[FFmpeg-devel] Hacks on FATE

[Måns Rullgård]

Michael Kostylev <michael.kostylev at gmail.com> writes:

> fix building with i386-*-gcc-3.3 -fPIC
>
> --- ffmpeg.orig/libavcodec/x86/cavsdsp_mmx.c
> +++ ffmpeg/libavcodec/x86/cavsdsp_mmx.c
> @@ -280,7 +280,7 @@
>          VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
>          \
>          : "+a"(src), "+c"(dst)\
> -        : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride), "m"(ADD), "m"(MUL1), "m"(MUL2)\
> +        : "S"((x86_reg)srcStride), "rm"((x86_reg)dstStride), "m"(ADD), "m"(MUL1), "m"(MUL2)\
>          : "memory"\
>       );\
>       if(h==16){\
> @@ -295,7 +295,7 @@
>              VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
>              \
>             : "+a"(src), "+c"(dst)\
> -           : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride), "m"(ADD),  "m"(MUL1), "m"(MUL2)\
> +           : "S"((x86_reg)srcStride), "rm"((x86_reg)dstStride), "m"(ADD),  "m"(MUL1), "m"(MUL2)\
>             : "memory"\
>          );\
>       }\

An 'm' constraint requires an lvalue operand (or ought to do (llvm
does)), which a cast expression is not.

> fix building with i386-*-gcc>=3.4 -fPIC
>
> --- ffmpeg.orig/libavcodec/x86/h264dsp_mmx.c
> +++ ffmpeg/libavcodec/x86/h264dsp_mmx.c
> @@ -973,7 +973,7 @@
>          "add %4, %1                 \n\t"\
>          "decl %2                    \n\t"\
>          " jnz 1b                    \n\t"\
> -        : "+a"(src), "+c"(dst), "+g"(h)\
> +        : "+a"(src), "+c"(dst), "+m"(h)\
>          : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
>          : "memory"\
>      );\
> @@ -1164,7 +1164,7 @@
>          "add %4, %1                 \n\t"\
>          "decl %2                    \n\t"\
>          " jnz 1b                    \n\t"\
> -        : "+a"(src), "+c"(dst), "+g"(h)\
> +        : "+a"(src), "+c"(dst), "+m"(h)\
>          : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
>          : "memory"\
>      );\
> @@ -1649,7 +1649,7 @@
>          "add %4, %1                 \n\t"\
>          "decl %2                    \n\t"\
>          " jnz 1b                    \n\t"\
> -        : "+a"(src), "+c"(dst), "+g"(h)\
> +        : "+a"(src), "+c"(dst), "+m"(h)\
>          : "D"((x86_reg)srcStride), "S"((x86_reg)dstStride),\
>            "m"(ff_pw_5), "m"(ff_pw_16)\
>          : "memory"\

A 'g' constraint allows a memory operand, so this is definitely a bug
in gcc.  Needless to say, the patch is unacceptable.

Is no more recent gcc version available for OpenBSD?  If yes, I would
suggest officially not supporting the older versions that fail.  If
not, I suggest someone at OpenBSD get a clue and use it.

-- 
M?ns Rullg?rd
mans at mansr.com