[FFmpeg-devel] FATE & Regressions (and PPC is broken)

[Mike Melanson]

Michael Niedermayer wrote:
> Someone posted something about FFV1 and ICC on the list ...
> IIRC there was an issue with variable length arrays and the way they
> are memset()
> Actually iam not completely sure if that is C standard compliant, if someone
> knows dont hesitate to speak up ...

Are you talking about sequences like this?

int_fast16_t sample_buffer[3][ring_size][w+6], *sample[3][ring_size];
[...]
memset(sample_buffer, 0, sizeof(sample_buffer));


BTW, gcc 4.3.0 on x86_64 fails the same way as 4.4-svn on the Snow 
regression tests. Not sure how much that matters in the grand scheme 
given Snow's current status, but it is certainly cause for suspicion 
that earlier gcc versions are okay with it.

4.3.0 on PPC fails exactly the same as the other PPC configs, as expected.

I wonder if I could automatically log the error output of a 'make test' 
run for db logging by combining grep and tail somehow?

-- 
	-Mike Melanson