(Cc'ing xine-lib maintainers) Trying to play back something with "fbxine -V DirectFB" on F-14 results in an immediate stack smashing crash: (*) DirectFB/Core: Single Application Core. (2010-11-29 16:48) (*) Direct/Memcpy: Using libc memcpy() (*) DirectFB/FBDev: Found 'MATROX DH' (ID 0) with frame buffer at 0xea800000, 8192k (MMIO 0xfe7fc000, 16k) (*) Direct/Modules: suppress module 'keyboard' (*) Direct/Modules: suppress module 'linux_input' (*) Direct/Thread: Started 'PS/2 Input' (-1) [INPUT OTHER/OTHER 0/0] <8388608>... (*) DirectFB/Input: IMPS/2 Mouse 1.0 (directfb.org) (*) DirectFB/Genefx: MMX detected and enabled (*) DirectFB/Graphics: Matrox G450 0.7 (directfb.org) (*) DirectFB/Core/WM: Default 0.3 (directfb.org) *** stack smashing detected ***: fbxine terminated ======= Backtrace: ========= /lib64/libc.so.6(__fortify_fail+0x37)[0x7f9a0653a1b7] /lib64/libc.so.6(__fortify_fail+0x0)[0x7f9a0653a180] /usr/lib64/xine/plugins/1.29/xineplug_vo_out_directfb.so(+0x26b6)[0x7f99fddef6b6] /usr/lib64/libxine.so.1(_x_load_video_output_plugin+0xbd)[0x7f9a06ea357d] /usr/lib64/libxine.so.1(xine_open_video_driver+0x9)[0x7f9a06ea3639] fbxine[0x40334d] /lib64/libc.so.6(__libc_start_main+0xfd)[0x7f9a0645ae7d] fbxine[0x402d49] ======= Memory map: ======== [...] $ rpm -q directfb xine-lib xine-lib-extras xine-ui directfb-1.4.11-2.fc14.x86_64 xine-lib-1.1.19-2.fc14.1.x86_64 xine-lib-extras-1.1.19-2.fc14.1.x86_64 xine-ui-0.99.6-25.fc14.x86_64 This is on x86_64, dunno about other archs. Fix: revert changes between directfb-1.4.11 -1 and -2 (the "abi" patch) and rebuild xine-lib against it. Tested locally, works. Causes the pain of soname bump, though. I also tried just rebuilding xine-lib against directfb-1.4.11-2.fc14 but that didn't help.
(In reply to comment #0) .. > Fix: revert changes between directfb-1.4.11 -1 and -2 (the "abi" patch) and > rebuild xine-lib against it. Tested locally, works. Causes the pain of soname > bump, though. > > I also tried just rebuilding xine-lib against directfb-1.4.11-2.fc14 but that > didn't help. Hello Vile, It sounds like contradictory. To me, rpmsodiff showed no ABI change between the previous version of directfb which I suppose was used to build xine-lib. Now maybe something bellow directfb has change it's ABI in which that that could explain a rebuild solving this case. Can you recreate your test case with the current version of directfb-1.4.11-2 just rebuilt ? Thx for you report.
Tested, does not help. Are you able to play something with fbxine -V DirectFB without a crash?
With the ABI change, the module directory is changing too, but maybe others part have it hardcoded. Can you give a try with create a symlink from %{_libdir}/directfb-1.4-5/ pointing to %{_libdir}/directfb-1.4-0/ Once that said, the Backtrace seems related to the fortify feature. I never succeed in using directfb on hardware that are KMS capable. Last that has succeed was with viafb, witch was the one I tested directfb on. I need some time to retest on this target. In other cases fbxine -V DirectFB doesn't produce a crash with me.(It just shows a black screen). I reverted the ABI change in Rawhide for now, still investigating for el6 and fc14.
I tried a few things over the weekend already, and nothing except the rebuild I mentioned helped. Given that the only packages in Fedora and RPM Fusion that depend on directfb are xine-lib-extras, vlc-nox and vlc-extras, I think it would be quite acceptable to just revert the abi patch and rebuild these also for F-14 and maybe even EL-6 - I have no interest in trying to pursue further hacks. The only system where I use directfb (my PVR) is already running my local rebuilds that revert the ABI changes, and they have some non-Fedora stuff already built on top of the "reverted" directfb sonames so switching back would be more work than I think it's worth for me.
If the directfb and xine-lib maintainers ack it, I can take care of reverting the ABI patch for directfb and rebuilding xine-lib against it for F-14.
No objections from the xine-lib side. (The ABI-hacked DirectFB apparently isn't working at all, so we should just use the upstream ABI and rebuild xine-lib against it.)
directfb-1.4.11-3.fc14 was built droping the ABI patch. Unfortunately this doesn't put a light on what's happened in the way upstream mess with uneeded ABI changes...
I tested that build locally against my local xine-lib rebuild and it works. Buildroot override requested: https://fedorahosted.org/rel-eng/ticket/4529
In the absence of an ABI-patched version that actually WORKS, I wouldn't be so fast to call upstream's ABI changes "unneeded"…
directfb-1.4.11-3.fc14,xine-lib-1.1.19-2.fc14.2 has been submitted as an update for Fedora 14. https://admin.fedoraproject.org/updates/directfb-1.4.11-3.fc14,xine-lib-1.1.19-2.fc14.2
directfb-1.4.11-3.fc14, xine-lib-1.1.19-2.fc14.2 has been pushed to the Fedora 14 stable repository. If problems still persist, please make note of it in this bug report.