[cairo-bugs] [Bug 35122] QNX 6 configure diff

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=35122

--- Comment #6 from seanb at qnx.com 2011-03-10 07:21:57 PST ---
(In reply to comment #5)
> (In reply to comment #4)
> > The issue is this section:
> > 
> > configure:33670: checking whether cairo's cairo-trace feature could be enabled
> > configure:33674: result: no (requires dynamic linker and zlib and real
> > pthreads)
> 
> Right, you did say that it was to get the trace feature working. Okay, so
> here's what I think is happening:  cairo-trace needs real pthreads for
> pthread_once() and friends, but our configure test doesn't think libc provides
> for real pthreads. This is because many other platforms have weakly linked stub
> implementations of the pthread primitives for efficiency reasons in case the
> application isn't actually multithreaded.  A pure link test against libc
> doesn't work on those platforms as it will give a false positive signal for
> pthreads being available in libc when in reality the libc implementation will
> error out when it comes time to do something actually thready.  This makes it
> impossible to test for real pthread support in libc without actually running
> some code and seeing if it fails.
> 
> Your patch works around the configuring logic by making it believe that the
> user set PTHREAD_LIBS. Our basic bug there is that we consider
> PTHREAD_LIBS/CFLAGS=""
> in the environment to be unset, even though the user has explicitly told us not
> to use any flags or libs at all.  Our deeper bug is that we don't trust libc to
> implement pthreads properly, but I'm not sure we can fix that without some old
> fashioned "if <your OS>" conditionals.
> 
> The reason cairo-trace utility needs to use real pthreads because it would be a
> disaster if the application were actually threaded, but cairo trace used libc
> stubs (this may be an issue due to cairo-trace LD_PRELOADING the interposer
> libcairo-tracer.so.)  On the other hand, using cairo-trace on a threaded
> application seems to be a disaster anyway right now, so changing the
> cairo-trace requirement from real pthreads might not make a practical
> difference. :)

So does that mean you'll accept the patch?

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.