[FFmpeg-devel] [PATCH/RFC] Constify src in av_fifo_generic_write()

[Tomas Härdin]

On Tue, 2010-02-23 at 15:55 +0100, Michael Niedermayer wrote:
> On Tue, Feb 23, 2010 at 01:41:49PM +0100, Tomas H?rdin wrote:
> > On Tue, 2010-02-23 at 13:14 +0100, Michael Niedermayer wrote:
> > > On Tue, Feb 23, 2010 at 10:50:03AM +0100, Tomas H?rdin wrote:
> > > > Good morning
> > > > 
> > > > This is a rather simple patch that merely makes av_fifo_generic_write()
> > > > take const void *src instead of just void *src. It causes no warnings
> > > > and regression tests pass.
> > > > 
> > > > The reason is simple: guarantee the user that the function does not
> > > > modify src.
> > > > 
> > > > One small issue though: this necessitated changing func to int(*)(const
> > > > void*,void*,int), which is OK for all code within FFmpeg. However, this
> > > > seems to count as a minor API change since g++ (and gcc I presume) won't
> > > > allow implicitly casting int(*)(void*,void*,int) to int(*)(const
> > > > void*,void*,int). I presume that part requires a bit of discussion,
> > > > which is why I marked this thread "RFC".
> > > > 
> > > > Patch attached.
> > > 
> > > ive added a comment explaining why src cannot be const as you are at
> > > least the 2nd person trying to introduce this bug
> > 
> > Ah, ok. Might I suggest that you put a comment in fifo.h? That way users
> > can see it too (especially if only the headers are installed), and there
> > is less chance anyone else gets confused about this :o
> 
> EPATCHWELCOME

Changed to:

@param *src data source; non-const since it may be used as a modifiable
context in func

Should be sufficient. Patch attached.

/Tomas

-------------- next part --------------
A non-text attachment was scrubbed...
Name: fifo_h_const_doc.patch
Type: text/x-patch
Size: 579 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100224/9cf449dd/attachment.bin>