Wed, Nov 15
Updated the dependency tree with these.
Tue, Nov 14
anchorencode fix at https://gerrit.wikimedia.org/r/#/c/391141/
As soon as we fix all the problems discovered on Russian projects.
Sun, Nov 12
The problem with unsafe dependencies is usually not someone introducing a dependency on something broken - it's holes being found in existing dependencies. That would just result in unrelated commits breaking.
Fri, Nov 10
Thu, Nov 9
It appears in planet_osm_line but not planet_osm_polygon. @Pnorman, any ideas why?
So, basically, what's going on is:
- Something needs user language
- User information is loaded to get their language
- Because cookie blocks are enabled, User::loadFromSession() retrieves block information
- When there is a block not caused by something that an admin leaves a block message for (in this particular case, XFF but there are other similar cases around in User::getBlockedStatus()), we initialize the block message with an interface message:
$block->mReason = wfMessage( 'xffblockreason', $block->mReason )->text();
- Message needs to know user language
- Infinite loop!
Wed, Nov 8
Tue, Nov 7
Mon, Nov 6
Напишите об этом на странице обсуждения статьи, пожалуйста. Phabricator только для технических проблем.
Sun, Nov 5
gt_name is a name of particular geotag. It might be the same as article name, or it may not - because pages can contain multiple geotags (though only one of them can be primary). To get page title, you need to match gt_page_id with page_id from the table page (so use hewiki-latest-page.sql.gz).
Fri, Nov 3
I have a silly question: why do we even use reference names in IDs? Because these IDs are also using reference numbers, they're not stable anyway.
Thu, Nov 2
Turns out, I inadvertently fixed this in https://gerrit.wikimedia.org/r/#/c/362326/ not even knowing about this bug. This change is present in 1.30. Do we need to backport to older versions?
+1 to cscott. We shouldn't be enabling this.
This appears to be the PHP behavior: a simple <?php header( 'Location: http://localhost/#абвг' ); redirects people to #%D0%B0%D0%B1%D0%B2%D0%B3.
1.28 is out of support now.
1.28 is out of support now.
Try $this->command as a more readable alternative to $cmd?
Submitted the stuff reported here as T179520: Users are redirected to percent-encoded section fragments after section editing and T179544: Some characters in HTML5 fragments mess with wikitext.
Wed, Nov 1
He says he's getting failed notices, but successful seems like it would make more sense...
We should probably limit the characters allowed for reference names explicitly in Cite.
Tue, Oct 31
Russian projects are done, I think we can tentatively deploy everywhere next week unless new serious problems surface. Do we need any extra outreach for that?
Aaaaaaannnndd it's done!
This is a general restriction on page length in MediaWiki, and I don't think that raising this already quite generous limit would be a good thing. We aren't just storing this data, we are serving it to users in maps, and waiting for a huge chunk of data to be downloaded isn't nice. There are other ways to reduce the size of data, like limiting precision and simplifying geometries.
Mon, Oct 30
Today was all packed, but I grabbed a window tomorrow.
It's normal for that repo, it'll just need a rebase before merging.
Thu, Oct 26
Notifications aren't sent as strings. They're built from parameters by various PresentationModel implementations when users see them. As long as you don't actually slap the IP itself somewhere it's fine.
I don't think this will work:
- GeoIP databases aren't as accurate as people think.
- They are even less accurate outside the US
- Relying on GeoIP based locations can create serious problems
- Maxmind databases are in English only. We don't have any means to translate them.
Wed, Oct 25
Should be deployed with the next periodical portals update.
Tue, Oct 24
Mon, Oct 23
Why are you asking me? :P Yes, of course the abuse of FauxRequest should be prohibited with fire and brimstone, however practically speaking there are huge obstacles around this. MediaWiki should be an example for extensions of how to share a common model between API and web interface, yet we have that for how many core APIs, zero? Furthermore, the API actively encourages its modules to do things directly, for example ApiQueryBase has functionality to build direct DB queries, instead of maybe some code to integrate its pagesets with models outside of the API. Until this gets addressed, people will continue sticking FauxRequest where they shouldn't.
Fri, Oct 20
Wed, Oct 18
After a week, nothing in logs - removing.
Oct 18 2017
Yes, I will totally be there.
MW isn't going to support HHVM in the long term.
Oct 17 2017
The fix has been deployed, transcodes work now.