| Age | Commit message (Collapse) | Author |
|---|
|
Missed to update the migrations in the py3 update
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Seems django 1.11 automatically enables caching template loader, which
of course breaks the ability to make any changes to the pages of a
website without restarting it. And there is no way to turn it off other
than to explicitly configure individual loders (the logic to turn it on
in non-debug configurations is hardcoded and cannot be changed).
|
|
If a single fetch from a blog grabs more than <n> entries (the default
being 4), this usually indicates a misconfigured blog and it will end up
spamming the outgoing feed and twitter, even if later corrected. For
that reason, when this happens, mark all the newly fetched entries as
hidden, and send a notification as such. In the event that they are
actually correct, the user can go in and manually unhide those entries,
which will make them show up again.
|
|
Shouldn't use print directly, since it doesn't respect the verbosity
setting.
|
|
We only follow the redirect if the URL is *exactly* the same except it
has https intead of http in it. But this is a very common scenario these
days, so let's simplify that one.
In particular, it makes no sense to re-moderate a blog after that change
has been made.
|
|
This means that if the same blog post shows up both under http:// and
https:// (for example when somebody changes their URL from http to
https, but it still contains the old posts), we will treat them as being
the same and not fetch a second copy of it.
We handle both http->https (common) and https->http (would probably
indicate a misconfiguration) scenarios.
|
|
|
|
Required in newer django, and still works in the old one.
|
|
|
|
|
|
|
|
MOst of what we end up keeping aren't..
|
|
|
|
|
|
When sending an email to non-ascii names, the actual email address
should not be part of the encoded section. It should have a separate
section that's not encoded.
Ported over from the pgeu install
|
|
|
|
Do this by connecting to the twitter API and actually validate both that
the account exists, and that it's not protected. Protected accounts will
cause the follower-sync script to keep trying over and over again for
following, and there's not any point in promoting those anyway.
This adds a dependency on requests_oauthlib, but that's already used in
other pgweb projects so shouldn't be a problem.
|
|
Instead, give a proper error message as intended.
|
|
If somebodys blog url changes in the feed, pick it up and store the new
one. Also send an email to the person letting them know it happened, as
well as a notice to moderators. But don't require new moderation when it
happens, just do manual post-moderation in the event that something goes
wrong.
|
|
Makes it easier to do debugging. Patch copied over from the pgeu
repository.
|
|
|
|
|
|
This will cause extra emails to be sent, which isn't very user-friendly.
|
|
Any feed that doesn't complete the socket operation in 10 seconds
will fail - we'll still retry it on the next run of course.
|
|
This avoids cases where a blog entry might be unintentionally (or intentionally)
set with a date in the future in the incoming RSS feed and as such stays at the
frontpage of planet until that date is reached.
Instead, whenever something is parsed to be in the future, set the post date to
the time of the parsing (because it clearly exists at this time).
|
|
We updatet he newest entry date in other ways as well, but this code
clearly didn't run.
|
|
Reported by David Rowley, though Alvaro
|
|
Sideeffect of some re-indenting, any short URL that had a P
in it would get %20P instead, making it not work. Oops.
|
|
Previously these were loaded from www.postgresql.org, which caused a mixed-content
warning when using the registration interface.
|
|
|
|
Also add an anchor on the edit page to deal with it
|
|
|
|
Don't encourage using the django admin view.
|
|
Blogs would display as error in list and warning on the details page,
and the other way around. Oops.
|
|
|
|
|
