[Ffmpeg-devel] [PATCH] Fix compilation when using --disable-opts

Måns Rullgård mans
Thu Mar 15 20:14:41 CET 2007


Michael Niedermayer <michaelni at gmx.at> writes:

> Hi
>
> On Thu, Mar 15, 2007 at 05:57:31PM -0000, M?ns Rullg?rd wrote:
>> 
>> Panagiotis Issaris said:
>> > -----BEGIN PGP SIGNED MESSAGE-----
>> > Hash: SHA1
>> >
>> > Hi,
>> >
>> > M?ns Rullg?rd schreef:
>> >> Diego Biurrun said:
>> >>> On Thu, Mar 15, 2007 at 05:58:35PM +0100, Panagiotis Issaris wrote:
>> >>>> The attached patch fixes compilation when using the --disable-opts
>> >>>> parameter. This to help those interested in using a debugger to debug
>> >>>> FFmpeg.
>> >>> Looks good.
>> >>
>> >> No, the configure changes lack indentation.
>> > Like this?
>> 
>> Yes.  A couple of things though: won't this leave a core dump
>> behind if the ebp test fails?
>
> cant every test leave one behind?

Only if the compiler is badly broken.  This one is expected to crash
sometimes.

>> Someone said there was hardcoded use of ebx and ebp in some inline
>> assembler.  
>
> yes, on x86_64 we either MUST run the test too or unconditionally set 
> CONFIG_EBX/EBP and hope thats ok, i think its ok for ebx but not ebp
> either way wasting time on theories is silly commiting either a 
> uncnditionally enabled EBX/EBP for x86_64 to configure and waiting for
> bugreports or leaving the test there for x86_64 too is MUCH more efficient

I was thinking of the option to use some other register on x86_64, but
if you say ebp and ebx are available, that shouldn't be necessary.

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




More information about the ffmpeg-devel mailing list