[MPlayer-cvslog] r27786 - in trunk: configure libmpdemux/demuxer.h
diego at biurrun.de
Thu Oct 16 22:56:55 CEST 2008
On Thu, Oct 16, 2008 at 09:44:00PM +0300, Uoti Urpala wrote:
> On Thu, 2008-10-16 at 20:26 +0200, diego wrote:
> > Move likely/unlikely macros to libmpdemux/demuxer.h where they are used.
> > config.h should only contain option definitions, no code.
> Even if they're only used in files including demuxer.h they're not
> related to demuxing, so IMO demuxer.h is no better place for them.
Do you have a suggestion for a better place?
> Also I wouldn't call them "code".
"code" is not a good word, yes. Maybe "logic" would be better.
> Your other commit moved the DEFAULT_CDROM_DEVICE logic from configure to
> stream/stream.h. I feel uncertain whether moving such platform-specific
> #ifdefs to individual header files is a good idea. More generally, what
> exactly is your criteria for "option definitions"? Do you mean option
> definitions that are expressed as #ifdefs in config.h instead of
> equivalent tests in the configure script? Or the value of the resulting
> definitions? Something like the likely/unlikely macros could be
> constructed by code in the configure script and so would have to appear
> in some output file of the script.
I would like to keep logic out of config.h and keep it either in other
header files or in configure. Yes, configure could (and maybe should)
generate some of this stuff directly. Does that make sense?
More information about the MPlayer-cvslog