[10:44:59] greg-g: Thanks! But it didn't help. :( [14:17:02] How is it possible that the line in ApiMain::executeAction immediately before a call to $this->checkMaxLag() is reached, but the first line of ApiMain::checkMaxLag is not reached? [14:17:11] I don't see anything that overrides it . . . [14:17:25] Oh, never mind me. [14:38:01] anomie: Trying to pull anything from gerrit gives me "fatal: internal server error\nremote: internal server error\nfatal: protocol error: bad pack header", which stops me from checking out extensions among other things. Do you know what might be wrong? [14:39:02] AryehGregor: try git remote prune origin [14:39:04] Please [14:39:06] AryehGregor: No. But I encountered that a few minutes ago and the `git remote prune origin` mentioned earlier fixed it. If `git remote` lists other remotes besides "origin", you might try pruning them too. [14:39:24] paladox: I did, didn't work. [14:39:25] * AryehGregor tries again [14:39:46] Oh [14:39:52] Aha, I have a separate remote called "gerrit". [14:39:56] Which repo is this please [14:40:04] You could also do git config remote.origin.prune true [14:40:16] Pruning that works. [14:40:55] Huh, I have a remote "origin" and a remote "gerrit" and they have the same url. [14:44:27] AryehGregor: That's a default for git-review. If you want to just use "origin" for everything, https://www.mediawiki.org/wiki/Gerrit/Tutorial#Configuring_git-review (and then `git remote remove gerrit`, I think) [14:50:05] Now I get: $ git clone ssh://simetrical@gerrit.wikimedia.org/r/p/mediawiki/extensions.git Cloning into 'extensions'... ssh_exchange_identification: Connection closed by remote host fatal: Could not read from remote repository. Please make sure you have the correct access rights and the repository exists. [15:20:33] AryehGregor: is that ssh? [15:20:36] Or https? [15:21:26] paladox: From your question, I fixed the problem. Thanks. [15:21:38] Having a non-default port and different paths for HTTPS vs. SSH is somewhat confusing. [15:21:56] Oh [15:21:57] Why a non-default port, just to avoid automated portscanning? [15:22:39] Because gerrit uses apache sshd [15:22:49] This is to prevent it conflicting with port 22 [15:24:41] Why do the extension submodules register HTTPS paths instead of SSH? [15:24:44] Presumably I want SSH? [15:25:53] If the https urls are used in .gitmodules then it will use https :) [15:27:16] How am I supposed to push updates? Change each one that I want to push to manually to SSH? [15:29:22] Change each sub module? [15:29:48] You should make changes in each of those repos instead of doing it through the submodule [15:30:14] Ie something like mediawiki/extensions stores all the extensions as submodules [15:30:39] You should clone the repo you want to make a change in and then push it [15:31:05] Gerrit is smart so it will update the submodule in mediawiki/extensions without you having to do anything [15:31:15] (As long as the change is merged) [15:41:29] AryehGregor: ^^ [15:42:07] paladox: Yes, I think I sort of know how git submodules work, although I don't think I ever actually pushed changes to a submodule before. [15:42:35] Yep [15:54:44] AryehGregor: i think you can configure Git to magically remap the URLs when pushing/pulling, like this: https://stackoverflow.com/questions/1722807/how-to-convert-git-urls-to-http-urls (but note that i never tried to use this myself) [17:11:47] 16:51:44 1) ApiBlockTest::testBlockWithExpiry [17:11:47] 16:51:44 Failed asserting that 1 is less than 1. [17:12:45] AryehGregor: oh wonderful, you already have a patch to fix it :D [17:40:23] legoktm: Can I convince you re. https://gerrit.wikimedia.org/r/#/c/416721/ ? ;-) [17:40:42] You must by spying on me [17:40:48] be* [17:40:49] Ha, timing! [17:40:57] I spent the past 20 minutes reviewing that :p [17:41:13] legoktm: You C+2'ed the rest of the stack previously, BTW. [17:41:31] But I had to rebase due to, well, release notes. [17:41:39] Yeah, I'll wait for that one to go through, I think it might need rebasing again [17:42:06] Hahaha, I remember all the merge conflicts in release notes. So sad that there's no solution to it. [17:42:38] AryehGregor: There's a solution (custom merge protocol), apparently, but it's scary and gerrit is… not the friendliest. [17:44:06] https://github.com/MatmaRex/mediawikireleasenotes-driver [17:46:10] it's not *that* scary. [17:48:42] Tell no_justification not me. ;-) [17:49:13] Huh? [17:51:13] no_justification: Talking about the endless rebase conflicts with RELEASE-NOTES files, and installing MatmaRex's custom merge driver to avoid them. [17:56:41] James_F: anything else you want review on? :) [17:57:41] legoktm: https://gerrit.wikimedia.org/r/#/c/415477/ which you want to do a different one alongside. ;-) [17:57:57] legoktm: Oh, and https://gerrit.wikimedia.org/r/#/c/414838/ ;-) [18:00:12] How would we call this? I'd rather us do it in a gerrit plugin [18:01:20] I don't remember. It's been so long [18:02:40] anomie: Why does ApiMain::checkConditionalRequestHeaders ignore empty ETags? It seems like those are