[MPlayer-cvslog] r18990 - trunk/libmpdemux/demux_lavf.c

[Rich Felker]

On Mon, Jul 10, 2006 at 04:52:40PM +0200, Diego Biurrun wrote:
> On Mon, Jul 10, 2006 at 10:27:24AM -0400, Rich Felker wrote:
> > On Mon, Jul 10, 2006 at 07:16:00AM +0300, Uoti Urpala wrote:
> > > On Mon, 2006-07-10 at 04:53 +0200, rfelker wrote:
> > > > Log:
> > > > more c++ decl crap
> > > 
> > > It's not C++, as you should well know. It's C, as opposed to legacy C
> > 
> > It's called C++ declarations for the same reason // comments are
> > called C++ comments.
> > 
> > > with workarounds for broken compilers. Seems nobody else had tried
> > > compiling with gcc 2.95 for 3 weeks. Time to drop support for it I
> > > think.
> > 
> > No, absolutely not. If you continue with such trolling I will
> > fight to have you removed from development. gcc3 generates SLOWER CODE
> > than gcc 2.95 and gcc 4 DOES NOT COMPILE on machines with less than
> > several gigs of virtual memory. Thus 2.95 is the only option.
> 
> gcc 3.4 does not generate slower code than 2.95, it just takes far
> longer but the end result is slightly faster.  Look up the numbers I
> generated ages ago.

Perhaps I will sometime. Still >double compile time is a ridiculous
price to pay for "slightly" faster, and I doubt it's faster on all
cpus even if it is on our K6-IIIs.

> Now stop this useless flamage.

I propose the following. Either Uoti stop committing broken code, or
send me a patch to gcc 2.95's parser to make it accept C++ style
declarations.

Even if he can patch gcc though, I find this kind of coding unreadable
and horribly broken except in very specific circumstances where it's
necessary for controlling stack usage without introducing many many
levels of braces, mostly in combination with vla's, and I recommend we
formally ban C++ style declarations in the policy and add the warning
flag for it when gcc>2 is detected so that the C++ coders with this
bad habit get a warning...

Rich