On 2013-10-23, Neil Horman <nhorman(a)redhat.com> wrote:
On Wed, Oct 23, 2013 at 04:58:34AM +0000, Petr Pisar wrote:
> On 2013-10-22, Neil Horman <nhorman(a)redhat.com> wrote:
> > On Tue, Oct 22, 2013 at 03:03:44PM +0000, Petr Pisar wrote:
> >> Is it healty to execute rpmbuild while building a package?
> >>
> >> I have tests for perl dependencency generator filters. I.e. the tests
> >> build a package using rpmbuild with redefeined all the `_*dir' macros,
> >> then use librpm to query requires and provides from built package, and
> >> then do checks on the values.
> >>
> >> I'm thinking how to plug the tests into Fedora. The simplest solution
is
> >> to run the tests in %check phase of a package (I don't know which one,
> >> but it does not matter which one).
> >>
> >> I've already heard warnings that calling rpm from package scriptlets is
> >> not recommend.
> >>
> >> What the situation with rpmbuild?
> >>
> > I'm not sure how re-entrant-safe rpmbuild is, but doing the above
> > seems a bit dodgy in general. Could you instead package the test
> > separately from the dependency generator rpm, make the latter
> > depedent on the former, and then use chain-build in koji to build
> > them at the same time?
> >
> Do you say to create a dummy package in Fedora? Package which itself
> has dummy (possibly) unsatisfied dependencies? Package that ends up
> in Fedora repositories? That's fishy.
>
Um, no. All I was suggesting was that you build the package tests
separately from the package itself, as its own rpm. Neil
Well, but the purpose is to test the output of rpmbuild. If I had to
prebuild tested data manually or out of koji, it would not serve the
purpose.
However I know, this is the last option. I've already redesigned the
tests to be runnable from %check phase as well as on demand after installing
them into system.
-- Petr