[Ffmpeg-devel] [PATCH] Use ff_eval for AVOption parsing

[Panagiotis Issaris]

Hi,

On za, 2006-09-23 at 13:02 +0200, Michael Niedermayer wrote:
> [...]
> > exact reason for the segfault, but I am sending it anyway as I'm off for
> > lunch and shopping and hoped for some feedback on this patch.
> 
> iam against changes which by some mysterious reason fix a segfault ...
I understand, and in fact fully agree. Just wanted to show the work in
progress, as to see if there were other objections against it besides
the "if nan then return" hunks. No need to debug _another_ 3 hours if
the patch would be rejected for other reasons anyway :)

> maybe gcc misscompiles something? does -O0 or gcc 2.95 fix that segfault?
Tried GCC 3.4 (besides my default 4.0.3).

> what does valgrind say about it? and if all fails adding lots of printf/
> av_log to see exactly whats happening before the segfault seems to be
> the obvious next step
Yep, had been doing that this morning but couldn't find it
yet :-( Didn't put valgrnid on it yet though. I will have another go
this evening or tomorrow. I still have all the debugging code (logging)
in a separate GIT-branch.

The segfault appeared when using "min", "max", "default" not when using
"normal" expression or simple values. So "-b 100000", "-b 100k", "-b
100*10" "-b 100*QP2LAMBDA" all worked, "-b default" caused a segfault.

> also look at the warnings gcc outputs when compiling eval.c & opt.c
No warnings, unfortunately.

With friendly regards,
Takis