RESOLVED INVALID 86467
[WK2] crash detected by mistake on Qt-WK2-Linux 
https://bugs.webkit.org/show_bug.cgi?id=86467
Summary [WK2] crash detected by mistake on Qt-WK2-Linux
Balazs Kelemen
Reported 2012-05-15 04:57:06 PDT
This has shown up in connection with https://bugs.webkit.org/show_bug.cgi?id=80691. After I landed the patch that implements plugins with Qt 5 we experiences crashes in some compositing tests. I believe that nrwt detect these as crashes by mistake because of the followings reasons: 1. the crash log produced by nrwt is quite strange: crash log for WebProcess (pid <unknown>): STDOUT: <empty> STDERR: <empty> 2. I could not get any core dump for the tests I tried to debug this via core dumps, because the crashes only occur if you run all the tests, but no dump has generated. Obviously I set up the env for this, i.e. ulimit -c unlimited + export QT_WEBKIT2_DEBUG=1 (the latter makes the Qt WTR to not to install signal handlers). 3. There are no crash with ORWT
Attachments
Add attachment proposed patch, testcase, etc.
Csaba Osztrogonác
Comment 1 2012-05-15 06:55:20 PDT
I managed to reproduce it with ORWT too, so it isn't NRWT bug. You can reproduce it simple if you pass the tests_run0.txt to ORWT --nthly 99999 . (Because ORWT runs tests in different oreder and restart DRT/WTR after 1000 tests if you don't override it with --nthly option) ORWT and NRWT handles WebProcess crashes in the same way, they only parse the output of WTR. If there is any "#CRASHED - WebProcess" line, it reports that WebProcess crashed. So maybe the bug is in WTR.
Balazs Kelemen
Comment 2 2012-05-16 05:50:32 PDT
The real bug is that sometimes we don't differentiate between web process timeout vs. crash. Indeed, this is not an nrwt issue.
Note You need to log in before you can comment on or make changes to this bug.