Mon May 12 06:28:05 CEST 2008
Alexander Strange wrote:
> On May 11, 2008, at 11:30 PM, Mike Melanson wrote:
>> Carl Eugen Hoyos wrote:
>>> Michael Niedermayer <michaelni <at> gmx.at> writes:
>>>>> Icc warns about mixing of enumerated type with another type.
>>>>> Attached patch fixes the trivial cases.
>>>>> Please comment, Carl Eugen
>>>> looks ok
>>> I will try to fix a few warnings more, they can be seen for example
>>> on fate:
>> Nice work with that. It's now actually possible to view some of the
>> icc's stderr output (used to be too long to store).
> That's too bad, since make test+icc is the only regularly failing
> test. If I run it here it segfaults during ffv1 - is it the same there?
We're talking about 2 different things here. I was referring to the fact
that icc used to output so many warnings when building that, even
compressed, the entire stderr text could not fit in 64KB and was being
omitted from the database. You're referring to the fact that the "full
regression suite" test spec in FATE never stores any stdout/stderr.
That's on purpose, or else the database would get very full.
I would like to store the last 25 lines or so of the combined
stdout/stderr. But I can't find a good one-liner for doing so that
doesn't destroy the 'make test' return code.
For the record, whenever I build FFmpeg with icc and then run 'make
test', it crashes when running audiogen. If I run 'make test' again, it
gets past audiogen but crashes at FFV1, IIRC.
More information about the ffmpeg-devel