The idea of nesting phases seems nice & useful to me. Perhaps we could make use of the possible modification of journalling-related functions to also add support for phase aborting. Which brings some new questions:

 * How to implement a phase-skip?
 * Should the phases rather be written as functions?

+1 to phase nesting

Phase aborting is really a desired function to have - it lives in our specs from the very beginning. I just still don't know how to handle clean-up. No idea how to skip phases (I've experimented with noexec option, but once this is turned on, one can't turn it off again)

I don't like the idea of functions - it brings a lot of troubles (variable validity, nesting...).

I wonder if we could just ignore the cleanup.

+1:

Once we have phase nesting as suggested in bug 464155 comment #2, it would be nice if each rlPhaseEnd generates result as rlJournalPrintText, but only for that phase (and it's children).

When I click in the WebUI on "Test" phase which failed, I do not want to see "Setup" phase which was OK. We could also have some fake "This is a test's real end" phase, which would be parent of all phases (and so, it would contain all the phases).

And more tricky part: would be nice, to have TESTOUT.log divided phase-by-phase. Like TESTOUT-Setup.log, TESTOUT-Test.log and TESTOUT-Cleanup.log - in the simplest case.

I've checked in the first step, under obvious rule: this bugfix enables very simple nesting: it places asserts and logs into the right phases, but it doesnt reflect the nesting itself in the log.

Created attachment 341147 [details]
simple bugfix

Created attachment 341149 [details]
real simple bugfix

This package has changed ownership in the Fedora Package Database.  Reassigning to the new owner of this component.

As we could live without that until now I do not see much demand on this.