[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Bug 226377] Merge Review: rpm



On Tue, Aug 28, 2007 at 10:21:14AM +0200, Dominik 'Rathann' Mierzejewski wrote:
> On Monday, 27 August 2007 at 23:11, Axel Thimm wrote:
> > On Mon, Aug 27, 2007 at 07:15:07PM +0200, Dominik 'Rathann' Mierzejewski wrote:
> > > On Saturday, 25 August 2007 at 01:01, Axel Thimm wrote:
> > > > On Fri, Aug 24, 2007 at 11:44:36PM +0200, Dominik 'Rathann' Mierzejewski wrote:
> > > > > > I think rpm5 in Fedora is dangerous. At the very least it reverses
> > > > > > the ordering of letters and digits and thus breaks a ton of packaging
> > > > > > techniques. Any *-1.fc8 -> *.1.1.fc8 upgrade path is busted for
> > > > > > example.
> > > > > 
> > > > > Are you 100% sure? I've seen Jeff deny this on IRC.
> > > > 
> > > > Index: rpmvercmp.c
> > > > ===================================================================
> > > > RCS file: /cvs/devel/rpm/lib/rpmvercmp.c,v
> > > > retrieving revision 2.8.2.2
> > > > diff -u -b -B -w -p -r2.8.2.2 rpmvercmp.c
> > > > --- rpmvercmp.c 27 Aug 2006 12:58:19 -0000      2.8.2.2
> > > > +++ rpmvercmp.c 26 Apr 2007 22:36:26 -0000
> > > > @@ -76,7 +76,7 @@ int _rpmvercmp(const char * a, const cha
> > > >          /* different types: one numeric, the other alpha (i.e.  
> > > > empty) */
> > > >          /* numeric segments are always newer than alpha segments */
> > > >          /* XXX See patch #60884 (and details) from bugzilla
> > > > 	 #50977. */
> > > > -       if (two == str2) return (isnum ? 1 : -1);
> > > > +       if (two == str2) return (isnum ? -1 : 1);
> > > >          if (isnum) {
> > > >              /* this used to be done by converting the digit segments
> > > > 	     */
> > > 
> > > The code present in current CVS HEAD is a bit different. Also I can't find
> > > this change in rpm5.org's CVS. Which revision were you comparing to 2.8.2.2?
> > 
> > I copy and pasted this from an original mail to rpm-devel by rpm5's
> > developer. Any later refactoring didn't undo the change AFAIK.
> 
> I'm willing to test this with current CVS HEAD. Do you have a test case?
> If not, I'll try testing it on my own when I get some free time.

Test case? See above: "Any *-1.fc8 -> *.1.1.fc8 upgrade path is busted"
(where the second should had read hyphen instead of the first dot, but
you get the picture).

But you don't need to empirically test, read the code especially the
parts that advertize full dpkg compatibility achived by putting
letters higher than numbers.

Even refactored you either are compatible to rpm.org or dpdkpg in
ordering, both is not possible.
-- 
Axel.Thimm at ATrpms.net

Attachment: pgp85R03S0j41.pgp
Description: PGP signature


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]