Re: pkg/39734 (lang/perl5 upgrade screws (some?) p5-* binary packages)

[Robert Elz <kre%munnari.OZ.AU@localhost>]

The following reply was made to PR pkg/39734; it has been noted by GNATS.

From: Robert Elz <kre%munnari.OZ.AU@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: pkg/39734 (lang/perl5 upgrade screws (some?) p5-* binary packages) 
Date: Mon, 22 Dec 2008 20:47:22 +0700

     Date:        Sun, 21 Dec 2008 16:44:04 +0000 (UTC)
     From:        he%NetBSD.org@localhost
     Message-ID:  <20081221164404.D806663B121%narn.NetBSD.org@localhost>
 
   | While the problem may not have been fully solved,
   | I'm not sure there is much more we can do at the
   | moment to improve the situation, because the fix
   | should have been made a long time ago, and there
   | are too many p5-* packages built for 5.8 "in the
   | wild" at the moment.
 
 Actually, no, that presumes that the fix is to somehow have
 packages guess what later version of some dependency they will
 be incompatible with, and as a general solution, that's patently
 absurd.  (As a quick fix, right now, it is better than nothing.)
 
 The right solution is for packages to note when they are ABI
 incompatible with an earlier version (so perl 5.10 could have
 said it was incompat with perl 5.8), and then for pkg_install to
 know that if something was built to depend upon perl 5.8
 and an incompatible perl version(5.10) has been installed,
 then that one does not satisfy the dependency, even though a
 simple evaluation to perl>=5.8 says it does.
 
 That is, perl5.10 would effectively add '&&perl<5.10' to every
 perl dependency that used a perl < 5.10, rather than attempting
 to have guessed that in advance in every package that uses perl.
 
 As a concept this is not that hard, even as an implementation it
 would not be too hard, the problem is that just this, alone, only
 scratches at the surface of the dependency version issues with
 pkgsrc.
 
   | Do you have a better suggestion?  If not, I think
   | we should close this PR.
 
 This PR, sure, close it, its intent was just to make sure that a
 massive revbump (or some equivalent) was done on all of the p5-* packages
 (and anything else) do they'd all end up being recompiled or at least,
 be known to be out of date.   That has happened, so this PR served its purpose.
 
 kre