| Age | Commit message (Collapse) | Author |
|---|
|
We had some dependency issues between migrations that needed to be
explicitly defined in order to get things fully moving, and do to some
braindeadness in Django tests not including the project url config, we need
to do some clever business when using the url tag in the base template so
tests don't doe with a NoReverseMatch exception.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Now that we cache everything, we need to ensure anyone doing edits and
such gets the live data and not some cached version that was already
updated and is now stale. Add the never_cache decorator to any of the
CUD screens as well as a few others that might benefit from always being
regenerated.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
This is also another step in making sure we don't unnecessarily access the
Django session object and thus add a Vary cache control header we don't
want.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Instead of putting 'www.archlinux.org' all over the place, use the Django
sites framework to pull the site name out of the database. Now these
amazing things will work if you are running locally and decide to change the
site!
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
After adding filename to the database, this is a rather simple request (see
FS#19546). Right now the "randomly" chosen mirror happens to always be
mirrors.kernel.org as it is the only one filed under the 'Any' country which
is what we screen on. Perhaps this logic could be improved in the future but
I don't see these links being all that high traffic anyway.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
We were still using the user-facing package IDs here which we have
eliminated everywhere else Change it to use the same user-friendly URL
pattern we are using elsewhere.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
|
|
Attempt to mark all packages built from the same PKGBUILD out of date at the
same time by using pkgbase instead of pkgname. Ensure testing and
non-testing repos are segregated as well.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
This allows some of our queries to get a lot simpler as well as removing the
pkgbase_safe property added earlier.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
This is an attempt to fix our long-standing problems dealing with maintainer
information. Move the actual maintainer information off of the package model
into a PackageRelation object, which has some flexibility to later represent
more than just maintainership.
This solves multiple problems:
* If a package gets accidentally deleted, so did the maintainer info
* Testing packages have always shown up as orphans
* With split packages, it was easy to miss some of the sub-packages
This commit does not include the deletion of the original maintainer column;
that will come at a later time when I feel more confident that the data was
migrated correctly.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Get rid of the copy/paste by including the sub-template.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Since the same URLs serve two different responses based on the request being
AJAX or not, we want to ensure we don't cache the wrong one and serve it up
incorrectly.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
This will put the filelist inline on the package details page if using a
capable browser. It should still fallback to a separate page if necessary
(e.g. all those users using links on the site).
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Move it away from the numeric pkgid-based view of old to the new pretty URL
format. This does nothing to actually make the view show files (or even
provide a link to it), but that will come in future commits.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Use update() instead of for x in y: save().
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
We can pull the common queryset out into a local variable, as well as using
the mass-update function provided on a queryset instead of resorting to
iteration.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Django already provides an implementation of urlencode that is Unicode-safe.
Use it instead of our own version.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
The len() call was killing performance, especially in the default click-thru
case where we show the packages list without any filtering (only sorting).
Also use a well-placed select_related() to further drop the query count. As
noted by the TODO, there is still more that can be done here to improve
performance.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Unnecessary, and lets us standardize on not using it everywhere.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Implements FS#14185. It is a bit more complex than listed there as I wanted
to not hardcode the URLs in the descriptor file; to do this we need to make
it a template and fill some things in. We also need to serve the file using
the correct mime type.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
And honor the packager's notify flag, as Pierre pointed out.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Especially when looking at packages, we always want the arch and repo.
Another big hunk of changes deals with the very inefficient signoffs code.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Signed-off-by: Evangelos Foutras <foutrelis@gmail.com>
[Dan: made a few other small touchups]
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
Makes it easier to find a user since we have quite a few. FS#18055.
Signed-off-by: Dan McGee <dan@archlinux.org>
|
|
|
|
|
|
|
|
|
|
archweb_dev. This is the original commit message by Dan:
"Make package SVN links always work
This should clean up the links for all varieties of things- different arches
(including any), different repos (community and community-testing), and
split packages. All of the logic is in one place now and any further changes
should be made to the method on the package object."
|
|
|
|
|
|
disown. Probably occurred by a press of the back button.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
