On Jan 27, 2008, at 8:56 AM, Peter Kalbus wrote:
> as i'm currently porting rpm5 to qnx (which is nearly finished), i
> now think about, what is the best way in getting the port tested.
>
Here's the basic drill I use to test rpm:
Build rpm from cvs/tar, type "make install".
Build rpm using rpmbuild and rpm.spec.
Sign the *.rpm packages, and install with rpm -Uvh.
Import a pubkey using --import.
Verify the package signature using rpm -Kvv.
Query with rpm -qa to insure installed.
Rebuild the rpmdb using --rebuilddb.
Verify the installation with rpm -V rpm.
Erase with rpm -evv.
If the above sequence "works" (you will need
other options like --nodeps to succeed), then
rpm is functional.
> for sure, as time goes on, we will see, what happens ;-) but that's
> not, what makes user happy ;-)
>
users basically cannot be full satisfied.
> is there any testsuite or similar stuff, which could be used for
> that purpose?
> how this is performed for the other supported systems?
> perhaps there are already peaces included in the distribution?
>
There is no "official" testsuite. I usually just type
some variant of the sequence into a script when I need it.
There is a sanity check in devtool that kept HEAD functional
during rpm-5.0 devel. I also had a script that ran valgrind
across the sequence above.
There's a test harness in rpm-4.4.9 that was developed to get --rollback
(which is an order of magnitude more complex than basic
rpm operations) functional that could be extended.
> i know, that the question is a little bit general, but any small
> information is helpfull. i'm not searching for a final solution
> (okay, if existing i would be very happy), but rather trying to get
> an idea, about the way rpm5 will address that point.
>
The test harness in rpm-4.4.9 is likely the best choice if you
need/want a test suite. The tests that are there are perhaps
too complicated; a well formed test harness should give
a strong hint about what needs fixing, and always fail
the same way. Many of the tests in the test harness
copy the existing system rpmdb into place, which can
lead to obscure failure modes in the test suite. The tests
also focus on --rollbck, not basic rpm functionality.
hth
73 de Jeff
Received on Sun Jan 27 15:54:30 2008