[Xcb] More pthread stubs.

M Joonas Pihlaja jpihlaja at cc.helsinki.fi
Tue Jun 22 07:50:47 PDT 2010

On Tue, 22 Jun 2010, Julien Danjou wrote:

> On Tue, Jun 22 2010, M Joonas Pihlaja wrote:

> > The rule in the Makefile.am simply does not propagate 
> > correctly.
> I've pushed a fix. Tell me if it works for you.

It does, thanks!  This still leaves the problem of requiring m4 (or 
actually autoconf) at build time for the user, so I'm still favouring 
the autogen.sh solution.

We discussed on #xcb the use of m4 a little, since it wasn't clear why 
it's not a great choice for a build dependency:  The main issue with 
using m4 at build time is that the BSD m4 is not compatible with the 
(perfectly valid!) m4 code for looping which works with GNU m4. Also, 
the -P option to prefix m4_ to the builtins doesn't work with BSD m4.  
I know some other m4 implementations have their own quirks (aix and 
tru64 come to mind *shudder*), which make writing portable m4 a truly 
formidable challenge.  It says something about a language that the 
looping examples that come with GNU m4 in fact turned out to be buggy 
and underquoted (see autoconf's m4sugar for example.)  Even autoconf 
has kicked other m4s to the curb and requires GNU m4 now.

So, as I see it, the simplest way to get an autoconf that actually 
works for us is to use m4sugar from autoconf.  This is why we'd need 
autoconf to build stubs.c from stubs.c.m4 portably.  The guy running 
autogen.sh is guaranteed to have autoconf.

Another option I think would work well is just dropping stubs.c.m4 
totally and using stubs.c as is.



More information about the Xcb mailing list