RFE: Running createrepo_c on individual update packages

[Michael Schwendt]

On Thu, 22 Oct 2015 18:28:45 -0400, Matthew Miller wrote:

> >   C_CREATEREPOLIB: Warning: Bad epoch in version string "%{epoch}:3.0.0-3.fc23.git0.9.2" for dependency "blktap(x86-64)" in package "blktap-devel-0:3.0.0-3.fc23.git0.9.2.x86_64"
> >   C_CREATEREPOLIB: Warning: Skipping this dependency
> 
> Are there other problems you're hoping to catch or is it specifically
> this one?

If there are more "invalid input data" cases hidden under the carpet by
createrepo_c's RPM metadata parser instead of rejecting them, those could
be caught, too.

> Because it seems like making rpmbuild fail in this case, or at the
> very least _some check in koji_ like the rpmgrill thing.

I've had a look at rpmbuild:

  https://bugzilla.redhat.com/show_bug.cgi?id=1251453#c1

The source code is not commented enough. In a place where macros are
expanded already, it explicitly overrides the EVR character set checking
method with a whitelist that includes "%{}". No idea why it does that.

I've had a look at rpmgrill, too, yesterday. It's written in Perl and
doesn't look pretty. Lots of FIXME placeholders everywhere in the code,
and even the manual pages are plastered with FIXME placeholders.
F22 package is from Dec 2014. Upstream has moved to github, but lots
of files are as old as 6-36 months.