Summary: | REGRESSION (r125126): Multiple crashes introduced in GTK debug builds | ||
---|---|---|---|
Product: | WebKit | Reporter: | Zan Dobersek <zan> |
Component: | UI Events | Assignee: | Nobody <webkit-unassigned> |
Status: | RESOLVED DUPLICATE | ||
Severity: | Normal | CC: | abarth, dglazkov, dominicc, eric, hayato |
Priority: | P2 | ||
Version: | 528+ (Nightly build) | ||
Hardware: | Unspecified | ||
OS: | Unspecified |
Description
Zan Dobersek
2012-08-10 10:10:40 PDT
Okay. Let me take a look. It might take some time for me to set up GTK build environments... I could reproduce some of them on my local environment. Let me investigate further. (In reply to comment #2) > It might take some time for me to set up GTK build environments... GTK build takes too much time for me. Let me continue tomorrow. I've not found out the cause yet. It's okay for me to revert r125133 if it is absolutely needed. (In reply to comment #3) > I could reproduce some of them on my local environment. Let me investigate further. > > (In reply to comment #2) > > It might take some time for me to set up GTK build environments... use-instanceRoot-as-event-target.xhtml is probably related to r125251. I confirmed that the crash in #1 happend on the revision before r125133 on my local environment. I am afraid that we should bisect to find the cause. (In reply to comment #4) > GTK build takes too much time for me. Let me continue tomorrow. I've not found out the cause yet. > > It's okay for me to revert r125133 if it is absolutely needed. > > (In reply to comment #3) > > I could reproduce some of them on my local environment. Let me investigate further. > > > > (In reply to comment #2) > > > It might take some time for me to set up GTK build environments... I'd like to note that crash is very flaky. Hard to reproduce on my local gtk build environment. It makes bisect very difficult. Flakiness dashboard told me that limited recent results. Is there any way to see more previous results? (In reply to comment #7) > I'd like to note that crash is very flaky. Hard to reproduce on my local gtk build environment. It makes bisect very difficult. > Flakiness dashboard told me that limited recent results. Is there any way to see more previous results? You can check the 'Show all runs' checkbox in the top right of the flakiness dashboard. That shows that on the 64-bit release bot, the tests first started failing in the revision range r125121-r125128. That range is covered by the following builds: http://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release/builds/27330 http://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release/builds/27331 Unfortunately, the first buildbot cycle didn't make it through because of dependencies update failure. Other than that, it seems the regression is hard to find because of unfortunate combination of bad patches and outside factors. At least the 64-bit debug builder is clear that the regression started occurring somewhere in between r12103 and r125133: http://build.webkit.org/builders/GTK%20Linux%2064-bit%20Debug?numbuilds=200 This irritates me well enough that I'll take a look at it in the CET afternoon if you don't find the offending commit by then. Thanks for the effort though, much appreciated! The bisection outlined r125126 as the offending commit. http://trac.webkit.org/changeset/125126 CC-ing proper people. This is on my list for today. *** This bug has been marked as a duplicate of bug 93654 *** |