[FFmpeg-devel] [RFC] MinGW+FFmpeg survival guide

[Axel Holzinger]

On Saturday, June 25, 2011 7:36 PM Reimar Döffinger wrote:
> On Sat, Jun 25, 2011 at 07:29:12PM +0200, Axel Holzinger wrote:
> > On Saturday, June 25, 2011 6:40 PM Reimar Döffinger wrote:
> > > On Sat, Jun 25, 2011 at 06:19:18PM +0200, Axel Holzinger
> > > wrote:
> > > > On Saturday, June 25, 2011 5:32 PM Reimar Döffinger wrote:
> > > > > On Fri, Jun 24, 2011 at 09:03:52PM +0200, Axel Holzinger
> > > > > wrote:
> > > > > > Do you know off-hand if using a different token '/' is
> > > > > > already supported? The last time I tried (nearly a year
> > > > > > ago) I think it wasn't.
> > > > >
> > > > > The answer isn't quite as simple unfortunately.
> > > > > There are unfortunately some flags that are used directly
> > > > > in the Makefile, like -c, -o etc. they will be an issue.
> > > > > For everything else the filter function can replace
> > > > > absolutely anything with anything else.
> > > >
> > > > I will have a look, thank you.
> > >
> > > Note that due to above, it might be best to have a wrapper
> > > script in addition that can replace some of the simpler
> > > command-line options.
> > > Possibly such a wrapper script around icl that emulates gcc
> > > might even work best at all, without configure changes.
> >
> > :-)
> >
> > That's what I did a year ago. Well I wrote a little C++
> > translation program that did convert the gcc/icc switches to
> > icl. But that's still a pita.
> 
> Why?

I was simply not good enough in scripting and understanding the
configure script. I tried, but then it was more feasable to write a
C++ program that reads in a translation file. So in the end make
called the translation tool and the translation tool called icl.

> It _could_ be made a script included in FFmpeg and that
> automatically is used when compiling with ICL.
> You could also support most via the normal configure filter
> function (so that functionality-testing is available), it's
> just that replacing standard options like -o and -c seems
> rather messy.

I would very much appreciate this, but I think I'm still not the right
one to do it, cause I'm not good enough in this.

> > Thinking about that I remember that still stdint and stdbool
> > headers are missing for icl, because it just relies on the ms
> > headers and there they're missing.
> 
> Hm, but then the compiler is not C99-compatible.

Well, the compiler program is, but they don't care of supplying the
right headers, so in the end, well it's not really.

> Anyway, I do not know about others, but I would not mind
> including such stub headers for ICL either, assuming they are
> under a compatible license.

At that time last year I just downloaded them from a link of somebody
posted to this list. The one stdint.h is free "THIS SOFTWARE IS NOT
COPYRIGHTED", the other stdbool.h is derived from gcc, but extended
with a c++ and msc check. That one should be rewritten, I guess.

Could also be a possibility to make a package for mingw-get that
brings these headers. But I'm not familiar with how to write mingw-get
packages neither :-(