[VDPAU] Testcase for the libvdpau dlopen segfault bug
swarren at nvidia.com
Wed Apr 4 09:57:16 PDT 2012
Aaron Plattner wrote at Wednesday, April 04, 2012 9:42 AM:
> On 03/14/2012 06:45 AM, Christian König <deathsimple at vodafone.de> wrote:
> > On 13.03.2012 17:45, Anssi Hannula wrote:
> >> As it might be of interest, here's a testcase I wrote last week for the
> >> issue that calling vdp_device_create_x11() will cause a callback to be
> >> set that will be run at XCloseDisplay() time, causing a crash if
> >> libvdpau has been dlclose()d already when XCloseDisplay() gets called.
> >> There's nothing new in it but it can save some testing time if someone
> >> eventually takes a closer look.
> >> The segfault happens both when vdp_device_create_x11() is successful and
> >> when it fails (e.g. on any non-VDPAU-supporting system/driver).
> >> This is Debian bug #592204, NVIDIA internal bug #770623.
> >> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=592204
> >> This has in the past affected at least mplayer and XBMC, though current
> >> mplayer doesn't dlopen() libvdpau anymore and current XBMC has a
> >> workaround for this (it never dlclose()s libvdpau). This affects
> >> libvdpau 0.4+ (that version added the dri stuff).
> > Well that reminds me that I wanted to ask if we already have an unit
> > testing suite for VDPAU?
> > Something like testing the different rendering features, decoding sample
> > streams and compare them to software decoded reference frames etc...
> > E. g. something like piglit just for VDPAU?
> There isn't one. This might be a good excuse to learn gtest.
It isn't well publicized, but there is the traceplay application. I think
the version we published includes all the CRC checking stuff. It's not so
much a unit test as a regression test though:
More information about the VDPAU