On 21/07/15 5:16 PM, Michael Niedermayer wrote: > On Tue, Jul 21, 2015 at 09:18:58PM +0200, Andreas Cadhalpun wrote: >> Signed-off-by: Andreas Cadhalpun <andreas.cadhal...@googlemail.com> >> --- >> libavcodec/bswapdsp.c | 4 ++++ >> libavcodec/bswapdsp.h | 1 + >> libavcodec/h264pred.c | 7 +++++++ >> libavcodec/h264pred.h | 2 ++ >> libavcodec/h264qpel.c | 5 +++++ >> libavcodec/h264qpel.h | 1 + >> tests/checkasm/bswapdsp.c | 2 +- >> tests/checkasm/h264pred.c | 2 +- >> tests/checkasm/h264qpel.c | 2 +- >> 9 files changed, 23 insertions(+), 3 deletions(-) > > ok, but > avpriv would imply that the ABI doesnt change (incompatibly) without > soname bump > so either there would be the gurantee that the various H264 DSP > contexts wont change in a incompatible way or > it should be documented that these avpriv functions (or exported > ff_ functions) are an exception and only to be used by the selftest > code build from a matching checkout
Lets just make checkasm depend on static builds for the time being, until someone comes up with a better solution. Worst case scenario, if we end up adding avpriv functions for this then they should probably be like avpriv_float_dsp_alloc() and similar to avoid what you mentioned. _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
