Re: CVS commit: pkgsrc/devel

["David H. Gutteridge" <david%gutteridge.ca@localhost>]

> Module Name:    pkgsrc
> Committed By:   nia
> Date:           Thu Jan 23 11:42:16 UTC 2020
> 
> Modified Files:
>         pkgsrc/devel/gettext: Makefile PLIST
>         pkgsrc/devel/gettext-tools: Makefile
> 
> Log Message:
> gettext: check for glibc in the right package + check for glibc, not Linux
> 
> 
> To generate a diff of this commit:
> cvs rdiff -u -r1.86 -r1.87 pkgsrc/devel/gettext/Makefile
> cvs rdiff -u -r1.20 -r1.21 pkgsrc/devel/gettext/PLIST
> cvs rdiff -u -r1.39 -r1.40 pkgsrc/devel/gettext-tools/Makefile
[...]
> > Index: pkgsrc/devel/gettext-tools/Makefile
> > diff -u pkgsrc/devel/gettext-tools/Makefile:1.39 pkgsrc/devel/gettext-tools/Makefile:1.40
> > --- pkgsrc/devel/gettext-tools/Makefile:1.39    Sun Nov  3 10:39:12 2019
> > +++ pkgsrc/devel/gettext-tools/Makefile Thu Jan 23 11:42:16 2020
> > @@ -1,4 +1,4 @@
> > -# $NetBSD: Makefile,v 1.39 2019/11/03 10:39:12 rillig Exp $
> > +# $NetBSD: Makefile,v 1.40 2020/01/23 11:42:16 nia Exp $
> >  
> >  .include "../../devel/gettext/Makefile.common"
> >  
> > @@ -41,7 +41,7 @@ PLIST_VARS+=  glibc
> >  
> >  .include "../../mk/bsd.prefs.mk"
> >  
> > -.if ${OPSYS} == "Linux"
> > +.if !empty(GLIBC_VERSION)
> >  PLIST.glibc=   yes
> >  .endif

Hi Nia,

As-is, this will break the build on Linux distros that place libc.so
somewhere other than the one place that's presently checked to populate
GLIBC_VERSION in platform/Linux.mk. E.g., Debian, Ubuntu, et al. have
multi-arch support that places the x86_64 version of the library at
/lib/x86_64-linux-gnu/, not at /lib64.

With, say, Debian 9, we now get:

=> Checking file-check results for gettext-tools-0.20.1
ERROR: ************************************************************
ERROR: The following files are in /home/disciple/pkgsrc/devel/gettext-tools/work/.destdir/home/disciple/pkg but not in the PLIST:
ERROR:         /home/disciple/pkgsrc/devel/gettext-tools/work/.destdir/home/disciple/pkg/lib/preloadable_libintl.so
*** Error code 1

Linux.mk needs adjustments if it's really going to be used to provide
that variable. (There's more than one way to do it, but that perhaps
up a can of worms...)

Regards,

Dave