I'm bisecting the culprit.
It seems http://trac.webkit.org/changeset/146174 is the culprit or one of its buildfix: http://trac.webkit.org/changeset/146178 or http://trac.webkit.org/changeset/146195 See r146178 results with applied the fix from r146195: http://cute1.inf.u-szeged.hu:8010/builders/ARMv7%20Linux%20Qt5%20Release%20%28Test%29/builds/8093
The bug is valid with Thumb2 assembler too.
Oh no, thumb2 result can be invalid, because the fix from https://bugs.webkit.org/show_bug.cgi?id=103146 haven't been landed.
(In reply to comment #3) > Oh no, thumb2 result can be invalid, because the fix from https://bugs.webkit.org/show_bug.cgi?id=103146 haven't been landed. On r146178 + fix from r146195 + fix from bug103146, all inspector tests pass with Thumb2, so only ARM traditional is affected by this bug.
Not r146174 is the culprit, but http://trac.webkit.org/changeset/146164. r146146 + r146195 (buildfix): http://build.webkit.sed.hu/builders/ARMv7%20Linux%20Qt5%20Release%20%28Test%29/builds/8096
I had to skip _all_ inspector test (r146275) to make the ARM bot work again. (We can't wait for ~500x30 secs timeout)
inspector tests pass with disabled DFG JIT: http://build.webkit.sed.hu/builders/ARMv7%20Linux%20Qt5%20Release%20%28Test%29/builds/8209 So r146146 broke the ARM traditional with enabled DFG JIT only. :)
Fixed by https://trac.webkit.org/changeset/150748 and tests were unskipped by https://trac.webkit.org/changeset/150794 *** This bug has been marked as a duplicate of bug 116306 ***