[17:22:19] tgr: got a while to review https://gerrit.wikimedia.org/r/#/c/306616/ ? it's the first for the login form regression you reported [17:22:52] in a sec [17:22:56] i'd like to deploy it during the SWAT in 40 minutes [19:03:24] tgr: i'm trying to test some stuff on the login page, and i can't figure out how to make $wgSecureLogin=true actually work, without MediaWiki redirecting me to HTTPS immediately. [19:03:41] $wgServer = '//localhost:3080'; [19:03:41] $wgCanonicalServer = 'http://localhost:3080'; [19:03:41] $wgDefaultUserOptions['prefershttps'] = false; [19:03:41] $wgSecureLogin = true; [19:04:10] when i try visiting the wiki, i get redirected to HTTPS. nothing i do seems to prevent that, and i can't even figure out where in the code the redirect happens. [19:04:11] isn't redirection is the point of that setting? [19:04:23] or you mean redirection elsewhere than the login page? [19:05:03] isn't it just supposed to show a link to HTTPS when visiting the login page using HTTP? [19:05:32] AFAIK it means redirect to HTTPS for login and then redirect back if needed [19:06:36] you can use the CanIPUseHTTPS hook to override it for a given IP [19:10:04] tgr: i was trying to achieve this: https://i.imgur.com/dCj7ldZ.png [19:10:29] i got it now. you have to visit Special:userLogin?fromhttp=1… alright [19:12:41] well, you can disable wgSecureLogin, then you will see it [19:16:38] …huh [19:18:00] i clearly misread something horribly. thanks [19:19:30] $wgSecureLogin was a terrible kludge from when we half-supported https. It should die in the firey hell I brought it out of. [19:19:45] Either you have https, or you don't. Half-https is half-assed. [19:21:53] ostriches: i only care about the interface it generates, because i'm changing it and don't want to break stuff :) [19:44:02] ostriches: the forcehttps cookie should die with it :) [19:44:25] Yes. [23:18:24] * TimStarling is examining the differences between parserTests.php and NewParserTest with a view to merging them [23:21:57] currently musing over why phpunit is so much slower [23:22:37] possibly it has something to do with the fact that to run the parser tests, phpunit does 188043 select queries, whereas parserTests.php does 4559 [23:23:15] I think phpunit is more agressive in resetting the database and caches on every test run [23:23:32] when I use parserTests.php, my main page will be screwed up until I null edit it [23:25:55] made a note of that [23:27:05] MediaWikiTestCase is atrocious [23:27:44] I managed to cut away some of the work in I17ef1f5197 / Iaba1c8e1f etc [23:27:53] right, so parserTests.php saves into the host wiki's parser cache [23:28:16] but it's still atrocious, and not easy to fix, because you have to be wary of breaking extensions [23:28:56] a few memcached queries are properly prefixed but most are not [23:29:40] phpunit just uses a HashBagOStuff [23:46:12] so moving publishTestArticles() from per-test to global setup makes it a factor of 4 faster at the cost of a single parser test failure