strange failure for dependency against libdb-4.5.so

[Michael Schwendt]

On Sat, 18 Aug 2007 00:51:09 +0900, Mamoru Tasaka wrote:

> Mamoru Tasaka wrote, at 08/18/2007 12:45 AM +9:00:
> > Jindrich Novy wrote, at 08/17/2007 09:08 PM +9:00:
> >> On Fri, Aug 17, 2007 at 11:57:51AM +0300, Ville Skyttä wrote:
> >>> On Friday 17 August 2007, Jindrich Novy wrote:
> >>>> db4 was updated recently to 4.6.18, and older db-4.5.20 was moved to
> >>>> compat-db
> >>> compat-db seems to have some self-obsoletion issues, that may have 
> >>> something to do with the problems:
> >>>
> >>> Obsoletes: db4 < 4.6, db4-devel < 4.6, db4-utils < 4.6, db4-tcl < 
> >>> 4.6, db4-java < 4.6
> >>> Provides: db4 = 4.2.52, db4 = 4.3.29, db4 = 4.5.20
> >>> Provides: db4-devel = 4.2.52, db4-devel = 4.3.29, db4-devel = 4.5.20
> >>>
> >>> The Obsoletes should probably be < 4.2.52 instead of < 4.6.
> >>
> >> The original intention was a possibility to replace an existing old db4
> >> with compat-db without conflicts. I decreased the obsoletes and build
> >> new compat-db so it sould be ok now.
> > 
> > Umm.. still no good. Currently I cannot figure out how to resolve
> > this issue.
> > http://koji.fedoraproject.org/koji/taskinfo?taskID=107120
> > 
> > (according to i386 half‐finished (cancelled) root.log, buildroot seems to
> > install new compat-db 4.5.20-2.fc8:
> > http://koji.fedoraproject.org/koji/taskinfo?taskID=107111 )
> The latter one is not 107111, but:
> http://koji.fedoraproject.org/koji/taskinfo?taskID=107121

A "cvs diff -u -r 1.22 1.24 compat-db.spec" shows that the Provides/Obsoletes
were added in compat-db-4_5_20-1_fc8 and after compat-db-4_3_29-2_fc7

If RPM in the build environment is a version that obsoletes packages
based on virtual provides, these Provides/Obsoletes in compat-db are
the culprit. In short, when compat-db "Provides: db4 = some-low-version",
the "db4" packages with a higher version upgrades compat-db. Some
background can be found here: https://bugzilla.redhat.com/111071
It has effected multiple packages since FC =~ 1

Recently, Panu has mentioned that this "feature" in RPM will go away
in a future release of RPM. But until then, don't use such virtual
provides. They break.