On 09/07/11 10:54, Shawn Walker wrote:
On 09/06/11 17:29, Brock Pytlik wrote:
Patch:
--- old/src/web/en/catalog.shtml Tue Sep 6 17:24:32 2011
+++ new/src/web/en/catalog.shtml Tue Sep 6 17:24:32 2011
@@ -123,7 +123,7 @@
flist = []
if selected_val and selected_match:
ver = selected_match
- pattern = "*,%s-%s" % (ver.build_release, ver.branch)
+ pattern = "*,%s-%s.*" % (ver.build_release, ver.branch)
flist, unmatched = catalog.get_matching_version_fmris(pattern)
else:
flist = [f for f in catalog.fmris()]
Bug:
18862web package catalog version matching fails to list all valid
packages
I tested this by comparing the packages listed for a 173 repo when
browsed through the BUI
(found by taking the source of the webpage then applying this
transformation:
grep "info" catalog.shtml.htm | cut -f 2 -d '>' | cut -f 1 -d'<'>
catalog.list )
with the output of this pkg list command:
pkg list -af -g /tank/repo-173/repo/repo/ |grep '173' | cut -f 1 -d ' '
The same packages, in the same order, were listed. I also checked the
BUI output of a 151a repo with the
output of pkg list (this time just comparing the number of packages) and
found that the BUI produced the same
number of packages as pkg list -af -g /tank/repo-151a/repo/ | grep
"151.0.1" | wc -l.
I'm happy to do more testing if people have specific tests they'd like
to see.
While this gets us back to roughly status quo, I would note that this
doesn't address some of the comments in the bug, specifically:
* doesn't filter by incorporation surface, so userland packages not
delivered in the same build won't be shown
* shows more packages than actually delivered for a given build
(i.e. 151 now shows 151 and 151a packages, while 151a only shows
151a)
However, I think this is a reasonable hack for the moment until we can
institute proper incorporation filtering. It's better than the
current situation.
Yep, that's all I was going for. I couldn't see a reason to let the
perfect be the enemy of the slightly better. I you'd prefer, I can file
a new bug more specific to the issue I'm fixing and return 18862 to its
previous state or I can file a new bug for incorporation filtering. Let
me know which you'd prefer.
Brock
-Shawn
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
