On 04/14/10 12:54 PM, Danek Duvall wrote:
Shawn Walker wrote:
Fixing both is fine, but I don't see how the changes to
__filter_install_matches() are involved.
Oh, that's easy. The old code was wrong. You actually fixed several
bugs in how it tried to eliminate multiple matches when you changed
it. Since ambiguous name pairs could result in multiple matches, it
seemed relevant to fix. Did you not realise you fixed something else
too? :)
I vaguely remember fixing something in there, but I don't at all remember
the details of it. The actual change there, particularly with bits of your
catalog v1 change mixed in is pretty subtle. My changeset primarily
rewrote that routine to use sets instead of dicts and, IMHO, made it a bit
easier to use. But aside from the change to take obsolete packages into
account, the only substantive change I made was in the "Next" stanza
(starting line 2283 at the time), to always add the package stem to mnames.
That same change isn't in your version of the change because the catalog v1
change hadn't happened to make it necessary.
Right, I realise your changes were primarily focused on fixing
obsolete/renamed bits which I extracted as best I could.
The short summary is the old filtering code was wrong because it based
eliminating matches purely on the stem instead of the installed, etc.
state of specific stem versions.
So yes, I think there's a few unfiled implicit bugs that got fixed
whenever this section got reworked.
Cheers,
-Shawn
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