valid per spec? [18:02:49] Specifically the array_diff near the top. [18:08:10] anomie: oh, have you seen https://phabricator.wikimedia.org/T190539 ? re: your postgres patches [18:09:26] AryehGregor: An empty ETag would be represented as "" or W/"" in If-None-Match, wouldn't it? [18:10:11] anomie: Ah, you're saying that just filters out '' and not '""'. [18:13:24] legoktm: Personally, I'd be happy with dropping support for such old PG versions. [18:14:21] 9.1 was released in 2011; 8.3 was released in 2008 and has been unsupported upstream since 2013. [18:17:25] anomie: 9.2 was approved in https://phabricator.wikimedia.org/T164898 so I think someone just needs to do it? [18:22:43] legoktm: OTOH, now 9.3 is the lowest version currently supported upstream. [18:29:53] legoktm, anomie: I have again been thinking about moving support PG, MSSql and Oracle into extensions. Shouldn't be too hard, should it? [18:31:03] AS LONG AS MSSQL GETS THROWN AWAY AFTER SPLITTING OFF!!1 [18:31:05] DanielK_WMDE_: IMO we should add PG to CI so it gets tested, rather than removing it. It would be helpful to catch where people make MySQL-specific assumptions. [18:31:33] MaxSem: :D [18:32:27] anomie: sure, why not. doesn't have to be in core for that. [18:32:34] Also, as things stand now splitting off is likely to mean killing it, because someone would have to duplicate all the schema change logic in the extension. Something I'd like to do someday is abstract the schema changes so the extensions would instead just have to implement generic operations like "add field", "drop index", and so on. [18:33:33] But mainly I want to get rid of the need for WMF to support MSSQL and Oracle. It sucks that we can't make schema changes to core without testing against proprietary databases. we either merge untested code, or require developers to get a license. [18:34:10] I personally just ignore them [18:34:20] anomie won't let me do that [18:34:20] * subbu is waiting for a gap in the discussion before throwing in a new qn. [18:34:21] I've been using sqlfiddle.com a lot lately. Usually that at least lets me see if a schema change has the right syntax. [18:35:39] * subbu takes the opportunity ... [18:35:40] anyone know of a reason why we couldn't / shouldn't deprecate $wgRawHtml? [18:36:20] is FR still using it? [18:36:28] FR? [18:36:43] subbu: sure, lots of in-house wikis use it to do Nasty Things (tm) - embedded forms, videos, all kinds of stuff [18:36:45] otherwise, i'd say gogogogo for it [18:37:00] subbu: fundraising [18:37:04] ah. [18:37:14] I know WMDE is using it on the internal wiki [18:37:40] raw HTML is the lazy solution where you don't want to write an extension [18:37:49] indeed. [18:38:54] I have an alternative proposal: an HTML content model, plus an HTML namespace with privileged access, and HTML based template transclusion. [18:39:14] the last bit needs a bit of work [18:39:16] parsoid has had no support for $wgRawHtml and arlo added a patch to fix crashers on foundationwiki which is because of that wiki raw html (and in a broken unbalancd way) .. but, that is going away .. so, wondering if there is a path we can get away without supportin git. [18:39:17] the rest is trivial [18:40:10] James_F mentioned that foundationwiki's use case is going away soon .. so, that is why this deprecation thought popped up. [18:40:11] subbu: {{#html:foobar}}, where foobar is in the HTML namespace, with the HTML content model. [18:40:28] do you think that could work? [18:41:02] would be even nicer if plain {{HTML:Foobar}} would work, but thatn needs more work in the parser/preprocessor [18:43:12] maybe ... since it fits with the existing extension / parserfn / transclusion model .. and we could then force that html to use balancing types like anything else. [18:43:32] so, breakage in that html doesn't screw up rest of the page. [18:43:47] that's the idea, yea [18:44:07] ok .. so, tldr seems to be that $wgRawHtml deprecation is not a quick thing. [18:52:22] anomie & legoktm: https://gerrit.wikimedia.org/r/421972 [18:52:37] anomie: that's on my roadmap too (abstracting schema changes) [18:52:47] it'd go a long way into improving dbms support [18:53:00] Plain {{HTML:Foobar}} could probably kind-of work, by having HtmlContent return "{{safesubst:#html:Foobar}}" from getWikitextForTransclusion(). [18:53:58] should make an RFC for it probably [18:54:16] rather than just submitting a patch which changes like everything :P [18:54:22] Skizzerz: Since you said that, I'm motivated to start writing down all the thoughts about that that have been floating around my head. [18:54:48] go for it! and send me a link and I'll fill it in with some of my thoughts on the matter too [18:55:59] It'll be a few days at least before I get it all written. [18:56:09] np, I'm super busy right now anyway [18:56:24] and I think this probably should go through the full RFC process, unless you don't think we need that [18:56:50] Definitely the full RFC process. [18:57:03] hmm, 1.32 is a decent target for it too, since it's the version after an LTS, so we have lots of time