[MPlayer-users] Problem building mplayer in ffmpeg from latest svn

[Erik Auerswald]

Hi George,

On 28.09.22 22:29, George R Goffe wrote:
> Hi,
> 
> I am trying to build mplayer on a fedora core 38 x86_64 system but it's failing in the ffmpeg snapshot directory. I have the "EXACT" verson of ffmpeg from their source repository and it builds with "NO" changes except for --prefix=/usr/lsd/Linux which is what I'm using to build mplayer.
> 
> I am at a loss as to why this build is failing from the mplayer snapshot of ffmpeg and my separate build directory.

A build of ffmpeg uses the build system of ffmpeg ("./configure")
to determine which system features are available and which are not.

But building ffmpeg inside the MPlayer source tree does not do
this.  Instead the MPlayer build system (also "./configure") checks
for features used by MPlayer itself, and additionally sets some for
ffmpeg as "available" or "not available" while others are not set
at all.

> Attached below here is a sample of the error messages I'm seeing as well as (I hope) a complete list of ffmpeg build results. Can anyone give this a peek and point me in the "right" direction please?
> 
> I have a full build log available. I have attempted to build with the "current" released gcc as well as the latest (sort of) version of gcc. Both fail in the same manor.
> 
> Best regards,
> 
> George...
> 
> 
> libavformat/dashenc.c:1960:13: note: 'snprintf' output between 1 and 2047 bytes into a destination of size 1024
> 1960 |             snprintf(os->full_path, sizeof(os->full_path), "%s%s", c->dirname, os->initfile);
> |             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> In file included from ./libavutil/aarch64/cpu.h:23,
> from libswscale/utils.c:52:
> libswscale/utils.c: In function 'initFilter':
> ./libavutil/cpu_internal.h:27:6: error: 'HAVE_LASX' undeclared (first use in this function); did you mean 'HAVE_AVX'?

It seems to me as if the "HAVE_LASX" feature flag is not set at all
by MPlayer, but ffmpeg's ./configure would have set it to a value.

> 27 |     (HAVE_ ## cpuext ## suffix && ((flags) & AV_CPU_FLAG_ ## cpuext))
> |      ^~~~~
> ./libavutil/cpu_internal.h:43:31: note: in expansion of macro 'CPUEXT_SUFFIX'
> 43 | #define CPUEXT(flags, cpuext) CPUEXT_SUFFIX(flags, , cpuext)
> |                               ^~~~~~~~~~~~~
> ./libavutil/loongarch/cpu.h:29:26: note: in expansion of macro 'CPUEXT'
> 29 | #define have_lasx(flags) CPUEXT(flags, LASX)
> |                          ^~~~~~
> libswscale/utils.c:663:9: note: in expansion of macro 'have_lasx'
> 663 |     if (have_lasx(cpu_flags)) {
> |         ^~~~~~~~~
> ./libavutil/cpu_internal.h:27:6: note: each undeclared identifier is reported only once for each function it appears in
> 27 |     (HAVE_ ## cpuext ## suffix && ((flags) & AV_CPU_FLAG_ ## cpuext))
> |      ^~~~~
> ./libavutil/cpu_internal.h:43:31: note: in expansion of macro 'CPUEXT_SUFFIX'
> 43 | #define CPUEXT(flags, cpuext) CPUEXT_SUFFIX(flags, , cpuext)

I am sorry that I cannot tell you how to fix this.

Best regards,
Erik