Dan Price wrote:
At long last, here's a cut at my test cleanup wad. I would
appreciate comments.
http://cr.opensolaris.org/~dp/pkg-testfu/
8265 EINVAL on rmdir running "make test"
9828 testsuite leaves turds in my tempdir
12257 setup.py doesn't respect run.py options
12259 run.py -x can cause traceback for failed api tests
13715 commonize duplicated misc_files mechanism
13717 lift debug functionality from cli suite into pkg5unittest base class
14215 test suite can publish using relative paths, saving complexity
14216 forked indexer process can race with testcase teardown
14217 Need a run.py option to enable debug output from suite
14218 Need a run.py option to archive the remains of failed tests
14219 test suite vomits when $DISPLAY isn't set
14220 test cases which depend on sun studio compilers need work
To summarize:
- Fix a lot of bugs in the harness
- New -a option for archiving failed tests
- New -d option with extensive debug output including the
docstring for the test case.
- Make the testsuite treat baseline mismatch as the true
indication of failure
- Elide output for tests expected to fail (unless -f)
- Move functionality into common places.
- All test cases can now take advantage of e.g.
SingleDepotTestCase without gyrations.
- All test cases can now call self.debug().
- Output is (I think) cleaner, neater, easier to
understand.
- Commonized boilerplate misc_files code.
- Changed all publication to use relative paths (a lot simpler)
- We no longer create random files in /tmp.
- Depot shutdown more reliable when things go ill.
Thanks,
-dp
t_api_search.py:
Would lifting do_*install up a level make any sense? So they're not
duplicated in the two classes? It makes sense to commonize them with the
ones in t_api_install at some point, but I'm fine with that as a P4/5 bug.
t_smf:
nit: 44, 51-56, 86-91 I think we usually indent those lines 4 spaces,
and would also indent "passwd" 4 spaces
t_pkgdep:
nit: 131-132 indentation looks weird
nit: 969 wrap line?
t_pkgsend:
536: assertEqual?
pkg5unittest:
171-180: I'm just generally confused about how this makes test_root readonly
264: when would self not have killalldepots?
304: wrap this in a try/except as well for the case of somone spamming ^C?
441, 444: I think isinstance is the preferred method of testing this.
Also, as Danek recently pointed out to me, we should probably then
compare against basestr instead of str.
t_setup_teardown:
Might be worth having a test that blows up before the superclass's setup
method is called.
Thanks for making all these changes. I can't wait for this to make it to
the gate.
Brock
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
