999496 - Error with gcc 4.9: undefined reference to mozilla::AudioQueueMemoryFunctor::MallocSizeOfBuild

Status: RESOLVED FIXED

(Core :: Audio/Video, defect)

Description

[Brad Jackson]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:26.0) Gecko/20100101 Firefox/26.0 (Beta/Release)
Build ID: 20140421152514

Steps to reproduce:

Compiling trunk builds with gcc 4.9RC was working for a while, and then started giving an undefined reference error. Builds still work with gcc 4.8.2, but are still broken with the gcc 4.9.0 release. It it not clear if this is a Firefox or gcc bug.


Actual results:

../../../content/media/Unified_cpp_content_media1.o: In function `mozilla::AudioData::SizeOfIncludingThis(unsigned long (*)(void const*)) const':
/home/brad/projects/mozilla/content/media/MediaData.h:86: undefined reference to `mozilla::AudioQueueMemoryFunctor::MallocSizeOf(void const*)'
/bin/ld: ../../../content/media/Unified_cpp_content_media1.o: relocation R_X86_64_PC32 against undefined symbol `_ZN7mozilla23AudioQueueMemoryFunctor12MallocSizeOfEPKv' can not be used when making a shared object; recompile with -fPIC
/bin/ld: final link failed: Bad value
collect2: error: ld returned 1 exit status


Expected results:

Build should complete

Comment 1

[Simon Wilper]

I got the same error. What would happen if we do what the error messages "recompile with -fPIC" suggests? I am just investigating on how the build system works to test that out...

Comment 2

[(not currently active) Ted Mielczarek]

That's usually a red herring in my experience. It's almost always a symbol visibility issue.

Comment 3

[Brad Jackson]

I can get the link to complete without error if I move AudioData::SizeOfIncludingThis out of MediaData.h and into MediaData.cpp, like VideoData::SizeOfIncludingThis is currently. Firefox seg faults on startup, but that could be another problem.

Comment 4

[Brad Jackson]

Backtrace of crash on startup. Does not crash in safe mode, but still crashes with all plugins/addons disabled without safe mode.

#0  mozilla::detail::GuardObjectNotificationReceiver::~GuardObjectNotificationReceiver (this=0x349, __in_chrg=<optimized out>)
    at ../../dist/include/mozilla/GuardObjects.h:102
#1  0x00007ffff41323cc in ~AutoValueVector (this=0x2c0, 
    __in_chrg=<optimized out>) at /home/bsjacks/mozilla/js/src/jsapi.h:499
#2  js::jit::BailoutIonToBaseline (cx=cx@entry=0x7ffff7df7560, 
    activation=activation@entry=0x7fffffff4a80, iter=..., 
    invalidate=invalidate@entry=false, bailoutInfo=bailoutInfo@entry=
    0x7fffffff4728, excInfo=excInfo@entry=0x0)
    at /home/bsjacks/mozilla/js/src/jit/BaselineBailouts.cpp:1339
#3  0x00007ffff413268f in js::jit::Bailout (sp=0x7fffffff4730, bailoutInfo=
    0x7fffffff4728) at /home/bsjacks/mozilla/js/src/jit/Bailouts.cpp:91
#4  0x00007ffff7fe6331 in ?? ()
#5  0x00007fffffff4730 in ?? ()
#6  0x00007fffffff4728 in ?? ()
#7  0x00007fffffff4750 in ?? ()
#8  0x0000000000000000 in ?? ()

Comment 5

[Simon Wilper]

Attachment: 999496.patch

Comment 6

[Simon Wilper]

@Brad: I summarized your previous comment in the attached path. Here it worked fine and my browser does not crash.

Comment 7

[Brad Jackson]

Can we get this patch reviewed and committed?

Comment 8

[Matthew Gregan [:kinetik]]

Comment on attachment 8422385 [details] [diff] [review]
999496.patch

Review of attachment 8422385 [details] [diff] [review]:
-----------------------------------------------------------------

It helps if you ask someone for review, otherwise you're relying on a reviewer spotting the bug. :-)

Comment 9

[Matthew Gregan [:kinetik]]

Attachment: Move AudioData::SizeOfIncludingThis to MediaData.cpp

Comment 10

[Matthew Gregan [:kinetik]]

inbound is closed, so marked for later landing.

Comment 11

[Matthew Gregan [:kinetik]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/1f10a8067853

Comment 12

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/1f10a8067853

Comment 13

[Simon Wilper]

@Matthew Thanks. And: I know, I should really read the docs on how to submit patches...