[00:28:53] uuuughhhhh https://src.fedoraproject.org/rpms/mediawiki/blob/master/f/mediawiki.spec [00:29:15] what's wrong with it? [00:29:27] um, every line? :P [00:30:39] lines 85+ are especially funny [00:31:11] I think that part predates Fedora changing their policies to allow for bundled dependencies [00:37:17] though, honestly, RPM .spec looks less scary than Debian packaging :P [11:14:13] Wikimedia uses Varnish for the last several years instead of Squid, or what? I remember as of when I left, we were trying out Varnish and planning to replace Squid, but I don't know what happened in the end. [11:14:26] (It seems like the switch to Varnish did take place in the end, but I'm not sure) [11:19:07] Hmm, the response headers say yes. [11:32:22] Why is this not a super-critical problem that completely breaks caching on WMF sites? https://phabricator.wikimedia.org/T178629 [12:06:51] AryehGregor: hi yes we use varnish now [13:02:14] anomie: Note, I will not be working next week (Passover), so you'll have extra time to work through all my code submissions of this week. :) [13:43:44] paladox: FYI, another annoyance in the new Gerrit UI: typing in the textarea for the "Reply" popup is s-l-o-w for some reason. [13:44:12] hmm [13:44:58] It's fast for me [13:46:08] I typed "This is a test" and it took 4 seconds for all the letters to show up. [13:46:08] anomie see http://recordit.co/jZ6VbTJ6VE [13:46:50] i used a macbook pro 16gb of ram and 2.5ghz cpu [13:47:47] Also, oddly, Ctrl+Z for undo isn't working. I think something is trying to be a bit too smart in an event handler, resulting in stupidity instead. [13:49:04] yeh, pg had a big rewrite in 2.15+ (including some of it's ui) (it converted to es6) so you will notice performance improvements as the reply dialog was redesgned and again in master. [13:49:12] Both work fine for inline replies, it's just the popup to actually submit the replies. Are there other such popups that I can test it in? [13:49:33] anomie not sure what you mean by test it in? [13:50:20] and nope no other popups [13:50:30] except the reply dialog to submit the replys [13:54:33] anomie though i am aware pg needs to fix keyboard actions too. it lacks some that are in gwtui. [13:56:19] anomie if you type ? on your keyboard it will show the help screen [13:56:26] for keyboard actions (in pg) [14:14:18] * anomie keeps reading paladox's "pg" as "PostgreSQL" rather than "PolyGerrit" [14:14:27] heh [14:14:46] polygerrit is more secure then gwtui [14:14:53] as it protects against xss attacks [14:15:17] it uses https://github.com/Polymer/polymer-resin [14:16:47] thats in polygerrit 2.15+ though i think [14:26:49] How do I get 'git review' to stop trying to submit a ton of commits that I didn't write and just submit the head commit? [14:28:00] AryehGregor: git fetch --all [14:28:41] Oh, because I fetched origin but not gerrit and so it thinks all the commits from origin are really my changes and it doesn't realize that gerrit is the same as origin anyway? [14:28:50] AryehGregor: git-review is a bit stupid. it tried to submit everythign that isn't on the target branch (master). [14:28:53] DanielK_WMDE_: Worked, thanks! [14:29:13] AryehGregor: yes, exactly. you can actually sumit all of them, the submits are redundant and won't do anything. it will just take forever. [14:29:50] AryehGregor: also... if there are non-ascii characters in the commit messages, trying to list them will make git-review die. because it sucks. [14:30:04] so if you get fatal errors related to character encoding - just fetch --all :) [14:53:39] anomie: Is there something somewhere that ensures that configuration variables are set to standard values before running tests? [14:57:23] AryehGregor: Sort of. tests/common/TestSetup.php sets various things, but if something doesn't use $this->setMwGlobals() or the like when changing a config variable the change will persist for later tests. [14:58:52] And, of course, TestSetup doesn't set *all* the configuration settings. Most tests tend to assume the defaults in DefaultSettings.php are in effect for anything not changed by TestStep or required for basic operation. [14:59:12] Thanks. [15:42:20] AryehGregor: most of our "unit tests" aren't. many of them fail if the wiki doesn't have english as its content language. or doesn't have wikitext in the main namespace [15:43:23] DanielK_WMDE_: It's not really possible to write unit tests for something not written in independent units. :) [15:43:38] Nothing wrong with integration tests, although they might be misnamed . . . [15:45:32] stuff that relies on local settings isn't even an integration test. it's a system test. but anyway. yea. modularity, dependency injection, and all that [15:45:36] yay global variables. [17:15:08] Can I get PolyGerrit to remember that I want unified diffs instead of side-by-side? [17:51:00] anomie: How do I test that a warning was logged via LoggerFactory? Where does the warning go? [17:51:07] To a log file? [17:53:23] AryehGregor: I don't know. bd808 might. If you can inject a Logger instead of the code using LoggerFactory, you could inject a TestLogger (tests/phpunit/includes/TestLogger.php). [18:01:27] anomie: Any suggestions for why ApiMain::execute() would return null in a test? I'm mocking the WebRequest I pass to it instead of passing a FauxRequest, if it helps. [18:02:43] AryehGregor: Doesn't ApiMain::execute() always return null? Or, technically, void, but in PHP that comes out the same. [18:02:55] Oh, right. [18:02:59] I just got confused. :) [18:03:42] ok (: [18:17:20] anomie: Do you know why large chunks of the file might be appearing as non-executable and therefore not getting coverage-tested, locally? On the WMF coverage report they show up as executable. E.g., the whole method handleException. [18:18:56] AryehGregor: I don't see any reason at first glance why ApiMain::handleException would show up as non-executable. [18:24:22] I tried php 5.6.30 with xdebug, and php 7.0.28's and 7.2.3's phpdbg, all showed it as executable. [18:34:00] anomie: Any ideas on ? AFAICT, TranslateSandbox's userid param has 0 as a default, not '0'. [18:34:24] AryehGregor: I already left a comment ;) [18:35:10] Oh, thanks. [21:22:03] unless anybody objects to https://gerrit.wikimedia.org/r/#/c/421193/ I'm gonna merge it [23:22:17] AryehGregor: nowhere by default, you can send it to a a file if you add the channel to wgDebugLogGroups