[FFmpeg-devel] [RFC] How to fix DR+lavfi+vflip crash

[Stefano Sabatini]

On date Monday 2010-12-27 16:13:14 +0100, Michael Niedermayer encoded:
> On Mon, Dec 27, 2010 at 03:23:02PM +0100, Stefano Sabatini wrote:
> > On date Monday 2010-12-27 01:54:54 +0100, Michael Niedermayer encoded:
> > > On Sun, Dec 26, 2010 at 01:51:21PM +0100, Michael Niedermayer wrote:
> > > > On Sun, Dec 26, 2010 at 11:41:57AM +0100, Stefano Sabatini wrote:
> > > > > On date Saturday 2010-12-25 02:22:48 +0100, Stefano Sabatini encoded:
> > > > > > On date Thursday 2010-12-23 11:14:26 +0100, Stefano Sabatini encoded:
> > > > > [...]
> > > > > > > > > From ee6dc086923ed994a5dd70ab79513d9a510c29ce Mon Sep 17 00:00:00 2001
> > > > > > > > > From: Stefano Sabatini <stefano.sabatini-lala at poste.it>
> > > > > > > > > Date: Sun, 19 Dec 2010 16:35:50 +0100
> > > > > > > > > Subject: [PATCH] Introduce the AV_PERM_NEG_LINESIZES flag, which allows to explicitely
> > > > > > > > >  request a buffer which can have negative linesizes, so that if it is
> > > > > > > > >  not specified in the permissions for the requested buffer the returned
> > > > > > > > >  buffer must have positive linesizes.
> > > > > > > > > 
> > > > > > > > > This is required in particular with direct rendering as implemented in
> > > > > > > > > ffplay, as some codec cannot process an image buffer with negative
> > > > > > > > > linesizes.
> > > > > > > > > 
> > > > > > > > > Fix issue #1913.
> > > > > > > > 
> > > > > > > > Ping.
> > > > > > > 
> > > > > > > I'll apply in few days if I see no comments.
> > > > > > 
> > > > > > I'll apply in two days, Michael?
> > > > > 
> > > > > I'm going to apply tomorrow.
> > > > 
> > > > wait please
> > > 
> > > Heres a counterproposal to your patch, please tell me with what filter chain
> > > this fails for you
> > > Also keep in mind that if you have filters that cannot handle negative linesizes
> > > they must declare this their rej_perm
> > [...]
> > 
> > Thanks, the only problem I can see with your patch is that it doesn't
> > support slice filtering (i.e. the whole input frame is copied to the
> > output buffer in start_frame()),
> 
> There is no frame copying code in start_frame thus i doubt this
> 
> 
> 
> > on the other hand this approach is
> > much simpler and cleaner so I tend to prefer to keep this patch and
> > drop that feature (which is not critical anyway).
> > 
> > So if you're OK I'm going to apply your patch.
> 
> applied (without typo)

Unfortunately can't work with pad, I was too quick to push the button.
-- 
FFmpeg = Fiendish and Free Mind-dumbing Portentous Efficient Guide