Created attachment 157995 [details] build log from the bot There are an incremental build bug on the Qt bots intermittently. (~ once a week) Unfortunately it's impossible to reproduce it. But now I have a build log for it and I saved the full WebKitBuild directory to help to find the root of the bug.
Created attachment 157999 [details] build log from the bot - same revision, second build, success
Created attachment 158002 [details] diff between WebKitBuild directories left side (-): broken build right side (+): second and working build
I found many build failures which can be related to this bug: - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20NRWT/builds/15427/steps/compile-webkit/logs/stdio/text - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20NRWT/builds/17506/steps/compile-webkit/logs/stdio/text - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20NRWT/builds/17673/steps/compile-webkit/logs/stdio/text - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20NRWT/builds/17795/steps/compile-webkit/logs/stdio/text - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20NRWT/builds/17978/steps/compile-webkit/logs/stdio/text It's strange that all of these build failures are on our "x86-32 Linux Qt Release NRWT" bot, which builds on 4 threads. But I can't find any build failure on our "x86-64 Linux Qt Release" bot which builds on 8 threads.
More build failures from the "x86-32 Linux Qt Release - Qt5-WebKit1" bot: - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20-%20Qt5-WebKit1/builds/9610/steps/compile-webkit/logs/stdio - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20-%20Qt5-WebKit1/builds/9888/steps/compile-webkit/logs/stdio - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20-%20Qt5-WebKit1/builds/10019/steps/compile-webkit/logs/stdio - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20-%20Qt5-WebKit1/builds/10131/steps/compile-webkit/logs/stdio - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20-%20Qt5-WebKit1/builds/10417/steps/compile-webkit/logs/stdio - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20-%20Qt5-WebKit1/builds/10548/steps/compile-webkit/logs/stdio - http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20-%20Qt5-WebKit1/builds/10578/steps/compile-webkit/logs/stdio This bot builds with -j4 too.
Do we need more build logs from other bots?
One more occurance: http://build.webkit.sed.hu/builders/ARMv7%20Linux%20Qt5%20Release%20%28Build%29/builds/10491/steps/compile-webkit/logs/stdio
One more different strange build fail: http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20NRWT/builds/18644/steps/compile-webkit/logs/stdio
Two more occurances of this bug: - http://build.webkit.sed.hu/builders/ARMv7%20Linux%20Qt5%20Release%20%28Build%29/builds/10561/steps/compile-webkit/logs/stdio - http://build.webkit.sed.hu/builders/ARMv7%20Linux%20Qt5%20Release%20%28Build%29/builds/10564/steps/compile-webkit/logs/stdio
One more fail, now on the EWS bot: - bug: https://bugs.webkit.org/show_bug.cgi?id=95382#c5 - build log: http://queues.webkit.org/results/13689448 It would be great to fix this crazy bug as soon as possible ... Or else nobody will trust in Qt buildbots and EWS bot. :-(
The "make[4]: *** No targets. Stop." makes me wonder if this is perhaps the result of a corrupted make file due to two qmake instances writing the same file. That would also explain why the likely-hood of this is somewhat related to the number of parallel jobs.
Maybe it is related to https://trac.webkit.org/changeset/121966
One more fail: http://build.webkit.sed.hu/builders/x86-32%20Linux%20Qt%20Release%20NRWT/builds/18968/steps/compile-webkit/logs/stdio
Taking
Created attachment 162231 [details] Patch
Comment on attachment 162231 [details] Patch r=me, thanks for the fix.
Committed r127583: <http://trac.webkit.org/changeset/127583>