Apache OpenOffice (AOO) Bugzilla – Issue 116845
smoketest should fail in non-product build when assertions are raised during the test
Last modified: 2017-05-20 10:30:47 UTC
Currently, the smoketest (in smoketestoo_native) succeeds even if assertions are raised during one of the tests. In order to ... increase the awareness of the significance of assertions as tool to ensure product quality, the smoketest should fail in such cases.
@fs: You should close this as a duplicate of 109142.
@sb: don't think so - aborting assertions are still different from assertions which "just" make the smoketest fail, aren't they? *** This issue has been marked as a duplicate of 109142 ***
ouch, clicked the "resolve as duplicate" button too fast ... reopening
@fs: Different, in what way? Either you take "the significance of assertions as tool to ensure product quality" seriously, or you don't. I will object to any enlargement of our code heap aimed at a mechanism that lets smoketest fail upon assertions. Keep it simple. Let assertions abort---always.
"In what way"? Compare "every assertion ever thrown during running OOo kills the process" with "assertions thrown during the smoketest mark the smoketest as 'failed'", please. I will object any change to our code which we're in no way prepared for. We already have acceptance problems with the current state, where developers do not use non-product versions "because there are so many assertions". Having a non-product version where an assertion isn't an ignorable message box, but implies termination of the process, won't get any better acceptance. It is simply too early to implement this. However, implementing smoketest breakage upon assertions is something which won't change the daily work directly, but still give assertions at least a *little* more attention, and thus educate people.
@fs: I can understand your motivations, but do not agree with your conclusions. If we want to take assertions seriously, we have to bite the bullet and get rid of their occurrences. An intermediate step that increases the code heap is not worth it IMO.
@sb: Well ... we should discuss this with a broader audience, then. I am very reluctant to introduce the change you propose in issue 109142, simply because I think we won't get acceptance for this. And such a chance isn't worth anything without acceptance amongst the developers. If you can convince me that a majority of developers will accept this - then let's do it. If not, we should do this incrementally.
*** Issue 116814 has been marked as a duplicate of this issue. ***
fixed in CWS debuglevels
fs->sb: please verify in CWS debuglevels