[MPlayer-cvslog] r24158 - trunk/input/input.c

Reimar Döffinger Reimar.Doeffinger at stud.uni-karlsruhe.de
Sat Aug 25 18:50:51 CEST 2007


Hello,
On Sat, Aug 25, 2007 at 12:04:11PM -0400, Rich Felker wrote:
> On Sat, Aug 25, 2007 at 09:35:35AM +0200, reimar wrote:
> > Author: reimar
> > Date: Sat Aug 25 09:35:35 2007
> > New Revision: 24158
> > 
> > Log:
> > Hack around broken bitfields in gcc 4.2
> > 
> > 
> > Modified:
> >    trunk/input/input.c
> > 
> > Modified: trunk/input/input.c
> > ==============================================================================
> > --- trunk/input/input.c	(original)
> > +++ trunk/input/input.c	Sat Aug 25 09:35:35 2007
> > @@ -609,7 +609,8 @@ mp_input_add_cmd_fd(int fd, int select, 
> >    cmd_fds[num_cmd_fd].fd = fd;
> >    cmd_fds[num_cmd_fd].read_func = read_func ? read_func : mp_input_default_cmd_func;
> >    cmd_fds[num_cmd_fd].close_func = close_func;
> > -  cmd_fds[num_cmd_fd].no_select = !select;
> > +  if (!select)
> > +    cmd_fds[num_cmd_fd].no_select = 1;
> 
> Any explanation of what the "broken bitfields" bug is? This seems
> exceedingly bad of gcc...

It is only in 4.2.0 it seems:
http://bugs.gentoo.org/show_bug.cgi?id=190154
It only happens with signed bitfields, maybe the optimizer just doesn't
like numbers consisting only of the sign...

Greetings,
Reimar Döffinger



More information about the MPlayer-cvslog mailing list