Hi Peter,
* Peter Rosin wrote on Mon, Feb 14, 2011 at 09:14:03AM CET:
> Den 2011-02-12 11:10 skrev Ralf Wildenhues:
> > * Peter Rosin wrote on Sat, Jan 29, 2011 at 02:26:24PM CET:
> >> Or is plain 'ar' used somewhere instead of 'x86_64-w64-mingw32-ar'?
> >
> > Automake outputs 'AR = ar' in Makefile.in for rules creating old
> > libraries iff neither AC_PROG_LIBTOOL nor another method to define
> > AR correctly is used in configure.ac.
> > A good workaround, as already mentioned, is to use this in configure.ac:
> > AC_CHECK_TOOL([AR], [ar], [false])
>
> I just cannot understand why the workaround isn't always working in
> this case.
>
> There was a log posted with this in it
> (in http://archives.postgresql.org/pgsql-hackers/2011-01/msg02697.php):
[...]
> configure:6164: checking for x86_64-w64-mingw32-ar
> configure:6180: found /mingw/bin/x86_64-w64-mingw32-ar
> configure:6191: result: x86_64-w64-mingw32-ar
[...]
> Which seem to match this snippet from configure.in:
>
> ...
> AC_PROG_RANLIB
> PGAC_CHECK_STRIP
> AC_CHECK_TOOL(AR, ar, ar)
> if test "$PORTNAME" = "win32"; then
> AC_CHECK_TOOL(DLLTOOL, dlltool, dlltool)
> AC_CHECK_TOOL(DLLWRAP, dllwrap, dllwrap)
> AC_CHECK_TOOL(WINDRES, windres, windres)
> fi
> ...
>
> Sure, AC_CHECK_TOOL has under-quoted arguments and the last argument is
> 'ar' instead of 'false'. But that shouldn't really matter here. (Or
> does it?)
No, that's irrelevant.
> Still, elsewhere in the thread there's a report about the wrong ar being
> used.
> (in http://archives.postgresql.org/pgsql-hackers/2011-01/msg02713.php)
Well, the poster wrote that it worked now though:
http://archives.postgresql.org/pgsql-hackers/2011-01/msg02806.php
> Sure, the configure log and the "wrong ar"-report are not from the same
> person, but the configure script should be the same for everybody (git
> log hints that this part of configure has been stable for a couple of
> years).
>
> It just doesn't add up.
FWIW, I don't see enough evidence of breakage to be able to analyze it.
Thanks,
Ralf