ao_alsa used static variables to define the parameter values for
snd_pcm_hw_params_set_buffer_time_near() and
snd_pcm_hw_params_set_periods_near(). The variables were non-const and
the desired value was only set in the initializer. The ALSA functions
in question take a pointer argument to an in/out parameter which is
modified to reflect the value actually used. As a result, when playing
multiple files or otherwise reinitializing the AO the later instances
could use values that had been modified by earlier calls. Change the
code to always always use the same default values.
A couple of months ago MPlayer's ALSA driver started rounding the
amount of input data it was willing to accept in one call down to an
integer multiple of the value it set in ao_data.outburst. In some
configurations it was possible for this value to exceed the 64 KiB
limit on the amount MPlayer was willing to write in a single call to
the AO. As a result ao_alsa accepted 0 bytes in each play() call and
audio playback failed. Fix this by removing the fixed 64 KiB limit on
the amount of audio sent to AO at once; the limit was mostly a remnant
of older code anyway.
Revert 3 old code uglification changes that were done with the excuse
of gcc-2.95 support. The last reverted change was a fix to a bug
introduced in the middle change.
Revert "10l, len may change after initialization time"
This reverts commit ae9db277c7.
Revert "fix declaration after statement, take 2"
This reverts commit 4bceedee93.
Revert "fix declaration after statement"
This reverts commit aef0374c1c.
When using libass to render plaintext (non-SSA/ASS) subtitles the code
in update_subtitles() still called set_osd_subtitle() in one case,
causing the global vo_sub variable containing non-libass subtitles to
be set. Under some circumstances this resulted in both a
libass-rendered and non-libass-rendered version of the same subtitle
appearing on screen. Fix by running the subtitle clearing code (which
called set_osd_subtitle) only when libass is not used.
variable instead.
should fix seeking after starting playback with -force-idx broken by
r29914.
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@30052 b3059339-0415-0410-9bf9-f77b7e298cf2
video surfaces than absolutely necessary.
I will revert if somebody reports real-world resource issues introduced
by this commit.
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@30050 b3059339-0415-0410-9bf9-f77b7e298cf2
Use the language variable directly in the template instead of passing
the language variable as a parameter to $(call).
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@30045 b3059339-0415-0410-9bf9-f77b7e298cf2
Based on a patch by Tim Wojtulewicz (timwoj at ieee dot org).
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@30043 b3059339-0415-0410-9bf9-f77b7e298cf2
Some code used an invalid '%lf' conversion specification for double
arguments. Maybe they were written that way due to confusion with
scanf where doubles are indicated by '%lf'; however it is not a valid
printf format specifier. Change those cases to use '%f'.
Thus those files can be symlinked directly without testing if they exist.
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@30040 b3059339-0415-0410-9bf9-f77b7e298cf2
Instead, employ make syntax to generate the shell command arguments.
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@30037 b3059339-0415-0410-9bf9-f77b7e298cf2
The feature was never properly used and provides little benefit.
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@30020 b3059339-0415-0410-9bf9-f77b7e298cf2