I see journal as the primary source of information on a test run, that should contain all necessary information (see also 473911). All other reports should be derived/generated from it. And because assert command output is an important information, it should be included (see discussion of bug 477138). As of now assert part of journal looks like this: <log> <phase name="fazicka" result="FAIL" score="1" type="FAIL"> <test message="Running 'echo abc' (Expected 2-7-3-5, got 0)">FAIL</test> </phase> </log> after output inclusion: <log> <phase name="fazicka" result="FAIL" score="1" type="FAIL"> <test message="Running 'echo abc' (Expected 2-7-3-5, got 0) result="FAIL"> abc </test> </phase> </log>
Huh, is this really a good idea? Some outputs can be pretty huge and I would like to keep the journal human readable... But you are definitely right, this should be stored somewhere... I'm thinking about an alternate solution, by creating something like a "output log" supplement to the journal, where we would put the rlRun command's output, and the journal would contain a link of some sort to this supplemental log. It's just a quick thought from the top of my head - and it's a bit "cosmetic" compared to your original proposal...
Actually this can be nicely hidden in the journal generated by rlJournalPrintText, but `DEBUG=true rlJournalPrintText` can be another story :) IMHO journal (XML one) should contain all info generated by the test.
Seems to me a good idea to include all test related output in the journal. Perhaps we should just set up some reasonable limit X to prevent the journal to grow too much (maybe include head -X and tail -X or grep -i FAIL when logs are really very long, e.g. for compiling). If we need all the output there's always the TESTOUT.log available.
*** This bug has been marked as a duplicate of bug 840618 ***