The following reply was made to PR pkg/36603; it has been noted by GNATS.

From: Robert Elz <kre@munnari.OZ.AU>
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: pkg/36603 (pkgsrc "make checksum" even more broken than I suspected) 
Date: Fri, 31 Aug 2007 21:49:28 +0700

     Date:        Fri, 31 Aug 2007 14:20:02 +0000 (UTC)
     From:        "Johnny C. Lam" <jlam@pkgsrc.org>
     Message-ID:  <20070831142002.B0C7E63B8B5@narn.NetBSD.org>
 
   |  Okay, this was the main part of the PR that I was concerned about
   |  fixing, and I'm glad the changes did fix it.
 
 I am fairly sure it should be fixed, but it isn't easy to produce the
 exact test case where I detected the problem...
 
   |  At the time I committed the fix for this PR, I was thinking about
   |  splitting the checksum target into a part that is run as part of the
   |  normal build process, and another that could be run "on-demand" by
   |  the user.  I think we're  thinking the same thing here.
 
 Yes, it sounds so.   Great.
 
   |  I will leave this PR as analyzed for now, and I hope to commit a change
   |  with your proposed behavior later today, at which point I'll request
   |  testing and feedback again.
 
 No problem, if you allow me enough time - just getting the updates (I have
 my own copy of the repository that I sync from one of the mirrors - so the
 mirror needs to update first, and then I need to get that, it all takes
 a while...)
 
 What I haven't tested yet is whether we're now back in the state where
 "make checksum" calculated the checksums twice if the distfile needed to
 be fetched (once while fetch verifies it worked and doesn't have to try
 again at a backup MASTER_SITE, and then once again for the checksum target).
 That never bothered me, but (I am assuming) was something of the inspiration
 for the earlier change.
 
 Thanks again.
 
 kre