||||||||
Issue descriptionTrybot win_chromium_rel_ng (https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng?numbuilds=200) is often running into exception. Sometimes the exception comes when the bot has finished running everything and stays at "recipe_results" status forever until interrupted exception occurs. Sometimes the exception comes when webkit_tests are running into TEST_INVALID_RESULTS. Example builds: https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/240949 https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/240936 https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/240977 This particular bot is currently blocking CLs that require running webkit_tests from landing.
,
Jun 17 2016
,
Jun 17 2016
-> current trooper
,
Jun 17 2016
,
Jun 17 2016
My current hypothesis is that WerFault.exe window (http://1.bp.blogspot.com/-GciIoIvrPLc/Vk7gSorf0hI/AAAAAAAAAas/6C9KNgTpv5c/s1600/werfault-exe-02.png) blocks recipe engine from exiting somehow. In all builds above webkit_tests lists "C:\Windows\SysWOW64\WerFault.exe" as one of the running processes. I thought we disable Windows Error Reporting on bots... Looks like it isn't working. I'll try to figure out: 1) Why Error Reporting process appears. 2) Why it block recipe engine. (Also I think it's a dup of 619742, though in 619742 webkit unit tests specifically get stuck).
,
Jun 17 2016
Oh, this is old and hairy problem :( crbug.com/522396
,
Jun 17 2016
Moving to this other bug. There's lot of info there.
,
Jun 17 2016
Hmmm.Are you sure it's a duplicate? The old bug has been there fore a very long time. And I encountered this webkit_tests not finishing only in these few days. At the current latest builds, I couldn't find anyone ends up with a green bot as long as it's running webkit tests.
,
Jun 17 2016
I'm still looking, but so far symptoms look consistent with 522396. Probably something increased the frequency of WerFault.exe happening recently. (Here's recent green build with webkit_tests https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/241129)
,
Jun 18 2016
vadimsh@: Hi, from the build link you provided, I didn't see it runs webkit_tests...It's a long build but it doesn't run webkit_tests...All those running webkit_tests end up being like this: https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/241165 In fact, if you look at the list of "Currently building", those builds that are stagnated at "recipe result" status are all those builds that run webkit tests..Wait for 1 hr and those builds will certain run into interrupted exception...
,
Jun 18 2016
Err.. I probably copy-pasted wrong build. Here's a green one 1h ago: https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/241135 Generally, commits are landing in third_party/WebKit, which means some builds are passing. --- > In fact, if you look at the list of "Currently building", those builds that are stagnated at "recipe result" status are all those builds that run webkit tests..Wait for 1 hr and those builds will certain run into interrupted exception... Yes. In all of them, if you open "webkit_tests" step, you'll see (at the bottom) it lists "C:\Windows\SysWOW64\WerFault.exe -u -p 5824 -s 2368" process. This process stays after the tests and blocks buildbot from collecting step results. See crbug.com/522396 for details. |
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by xlai@chromium.org
, Jun 17 2016