to stabilize before next LTS [18:57:45] ALL versions should be stable [18:57:54] yes [18:58:27] LTS is not "more stable", it's "we provide extended support for it so that people don't have to migrate every half year" [18:58:32] but a sweeping change like this is likely to have sneaky bugs, especially with respect to extension schema changes [18:59:02] hopefully the RFC process shakes all of those corner cases out though, and unit tests should be able to cover them as well [18:59:39] but I'm sure we won't get iteration 1 completely perfect even with all of that (we can certainly try, but reality has this nasty habit of spiting you) [19:00:32] by targeting it for version after LTS, we reduce the risk of having to backport critical bugfixes into the LTS a year or so down the road, when a normal release would already be out of support [19:02:01] hope that makes sense :) [19:13:54] We have stats for this now [19:13:58] Nobody uses Oracle or MSSQL [19:14:33] *basically nobody [19:22:22] the stats are opt-in though, you may be underrepresenting them by like a whole 2 installs! [19:22:24] :P [19:23:19] not surprised though, they're broken more often than not. The only contact I've had with people who use mssql are because it's a mandate on-high that they have to be 100% microsoft [19:24:51] I'm surprised they're not using the sharepoint wiki [19:26:16] idk. I usually gave advice that they should work it out with management to switch to mysql since it's far better supported [19:38:34] Skizzerz: but PHP isn't MS! [19:44:58] no_justification: Re the new Gerrit UI, whenever I try to highlight something it jumps all over the place unless I finish really quick. :( [19:50:32] https://www.mediawiki.org/wiki/Requests_for_comment/Moving_database_abstractions_out_of_MediaWiki_core [19:54:38] legoktm: what we were discussing above was refactoring schema updates to no longer need .sql scripts, which would in theory improve support for the various other dbmses. Should be less effort than the one you linked, I think [19:55:54] might be able to nix tables.sql too [19:56:14] and ship a maintenance script to generate the sql that you can manually run should your workflow require that [19:57:03] can we just use doctrine or something similar? [19:57:10] http://www.doctrine-project.org/ [19:57:35] legoktm: I'm actually opposed to that RfC nowadays. I've slowly come around to the idea that we should drop all DB backends other than Mysql/Maria & Sqlite (the latter only for testing/dev purposes and the fact that differences are minimal) [19:58:29] I've seen a number of projects go the direction of "we support postgres. if you don't like postgres tough find new software" [19:58:46] Phab does same with Mysql/Maria [20:00:15] what are the stats like for people using mediawiki + postgres? [20:00:20] Too many :( [20:00:27] We'd have to announce an EOL for it [20:14:10] MaxSem: +2'd [20:14:18] wee [20:18:43] no_justification: at the very least, we should have a way to migrate a DB off PG if you want to kill it [20:22:05] Sure. [20:22:15] I'd support that. And Oracle/Mssql too. [20:23:28] yeah, same [21:03:34] anomie do you mean jump around like https://bugs.chromium.org/p/gerrit/issues/detail?id=8596 ? [21:04:46] paladox: Considering my complaint is about trying to highlight something to use the "C" key, not about during the page loading, no. [21:04:57] ah ok [21:05:04] they have done alot of work since 2.14 [21:05:11] mostly with the keyboard too [21:05:59] If I had to guess, I'd guess that something about it trying to place little "Press C to comment" popup is making it jump like crazy. [21:12:44] anomie you could open the diff like this: [21:12:50] 1s generating the photo :) [21:13:21] anomie https://phabricator.wikimedia.org/F16147298 [21:13:25] click expand all diffs [21:13:37] or show diffs [21:14:42] paladox: What does that have to do with my issue? [21:14:57] It could work around it [21:15:34] though i carn't seem to reproduce your problem. [21:17:31] paladox: It doesn't work around it. [21:17:37] ok [21:19:02] anomie is it something like this http://recordit.co/uEh1U4k2CT (so that i understand how your doing it) [21:19:29] you can report the issue here https://phabricator.wikimedia.org/project/view/330/ and we can try to forward it upstream [21:24:15] paladox: https://phabricator.wikimedia.org/F16148257 [21:25:12] anomie thanks, i wonder how do i play mkv files on a mac? [21:29:26] anomie ah i managed to convert it to mpv4 [21:29:29] i see your problem now [21:29:35] your using firefox [21:29:51] lol [21:29:54] :( [21:30:09] is gerrit going to get into the "only works with chrome" business too? [21:30:16] Nope [21:30:24] they are fixing bugs to do with firefox too [21:30:30] apparently firefox works differently [21:31:33] https://gerrit-review.googlesource.com/q/firefox [21:32:49] We should add firefox to there ci, but i need to figure out how to get headless to work [21:35:10] hmm though i carn't reproduce the issue in firefox on the mac [21:35:14] with firefox 59 [21:40:56] they are also fixing bugs for safari and ie.