606179 - maemo builders inconsistent

Status: RESOLVED INVALID

(Release Engineering :: General, defect)

Description

[Brad Lassey [:blassey] (use needinfo?)]

When I pushed the android breakpad patches, half the maemo builders turned red while the other half built successfully. Specifically ARM_sp and ARM_ip weren't defined on the ones that burned. My local sdk had the same problem, <sys/ptrace.h> doesn't include <asm/ptrace.h> which is where ARM_* is defined.

burning builders:
http://tinderbox.mozilla.org/showlog.cgi?log=Mobile/1287628523.1287631067.30023.gz
http://tinderbox.mozilla.org/showlog.cgi?log=Mobile/1287625133.1287625394.3995.gz
http://tinderbox.mozilla.org/showlog.cgi?log=Mobile/1287625681.1287626798.10705.gz

non-burning builders:
http://tinderbox.mozilla.org/showlog.cgi?log=Mobile/1287625133.1287626159.7757.gz
http://tinderbox.mozilla.org/showlog.cgi?log=Mobile/1287625681.1287626022.6862.gz
http://tinderbox.mozilla.org/showlog.cgi?log=Mobile/1287628523.1287628853.20315.gz

Comment 1

[John Ford [:jhford] CET/CEST Berlin Time]

slaves:

burning:
moz2-linux-slave38 - maemo5-gtk
mv-moz2-linux-ix-slave15 - maemo5-gtk
mv-moz2-linux-is-slave20 - maemo5-gtk

non-burning
mv-moz2-linux-ix-slave16 - maemo5-qt
mv-moz2-linux-ix-slave19 - maemo5-qt
mv-moz2-linux-ix-slave19 - maemo5-qt

Comment 2

[John Ford [:jhford] CET/CEST Berlin Time]

Burning:
[cltbld@mv-moz2-linux-ix-slave16 FREMANTLE_ARMEL]$ sha1sum usr/include/asm/ptrace.h
9c72e6521557281453946b8773f42d306560bba0  usr/include/asm/ptrace.h
[cltbld@mv-moz2-linux-ix-slave16 FREMANTLE_ARMEL]$ sha1sum usr/include/sys/ptrace.h
f819762f2d25476f3827bff773c45e818af302df  usr/include/sys/ptrace.h

Non-burning:
[cltbld@mv-moz2-linux-ix-slave15 FREMANTLE_ARMEL]$ sha1sum usr/include/asm/ptrace.h 
9c72e6521557281453946b8773f42d306560bba0  usr/include/asm/ptrace.h
[cltbld@mv-moz2-linux-ix-slave15 FREMANTLE_ARMEL]$ sha1sum usr/include/sys/ptrace.h 
f819762f2d25476f3827bff773c45e818af302df  usr/include/sys/ptrace.h

It looks like these files are the same on both passing and failing slaves.  It also looks like this is something that is dependent on whether you are building with GTK or QT.  We currently build QT without the crashreporter (http://hg.mozilla.org/build/buildbot-configs/file/ca43eb7d3c8e/mozilla2/mobile/maemo5-qt/mobile-browser/nightly/mozconfig#l13).

Have you tried locally with the crash reporter enabled for QT?  Does it fail the same way?

Comment 3

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

from irc, jhford is already looking at this.

Comment 4

[(not currently active) Ted Mielczarek]

Oh, uh, comment 2 would explain it then. The Qt builds didn't burn because they...weren't building this code.

Comment 5

[Brad Lassey [:blassey] (use needinfo?)]

That certainly explains it, if we're not building crash reporter (which we're not on qt) then we wouldn't hit this bustage. 

Perhaps QT and Gtk builds should be on separate lines in tbpl to avoid this confusion in the future. This bug is invalid though.