Subject: Re: MASTER_SITE_SUSE91 ??
To: Geert Hendrickx <ghen@NetBSD.org>
From: Robert Elz <kre@munnari.OZ.AU>
List: tech-pkg
Date: 01/10/2006 13:58:23
    Date:        Mon, 9 Jan 2006 21:45:30 +0100
    From:        Geert Hendrickx <ghen@NetBSD.org>
    Message-ID:  <20060109204530.GA15970@lori.ghen.be>

  | Jaromir Dolecek wrote:
  | > Check emulators/suse91_linux/Makefile.common

Ah, didn't find that one.   Thanks.

  | Geert Hendrickx:
  | Apart from that, you can always check the value of a variable with "make
  | show-var VARNAME=MASTER_SITE_SUSE91", even if you don't know where it's
  | defined.  

And I get nothing (as in am empty line).    I didn't ask from idle
curiosity, I asked because it clearly wasn't set for me (for the one
package that gave me an error - I admit to not thinking too hard as
to why it was working for all the others...)

I suspect a bug in emulators/suse91_aspell which is the package that I
noticed I was unable to fetch the distfile (the only place it tried was
ftp.netbsd.org which doesn't have it - or at least doesn't have it
using the pathname that pkgsrc was attempting to use to find it).

I suspect the problem relates to the ordering of suse91_aspell/Makefile.

That includes suse91_linux/Makefile.application, which in turn includes
suse91_linux/Makefile.common - but the latter only if RPMFILES is set.

In suse91_aspell RPMFILES comes from Makefile.i386 (or Makefile.x86_64)
which is included after suse91_linux/Makefile.application - so RPMFILES
doesn't end up defined at the point it is needed to get
suse91_linux/Makefile.common included - so MASTER_SITE_SUSE91 remains
undefined.

suse91_aspell is the only user of suse91_linux/Makefile.application that
I could find (very quick grep), so who knows, but some other suse91
Makefiles include the Makefile.${MACHINE_ARCH) relatively
early, rather  than relatively late - and sure enough, rearranging the
suse91_aspell Makefile corrects this problem for me.

I can file a PR if needed to get this fixed, but I suspect that rather
than tell me to do that, and then have to close the PR again when it
gets fixed, someone could just fix the problem ...   (It is unlikely
that the actual change I made to correct this is really the proper
way to fix it).

kre