On 02/25/10 07:23 PM, Danek Duvall wrote:
Shawn Walker wrote:
Greetings,
The following webrev contains fixes for the following issues:
14832 pkg info traceback for renamed packages with no
variant-applicable dependencies
14866 missing info retrieval unit tests for renamed packages
webrev:
http://cr.opensolaris.org/~swalker/pkg-14832/
So now if the package is marked renamed, it'll show up as "State: ...
(Renamed)", but there won't be a "Renamed to:" line. Should we have that
line, but have it be empty? If so, then I think all you have to do is
change pi.dependencies[0] to pi.dependencies[0:1] (assuming that we get
back an empty iterable instead of None, if there aren't any dependencies).
I thought there could be multiple dependencies for a rename?
I'm a bit sketched out by the fact that you're groping through the raw
action strings, rather than converting to actions, so it's good to see this
change for the renamed case. Care to fix the obsolete case, too?
There's a massive performance difference between using the raw string
vs. parsing the action string minimally. Tens of seconds last I
checked, although that could have changed since we moved to python2.6
and have made various other changes.
As far as the obsolete case, I didn't think we'd reached agreement about
whether those were variant-specific or not. Given that the solver
currently does account for that...
I'd rather address the obsolete case separate from rename since there's
a nice, clear delineation between the two sets of changes needed.
The other tests in api.t_api_info.py are missing from baseline.txt. Should
you bother adding this one? I presume at some point we should get rid of
all passing tests from that file.
I wasn't aware we had adapted that approach for baseline.txt; the -g
option still adds passing tests to baseline.txt (which is what I used).
Might it not be safer to create an image with a dummy variant, rather than
relying on the arch variant?
Hmm. That would work too; I honestly hadn't thought of using a dummy
variant.
Is [email protected] actually used anywhere in test_2_renamed_packages?
Nope; removed.
Is it worth having a test where renamed is explicitly set for the current
architecture but the depend target is explicitly set for the opposite?
Sure.
line 456 (in the api test): no need for a continuation character inside
parens.
Done.
Thanks,
--
Shawn Walker
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
