On Mon, May 28, 2007 at 11:39:34AM -0400, Dan McGee wrote:
> On 5/28/07, Roman Kyrylych <[EMAIL PROTECTED]> wrote:
> > 2007/5/28, Michael Towers <[EMAIL PROTECTED]>:
> > > Is it now allowed to have a '.' in the 'pkgrel'? Was it always allowed?
> > > If so, is it necessary? If not, why does pacman accept it?
> >
> > It is allowed to state that this is a change for one arch only (i.e.
> > x86-64-specific).
> 
> I've brought this up elsewhere, but I will state an opinion on this
> thread- I think keeping pkgrel to an integer is a much smarter idea.
> First off, that version number is hard to parse by a human- far too
> many dashes and dots to know which is what. Second, the only change
> I've ever seen that uses a .1 increment is adding an architecture to
> the arch array. This is completely unnecessary- there is no need to
> bump the pkgrel at all, and if you do want to bump it, then just bump
> it a full integer.
> 
> Thoughts?

Are you sure about that?  When Andy and I first talked about it, it was for
when arch64 needed to change a PKGBUILD to release their version of a
package.  That way foo-1.2.3-1 always meant exactly the same PKGBUILD.
foo-1.2.3-1.1 was a different PKGBUILD that just didn't happen to be
released for x86 but *was* released for arch64.  The next time that package
is released it will probably not need to be modified for arch64, so the
arch64 version will be foo-1.2.3-2 just like the x86 version.

I never thought the pkgrel was so confusing.  Maybe it's because I looked
at the pacman code and realized it's treated exactly the same as the
version.

My fear with updating a full integer is having people say, "why hasn't x86
updated to -5?!? arch64 has -5, why does x86 only have -4?!? are we that
far behind?!?"  I always thought of the -1.1 notation as being a revision
of the revision.

Jason

Attachment: pgp0KgQkphR41.pgp
Description: PGP signature

_______________________________________________
arch mailing list
[email protected]
http://archlinux.org/mailman/listinfo/arch

Reply via email to