[MPlayer-cvslog] r31783 - in trunk: Makefile configure

[Diego Biurrun]

On Sun, Jul 25, 2010 at 11:53:22AM +0200, Reimar Döffinger wrote:
> On Sun, Jul 25, 2010 at 11:39:47AM +0200, Diego Biurrun wrote:
> > On Sat, Jul 24, 2010 at 05:14:44PM +0200, reimar wrote:
> > > 
> > > Log:
> > > Fix dependency generation for libav*
> > > 
> > > --- trunk/configure	Sat Jul 24 16:43:55 2010	(r31782)
> > > +++ trunk/configure	Sat Jul 24 17:14:43 2010	(r31783)
> > > @@ -2562,7 +2562,7 @@ else
> > >  
> > >  cflag_check -mno-omit-leaf-frame-pointer && cflags_no_omit_leaf_frame_pointer="-mno-omit-leaf-frame-pointer"
> > > -cflag_check -MD -MP && DEPFLAGS="-MD -MP $CFLAGS"
> > > +cflag_check -MMD -MF "$mplayer_tmpdir/test.d" -MT test.o && DEPFLAGS='-MMD -MF $(@:.o=.d) -MT $@'
> > 
> > This is not equivalent.  -MP works around compile failures when header
> > files get renamed.
> 
> First, that was by accident, but...
> Are you sure this does not have some unwanted side-effects?
> FFmpeg's common.mak has a dummy rule for .h files, but that isn't
> quite right either since we have .c template files.

Mans distrusts -MP for some reason.  I don't see such a reason and
I don't like the troubles caused by renamed header files.

> Also this workaround seems to substitute compile failures for
> silent failures to rebuilt correctly, honestly failing to
> compile IMO is the _correct_ behaviour, you simply have to do
> a make clean after a header rename.

I haven't seen such a failure yet, but the issues with renamed
headers bit me again over the weekend.  I'm changing this back.

Diego