[Ffmpeg-devel] Re: On2 vs libvp62

Rich Felker dalias
Sat Apr 22 00:37:00 CEST 2006

On Fri, Apr 21, 2006 at 10:19:28AM -0700, Mike Melanson wrote:
> M?ns Rullg?rd wrote:
> >Glibc also implements many non-standard functions found in BSD, SysV
> >and other Unixes, making porting code from those systems to glibc
> >easier. This is not fundamentally a bad thing.

My position is that it _is_ fundamentally a bad thing. It encourages
new programs to keep using these extensions and forces everyone to pay
the price of having them always loaded in memory, slowing down symbol
resolution at dynamic-link-time, etc. Putting legacy stuff in a static
library for legacy code to link to would have worked just as well and
avoided all the problems.

> >I agree that
> >inventing extensions of their own is bad.
> 	What are you talking about? I can't count the number of times that 
> strcpy() and memfrob() have saved me in a pinch! Of course, I can't 
> count very high anyway so that might not be such a rave review.
> $ man strfry
> $ man memfrob


Try some of the truly gnu-specific ones: obstack, dprintf, the
gnu-specific regex api, reentrant versions of legacy functions (such
as [efg]cvt_r). Then on top of that there are extensions to existing
functions such as the %a modifier in sscanf (which actually makes gnu
scanf _broken_ with regard to C99), extra flags to fnmatch/glob/etc.,

Eventually I should write a whole paper on the subject when I'm bored
but for now this is a good starting list of broken extensions.


More information about the ffmpeg-devel mailing list