[FFmpeg-devel] [PATCH 1/2] avutil/common: Add FF_PTR_ADD()

[Anton Khirnov]

Quoting Michael Niedermayer (2021-03-31 00:44:16)
> On Mon, Mar 29, 2021 at 12:19:57PM +0200, Anton Khirnov wrote:
> > Quoting Michael Niedermayer (2021-03-15 09:47:43)
> > > Suggested-by: Andreas Rheinhardt
> > > Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> > > ---
> > >  libavutil/common.h | 2 ++
> > >  1 file changed, 2 insertions(+)
> > > 
> > > diff --git a/libavutil/common.h b/libavutil/common.h
> > > index aee353d399..c2d47a45b4 100644
> > > --- a/libavutil/common.h
> > > +++ b/libavutil/common.h
> > > @@ -108,6 +108,8 @@
> > >  #define FFSWAP(type,a,b) do{type SWAP_tmp= b; b= a; a= SWAP_tmp;}while(0)
> > >  #define FF_ARRAY_ELEMS(a) (sizeof(a) / sizeof((a)[0]))
> > >  
> > > +#define FF_PTR_ADD(ptr, off) ((off) ? (ptr) + (off) : (ptr))
> > 
> > IIUC Andreas suggested to either put an AV prefix on this or move it to
> > an internal header. I agree with that and prefer the latter, this macro
> > doesn't seem very useful for external callers.
> 
> ill move it to internal.h
> 
> I think though our headers are not optimally named and the terminology used is
> a bit ambigous. for example what exactly is external ? outside the lib, all libs maintained
> by us, outside the soname object, ...
> 
> whats the header that is internal to libavutil ?
> libavutil/internal.h, nope
> 
> whats the header with API common to all our libs in libavutil ?
> libavutil/common.h, nope though it contains a good part of that API
> 
> I know these still it feels wrong every time i see a
> #include "libavutil/internal.h" outside libavutil
> 
> Probably its just me but even after years this naming system still feels odd to me

I certainly agree that it's messy and efforts to improve it are very
much welcome.

In this case at least, by external I mean the stable API we provide to
other projects. And I think there's broad consensus that
- all global external identifiers provided by us in installed headers
  should be AV/av-prefixed
- installed headers should, as much as possible, not declare anything
  that is not public API

Neither of those are completely true today, but it's a goal we should be
trying to approach.

-- 
Anton Khirnov