"Ronald S. Bultje" <[email protected]> writes:

> Hi,
>
> On Sun, May 15, 2011 at 12:16 PM, Diego Biurrun <[email protected]> wrote:
>> This resolves circular header dependencies and fixes 'make checkheaders'.
>> ---
>>  libavcodec/arm/vp56_arith.c |   89 +++++++++++++++++
>>  libavcodec/arm/vp56_arith.h |   92 ------------------
>>  libavcodec/vp5.c            |    1 +
>>  libavcodec/vp56.c           |    1 +
>>  libavcodec/vp56.h           |  193 -------------------------------------
>>  libavcodec/vp56rac.c        |    1 +
>>  libavcodec/vp56rac.h        |  224 
>> +++++++++++++++++++++++++++++++++++++++++++
>>  libavcodec/vp6.c            |    1 +
>>  libavcodec/vp8.c            |    1 +
>>  libavcodec/x86/vp56_arith.c |   53 ++++++++++
>>  libavcodec/x86/vp56_arith.h |   54 ----------
>>  11 files changed, 371 insertions(+), 339 deletions(-)
>>  create mode 100644 libavcodec/arm/vp56_arith.c
>>  delete mode 100644 libavcodec/arm/vp56_arith.h
>>  create mode 100644 libavcodec/vp56rac.h
>>  create mode 100644 libavcodec/x86/vp56_arith.c
>>  delete mode 100644 libavcodec/x86/vp56_arith.h
>
> The real problem here is that {arm,x86}/vp56_arith.h depend on stuff
> in vp56(rac).h, and there is no way to make that clear in the current
> system. I don't believe that moving stuff to a .c file just for that
> reason is a good idea. Better to fix checkheaders to exclude these
> known-not-to-be-used-directly headers.

Some headers simply don't make sense to include on their own.  Those
should be listed in the SKIPHEADERS make variable.

-- 
Måns Rullgård
[email protected]
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to