svn commit: r190943 - head/include

Coleman Kane cokane at cokane.org
Sat Apr 11 19:30:38 UTC 2009


On Sat, 2009-04-11 at 21:44 +0400, Stanislav Sedov wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On Sun, 12 Apr 2009 02:18:41 +0900
> Norikatsu Shigemura <nork at FreeBSD.org> mentioned:
> 
> > On Sat, 11 Apr 2009 21:07:02 +0400
> > Stanislav Sedov <stas at freebsd.org> wrote:
> > > > Log:
> > > >   GNU Pth has some fragile kludges that were broken by r189828.
> > > >   I've discussed this with the Pth maintainer and no clear solution
> > > >   has emerged on the ports side of things, so for now, hack around
> > > >   the issue in signal.h.
> > > Can't we just put a patch in ports tree itself? What meant under 'no
> > > clean solution emerged'? I can prepare a patch, if needed.
> > 
> > 	I think so, too.  I have a quick hack patch.
> > 
> > ports/devel/pth/files/patch-pth_p.h.in
> > - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
> > --- pth_p.h.in.orig	2006-06-09 02:54:03.000000000 +0900
> > +++ pth_p.h.in	2009-04-08 15:05:12.911807009 +0900
> > @@ -32,7 +32,7 @@
> >  #include <stdarg.h>
> >  #include <string.h>
> >  #include <setjmp.h>
> > -#include <signal.h>
> > +//#include <signal.h>
> >  #include <unistd.h>
> >  #include <fcntl.h>
> >  #include <errno.h>
> > 
> > - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
> > 
> 
> Or just drop the line. Vasil, what do you think about the possibility to
> commit this patch into the tree? I think we should have a workaround in ports
> tree (it is one of the things what it was designed for) than in the src tree.
> Better would be to implement a workaround upstream, though. I think something
> like
>  #if !(defined(__FreeBSD__) && __FreeBSD_version >= %%PTHREAD_CHANGE_VERSION%%)
>  #include <signal.h>
>  #endif
> where %%PTHREAD_CHANGE_VERSION%% would correspond to the __FreeBSD_version where
> the change was introduced, could be safely submitted upstream.
> 
> - -- 
> Stanislav Sedov
> ST4096-RIPE

I agree... if you've been discussing it with the Pth maintainer, then
what are the chances that you can get Pth to change their behavior?

Since Pth is supposed to be a support library that works on a large
collection of platforms to provide a consistent multithreading and I/O
API to users, I would imagine that their code should not #include
<signal.h> when: a) They don't need it on our platform, and b) It breaks
things on our platform.

-- 
Coleman Kane



More information about the svn-src-head mailing list