Subject: Re: CVS commit: pkgsrc/pkgtools/pkg_install/files
To: Thomas Klausner <wiz@NetBSD.org>
From: Alistair Crooks <email@example.com>
Date: 06/10/2005 22:11:09
On Fri, Jun 10, 2005 at 05:54:38PM +0200, Thomas Klausner wrote:
> On Fri, Jun 10, 2005 at 04:43:45PM +0100, Alistair Crooks wrote:
> > You may not, but others have valid reasons, it seems to me.
> Good, then let them voice them. I haven't seen any except
> for hubertf's, who was starting from a wrong assumption (that
> it _did_ compile on AIX before), which dillo corrected.
Dan bumped the version number in the first place.
Characterising Hubert's objections as being because of a
misunderstanding is pushing it a bit, I feel.
To put it bluntly, the sources changed. The version number should be
bumped. You may not like it like that, but you are proposing to have
two different versions with the same version number, and that is
unacceptable from a software engineering point of view.
> This is the question I want to answer here: The change was a
> compilation fix _only_, and historically we haven't bumped PKGREVISIONs
> for that. Are you arguing we should start that? If yes, why?
I'm saying that the sources changed (you don't dispute this, although
you tried to play down the significance of this). As such, the
version number must be changed. Not to do so is sloppy, and invites
problems. Citing historical precedence for this is neither here nor
there - we should endeavour to do the right thing.
You still haven't answered the question of why you do not want to bump
the version number of pkg_install. It used to get bumped whenever any
file in the CVS repository was changed. Then you changed it to a
version.h style hardcoded version number. If you are now saying that
the version should only get changed on major bumps, that is something
completely different, and something with which, as you've probably,
guessed, I'm very unhappy.