I think it would be a good idea to move check-buildroot and check-rpaths from
rpmdevtools to either rpm (upstream or as distributed by Red Hat) or
redhat-rpm-config. Reporting against rpm because I believe these scripts are
generic enough to benefit a wider user base than just Fedora/RHEL users.
The scripts are in the /fedora CVS root on cvs.fedora.redhat.com, in
For backwards compatibility, it would be nice to have these installed directly
in /usr/lib/rpm/check-* (for which rpm-build sounds better than
redhat-rpm-config to me). At least check-buildroot could be also enabled by
default in the default rpm-build/redhat-rpm-config configuration - it is already
enabled in all plague and koji build roots as far as I know - by adding it to
In addition to that these scripts would be available to a wider user base if the
move would be done, one reason driving this is that I'm in the progress of
adding some new goodies with new dependencies to rpmdevtools and don't want that
to result in those dependencies being pulled into minimal build roots.
F-8+ would be a nice time for this change.
Doing the work to get these into redhat-rpm-config. Since redhat-rpm-config
will be providing the macro to run these, I think having them in
/usr/lib/rpm/redhat/ is best for now. This is orthogonal to getting these in
upstream RPM. If they ever show up in upstream RPM, we can adjust
redhat-rpm-config to drop them and point to the new path for running them. I
would suggest filing this RFE at rpm.org Trac.
The bad thing about /usr/lib/rpm/redhat/ instead of /usr/lib/rpm/ is that a lot
of people and setups have their ~/.rpmmacros and friends configured to invoke
the scripts from /usr/lib/rpm/ - that's where rpmdevtools installs and
rpmdev-setuptree configures them.
I don't think these settings are fixable from a package, therefore /usr/lib/rpm/
would be very much recommended for backwards compatibility and less grief.
http://people.redhat.com/jkeating/git/redhat-rpm-config.git has changes
necessary to add these files to redhat-rpm-config, and enable them by default.
(also I cleaned up the dist stuff in a second commit)
Jeremy, can you apply this? This is part of the effort to remove the 'groups'
repo in mock.
Hrm. a bit of cross traffic there. I suppose I can make the scripts go into
/usr/lib/rpm, will just mean special casing them in the Makefile, and I do so
hate editing Makefiles :(
Villa, care to provide a patch to Makefile that moves them somewhere different?
Created attachment 158375 [details]
Move check-* to /usr/lib/rpm, add test suite
(In reply to comment #4)
That's wool in Finnish :)
> care to provide a patch to Makefile that moves them somewhere different?
Here goes. Also included is the check suite for check-rpaths from rpmdevtools.
My git repo now has those changes in it. Thanks Ville (<-- look, I got it right!)
FWIW, I've zero objections to adding the check-* scripts to upstream rpm even
for 126.96.36.199, but enabling by default (or not) will need to go to
redhat-rpm-config at this point.
How soon would 188.8.131.52 show up? I'd like to get these things done ASAP as they
are part of a larger change to mock.
I can stuff the scripts in, bundle up a 184.108.40.206-rc2 tarball and have it in next
rawhide build if you wish. Final 220.127.116.11 is due out very shortly anyway, these
rc's are simply about being very cautious.
That would be most awesome. IF you do that, I'll adjust my redhat-rpm-config
changes to just enable it by default (and fix the %dist stuff too). Next
rawhide build would be most awesome.
http://people.redhat.com/jkeating/git/redhat-rpm-config.git now just has changes
for enabling the post %install buildroot checking, and dist cleanups.
Added to rpm 18.104.22.168-rc2 and currently building that for rawhide. Assuming
nothing blows up, will be in next rawhide push.
Confirmed, in the new rpm. I'll consider this closed.
Thanks. One last bit:
--- scripts/check-rpaths-worker~ 2007-07-04 14:57:19.000000000 +0300
+++ scripts/check-rpaths-worker 2007-07-05 23:41:11.000000000 +0300
@@ -65,4 +65,2 @@
-* 'check-rpaths' is part of 'rpmdevtools'.
Looks like Panu got around to fixing that bit in hg some time ago. Thanks.