[FFmpeg-cvslog] r12930 - trunk/libavcodec/mimic.c

[Mike Melanson]

Reimar D?ffinger wrote:
> On Tue, Apr 22, 2008 at 03:20:04PM -0700, Mike Melanson wrote:
>> Robert Swain wrote:
>>> It might be useful if viewers of the fate.multimedia.cx page could see  
>>> which test(s) failed, possibly why and what the test is. I guess  
>>> you've already thought of this but just haven't got round to doing it  
>>> yet. :)
>> Oh yes. :) And it's coming sooner rather than later. Currently, I have
>> to do a manual query to figure out when a test broke for a configuration.
> 
> Ah, that reminds me, as a compromise between space used and usefulness,
> maybe FATE could keep the full compile and test logs of just the last
> run? Or at least a "tail -n 50" or so of it?

FATE already keeps the full stdout and stderr generated by configure &&
make for each build configuration. I am planning to retire older data.
But all the data is currently there unless it's too long (as is always
the case with Intel's pedantic C compiler).

Perhaps you are referring to storing the last 50 lines of 'make test'? I
agree that doing so would be very useful. The most straightforward
method I can think of is:

  make test 2&>1 > maketest-output.txt && tail -n 50 maketest-output.txt

Problem: The 'tail' portion of the command clobbers any return code from
the 'make test' portion, thus altering the FATE results. I could modify
the FATE test to capture the last n lines of stdout instead of the whole
thing.

Or I could move faster with adapting all of the constituent regression
tests as individual FATE tests. :)

-- 
	-Mike Melanson