[FFmpeg-devel] MXF D10 regression tests

[Reimar Döffinger]

On Thu, Mar 19, 2009 at 07:49:12PM +0100, Michael Niedermayer wrote:
> On Thu, Mar 19, 2009 at 06:55:53PM +0100, Reimar D?ffinger wrote:
> > On Thu, Mar 19, 2009 at 03:09:13PM +0100, Michael Niedermayer wrote:
> > > On Thu, Mar 19, 2009 at 08:59:25AM +0100, Reimar D?ffinger wrote:
> > > > Your missing that PAL8 currently has linesize[1] = 4
> > > 
> > > so change linesize to 0 or add a check for PAL8
> > 
> > Changing linesize could be considered breaking API, and the alternative
> 
> what API?
> I didnt even remember that it was set to 4 nor do i know why its done or
> what meaning it is supposed to have if its not 4 or 0
> and iam pretty sure this is not documented anywhere

No disagreement, just wanted to make sure.

> > would need to check for more than just PAL8, it's a (small) load of
> > other formats that have linesize[1] == 4 currently.
> 
> linesize[2]==0 linesize[1]==4 but i prefer to change it to 0 if possible

I haven't test it extensively (i.e. beyond make test), but I think making it 0 works.

> > > its neither me who broke the regressions
> > 
> > Since it just uncovered an existing bug there is not really anyone else
> > to blame either though.
> 
> well last time i uncovered a unrelated bug (in the timestamping code, and
> in gif and ...) i was flamed, now someone else uncovers one and again i
> am the one pointed at

Well, blame is the thing you put where ever it is most convenient. ;-)
For the time stamp stuff I have to say for an outstander it often is not easy
to say what is a bug or feature.
Anyway let me say it like that: I do think that a maintainer always has
a extra bit of responsibility for the correctness of "his" code.
And I reserve for me the right to have human faults, in this case
particularly not taking your time constraints into account after two
rejected patches. Sorry if we end up a bit annoyed at each other
occasionnally.

> > > nor me who submitted the patch
> > 
> > I protest the (hopefully not intended by you) implication that whoever
> > tries to fix something suddenly becomes responsible for fixing it.
> 
> whoever wants something fixed or caused the bug, has to do the work.

I expect both apply to both of us in similar amounts :-)
Anyway. have a look at attached patch. I tested it on x86 with modified
STRIDE_ALIGN and it does fix the issue.
Maybe I have taken "avoiding cosmetics" too far but that should be
easily fixable. My dislike of goto shows a bit too compared to your
original suggestion...
-------------- next part --------------
A non-text attachment was scrubbed...
Name: align.diff
Type: text/x-diff
Size: 1762 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20090319/35c06046/attachment.diff>