RESOLVED FIXED 135283
REGRESSION(r171526): [SOUP] Massive crashes.
https://bugs.webkit.org/show_bug.cgi?id=135283
Summary REGRESSION(r171526): [SOUP] Massive crashes.
Carlos Alberto Lopez Perez
Reported 2014-07-24 20:46:08 PDT
Revision r171526 <http://trac.webkit.org/r171526> seems to have caused *massive* crashes on the GTK port. All the GTK test bots started to crash for *every* layout test / performance test after this revision. Last good build for the GTK Release test boot (r171525): http://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release%20%28Tests%29/builds/1678 First bad build for the GTK Release test boot (r171526): http://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release%20%28Tests%29/builds/1679 [Exiting early after 47 crashes and 3 timeouts.] And the same for the perf bot or the GTK Debug test boot :\
Attachments
GDB Backtrace for the GTK port when running the perft test Animation/balls.html (17.13 KB, text/plain)
2014-07-24 20:55 PDT, Carlos Alberto Lopez Perez
no flags
Patch (1.34 KB, patch)
2014-07-24 21:16 PDT, Pratik Solanki
no flags
Pratik Solanki
Comment 1 2014-07-24 20:47:16 PDT
Uh oh.. Looking.
Carlos Alberto Lopez Perez
Comment 2 2014-07-24 20:55:43 PDT
Created attachment 235495 [details] GDB Backtrace for the GTK port when running the perft test Animation/balls.html
Carlos Alberto Lopez Perez
Comment 3 2014-07-24 20:57:08 PDT
I have attached a GDB backtrace for the coredump generated when running the perf test Animation/balls.html Relevant part: Program terminated with signal 11, Segmentation fault. #0 0x00007f9c955a6f36 in WebCore::SharedBuffer::append(char const*, unsigned int) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-3.0.so.25 [...] Thread 1 (Thread 0x7f9c87619940 (LWP 6612)): #0 0x00007f9c955a6f36 in WebCore::SharedBuffer::append(char const*, unsigned int) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-3.0.so.25 #1 0x00007f9c95aec465 in WebCore::SharedBuffer::maybeTransferPlatformData() () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-3.0.so.25 #2 0x00007f9c955a6f2b in WebCore::SharedBuffer::append(char const*, unsigned int) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-3.0.so.25 #3 0x00007f9c955a7f11 in WebCore::SharedBuffer::append(WebCore::SharedBuffer*) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-3.0.so.25 #4 0x00007f9c95461ac1 in WebCore::ResourceLoader::addDataOrBuffer(char const*, unsigned int, WebCore::SharedBuffer*, WebCore::DataPayloadType) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-3.0.so.25 [...]
Carlos Alberto Lopez Perez
Comment 4 2014-07-24 21:00:23 PDT
(In reply to comment #0) > > All the GTK test bots started to crash for *every* layout test / performance test after this revision. > After a more detailed look it seems that causes crashes for many layout test, but not for every layout test. It seems however to cause crashes for every performance test.
Pratik Solanki
Comment 5 2014-07-24 21:16:44 PDT
Carlos Alberto Lopez Perez
Comment 6 2014-07-24 21:25:10 PDT
I have tested the attached patch. It fixes the issue. Thanks a lot :)
Pratik Solanki
Comment 7 2014-07-24 21:26:54 PDT
Since this is build breakage, I'll just land it unreviewed. Carlos has confirmed that it fixes the crashes.
Pratik Solanki
Comment 8 2014-07-24 21:28:52 PDT
Csaba Osztrogonác
Comment 9 2014-07-25 00:59:42 PDT
(In reply to comment #8) > Committed r171579: <http://trac.webkit.org/changeset/171579> It fixed the issue on EFL too since EFL also uses SOUP.
Note You need to log in before you can comment on or make changes to this bug.