[17:06:15] Does LuaSandbox consider various function calls cheaper than the standalone version? [17:57:12] Can someone review https://gerrit.wikimedia.org/r/#/c/324352/ so we can remove the FSRepo class from core? :) [18:06:27] "Haven't tested and not sure if this will work but +1." -- ugh [18:08:15] "Do you know even know how to test it?" [18:08:43] heh [18:08:48] left one really annoying comment for you [18:08:53] It's pretty fair [18:09:01] Cause it's a lie [18:09:02] I don't know how to test it :/ [18:10:42] I don't know if https://gerrit.wikimedia.org/r/#/c/324350/ helps or hinders [18:10:52] FSRepo was pretty much just a wrapper class... [18:12:11] deprecated in 1.19 seesm worth removing [18:12:17] less cruft is less cruft [18:12:56] there was only 2 usages of it in any of our extensions [18:14:12] Reedy: so if I blindly merge those 2 patches will you think less of me as an engineer? ;) [18:16:17] Reedy: Re LuaSandbox vs LuaStandalone, there's a method Scribunto_LuaSandboxEngine->getPerformanceCharacteristics() that returns data indicating whether calls into PHP require serialization. The only thing that uses it at the moment is Scribunto_LuaUstringLibrary, which uses it to decide whether to use PHP or pure-Lua implementations of various string functions. Why do you ask? [18:18:15] Reedy: https://gerrit.wikimedia.org/r/#/c/324350/ needs a manual rebase (probably RELEASE-NOTES) [19:01:16] anomie: Was just wondering if it was worth "upgrading" to using luasandbox for an install I was doing some work for someone [19:01:27] But they're mostly doing Wikibase stuff... [19:01:39] bd808: It's always release notes :P [19:03:15] Reedy: In general luasandbox will probably be faster since it doesn't have to do IPC for every invoke and every call back into PHP. Whether it's enough faster to make it worth the effort, I don't know. [19:03:32] Fair enough [19:03:46] I'll bare it in mind if they encounter other problems... [19:04:01] Mostly bumping the expensive function counts seem to work... And things aren't that slow anyway [19:07:26] Thanks anomie [19:12:21] bd808: rebase [19:12:22] d [21:30:56] legoktm: (or anyone else familiar with centalauth and Special:GlobalUserRights), guidance please: https://phabricator.wikimedia.org/T153578#2894725 [21:31:07] * greg-g resists pinging a lot of individuals for now :) [22:18:22] bd808: vagrant install works when forcing shallow cloning [22:18:31] is that useful or will it break git-review? [22:19:30] tgr: as I recall the shallow clone will be replaced with a deep clone the first time you try to actually use git-review [22:19:43] and then you get stuck waiting for that to happen [22:25:01] ugh, new gerrit does not offer you a Change-Id in the error message any more [22:33:54] tgr: I'm kind of unsure how hard we should try to hack around gerrit bugs in mw-vagrant vs trying to help figure out how to fix the origin repo itself [23:08:35] bd808: I'd agree, except I have no clue how to fix it [23:08:51] fwiw https://gerrit.wikimedia.org/r/#/c/328591/ was pushed instantly, and my repo is still shallow [23:09:28] shallow cloning might speed up the install process as well [23:10:11] I don't have a full baseline but vagrant up took 1500s with a shallow clone and 1200s with a deep one (which of course stopped after the git clone error)