I thought it would be easy to query rpm what's its name. But I do not have the rpm at the time I want to query it.

Bill, would not this be better handled by scheduler?
- It is scheduler sending out the data in form which is not safe to use
- Scheduler has access to the RPMs and could query them (I have heard claims it is likely to get N-V-R parsing wrong especially with regard top Epoch. However epoch is not used by our tests.)

This will still not catch the name changes due to the repo-prefix part not constant. Fortunately this does happen rarely enough to ignore.

Here my naive attempt to handle on harness side. Not that difficult but not sure if correct:

Let's get rid of epoch first:

  s/^\d\+://
  s/-\(\d\+\):/-/
  s/:\d\+//

Let's remove -V-R part of N-V-R:

  s/-[0-9.]\+-[^-]\+$//

Testing:

  for each rpm-package in rpms repo (e.g. `bkr task-list`):
  - get the N-V-R
  - get the N part
  - query the package by N (e.g. `yum info`)
  - if it can not be queried:
    - the parsing is wrong

I agree the server should provide the correct name for you. 

Currently the node looks like this:

<rpm name="rh-tests-distribution-install-ext4-lvm-1.0-2.noarch.rpm" path="/mnt/tests/distribution/install/ext4-lvm"/>

Should I just drop the VR.rpm part?

<rpm name="rh-tests-distribution-install-ext4-lvm" path="/mnt/tests/distribution/install/ext4-lvm"/>

(In reply to comment #2)

> Should I just drop the VR.rpm part?

Yes, please. Getting the name part only would be fine.

pushed to gerrit for review