On Thu, Feb 6, 2020 at 8:38 PM pmkellly(a)frontier.com <pmkellly(a)frontier.com>
wrote:
I have completed a first draft of my proposed revision to the
"QA:Testcase package install remove" Base test case. I got some input
based on my initial e'mail on this. I believe I have addressed those
items.
I believe it is an important part of testing if an install happened
properly to see if the application will open. To best enable that test,
I have specified the two packages to be used in the testing. They are
reliably available; even in the repos available to Rawhide drops. They
are stable. This helps avoid testing with packages that my not be stable.
I believe the purpose of the test case is to have a very generic test case
that works in all environments (not just graphical ones) and applies to all
package managers (notably dnf, packagekit, gnome-software, discover,
dnfdragora, etc -- but not rpm, that is one level below them). It relies on
rpm working correctly in the bottom, and it doesn't care about the actual
software that gets installed. The point is just to ensure the selected
package got installed or removed (which can be easily verified with rpm).
Checking app icons in the desktop menus or launching them is a departure
from this goal. And it doesn't work e.g. in Server environment. Also, I try
to avoid hardcoding package names like you did (install packages foo and
bar), because it's the random nature of things that helps up detect issues
during manual testing. We can hint the package names, if really needed.
I'm not saying the changes are wrong, it just seems to be a different test
case. Something that would perhaps fit more into desktop testing and be
tied to graphical package managers, than to be a completely general test
case.