[Bf-committers] OSX Mavericks 10.9 (mostly) success
info at jensverwiebe.de
Mon Sep 9 21:11:33 CEST 2013
Today looked too into this problem and found some hints:
While gcc/llvm-gcc produced fine U _InstallConnexionHandlers and U _SetConnexionClientButtonMask symbols,
clang aditionally creates U __Z24InstallConnexionHandlersv and U __Z28SetConnexionClientButtonMaskv
wich means demangled exactly the missing symbols while linking ( InstallConnexionHandlers() and SetConnexionClientButtonMask() )
Not exactly sure why this happens atm, but it seems triggered by the __attribute__((weak_import)) flag.
I tried to make these resolvable at runtime ( -Wl,-U,__Z24InstallConnexionHandlersv -Wl,-U,__Z28SetConnexionClientButtonMaskv ), but this
only pushd the problem. Also some hints i found to use -undefined dynamic_lookup is not a good solution here. ( just masks other linking probs if any )
I checked our demangling ( extern "C" ) and all looks fine so far, also in the 3DConexxion headers the #ifdef __cplusplus is set as well.
So atm. i am stuck with this odd c++ symbols occuring only with clang )
Have no idea tonight, but will try to catch my linking-expert soon.
Am 09.09.2013 um 20:16 schrieb Mike Erwin <significant.bit at gmail.com>:
> (sorry if this is a repeat post, I got a bounce alert from the list server
> a few days ago)
> Thanks for taking initiative on this! Looks like we're in pretty good shape
> when 10.9 ships. I'll look into the 3Dconnexion build issue soon -- I'd
> forgotten about it since I always have their driver installed.
> Yeah, it's weird that -weak_framework stopped working (was fine with gcc &
> friends) but there are other ways to do what we need. Will send an update
> when it's building fine with & without the driver.
> Mike Erwin
> musician, naturalist, pixel pusher, hacker extraordinaire
> Bf-committers mailing list
> Bf-committers at blender.org
Allerskehre 44 - 22309 Hamburg
Tel.: +49 40 68 78 50
mobil: +49 172 400 49 07
mailto: info at jensverwiebe.de
More information about the Bf-committers