[00:05:38] csteipp: Do you want/need to do a security review of the updated scholarships application? It has several updates since last year, but all the updated libs are the same versions that the IEG review app used. [00:07:28] csteipp: The only thing that has been added that you might want to really look at is the csrf token refresh code we added. -- https://gerrit.wikimedia.org/r/#/c/179471/ [00:14:19] 3MediaWiki-extensions-CentralAuth, MediaWiki-Core-Team, SUL-Finalization: Investigate and fix OOMs caused during account globalization - https://phabricator.wikimedia.org/T78727#961277 (10Legoktm) a:5Legoktm>3None [16:02:31] I talked with _joe_ about that earlier [16:02:40] the pending HHVM updat... [00:46:04] Fatal error: unexpected N4HPHP13DataBlockFullE: Attempted to emit 23 byte(s) into a 35651584 byte DataBlock with 1 bytes available. This almost certainly means the TC is full. If this is the case, increasing Eval.JitASize, Eval.JitAColdSize, Eval.JitAFrozenSize and Eval.JitGlobalDataSize in the configuration file when running this script or application should fix this problem. [00:46:19] ori, _joe_ ^^^ [00:46:32] (in prod) [01:25:34] bd808: I'm not too worried about it, but I'll look at the csrf patch when I get some free time. [01:26:37] csteipp: cool. I think we are going to re-launch early next week with the application period opening on 2015-01-19 [01:26:52] assuming things go as they should any way :) [01:29:46] (he just left) [01:34:39] 3MediaWiki-Core-Team, Services: Authn and authz as a service - https://phabricator.wikimedia.org/T84962#961654 (10GWicke) [01:53:43] <^demon|away> Aww, Brad's not here :( [02:46:27] TimStarling: has the agenda for the RFC meeting next week been set? If not, I think Jon Robson is interested in putting on the agenda. The mobile team is planning to implement it, and in fact he has an initial patch up at [02:47:54] (bbiab) [03:07:14] there is a draft agenda for next week [03:07:31] we could possible drop SOA Auth and have Jon's one instead [03:07:39] possibly [03:31:10] TimStarling: thanks! [03:31:56] no problem, thanks for the nice work [03:37:03] I'm wondering about how we should do the migration. Instead of bundling a shim-type thing in every extension like we did for i18n/json, I was thinking there could be an extension that has basically the same code for older core versions so people could still use wfLoadExtension(). Or we could just continue to maintain the PHP entry points for a while and duplicate all the info... [03:40:35] 3MediaWiki-Core-Team, MediaWiki-General-or-Unknown: Drop PHP 5.3 support - https://phabricator.wikimedia.org/T75901#784809 (10MZMcBride) The triplicate posts made me smile. <3 [03:50:32] <^demon|away> TimStarling: That's probably a good idea. We (Chris, Gabriel, myself, others) haven't met to likely revise the RfC yet. [03:50:45] <^demon|away> So swapping it out for another week is best methinks. [03:53:14] cool, that's a happy coincidence then. [03:59:38] Wasn't user-specific page lists about adding tags? [04:00:47] <^d> ori: Do you have any insight on https://phabricator.wikimedia.org/T85565? I think I might be running into an issue with hhvm and pcntl_* functions not working as expected. [04:02:14] <^d> here's a slightly more complete output from my command line, actually: https://phabricator.wikimedia.org/P197 [04:03:17] 3MediaWiki-Core-Team, CirrusSearch: Forking while reindexing fails on HHVM - https://phabricator.wikimedia.org/T85565#961740 (10Chad) [04:03:38] ^d: I'm just about to put Noam to bed, but I'd be happy to look in a little bit [04:04:15] <^d> I should still be about, ping when you're available :) [04:32:41] <^d> robla: Which wiki is this? I'm having trouble figuring out the lang code and the e-mail lacks any links. [04:33:11] ^d: I think it's not wiki specific; probably commons. [04:33:23] <^d> Blagh, that's more annoying. [04:33:25] <^d> Gotcha [04:33:34] the reason why WP0 is in the mix is because there are specific IP ranges associated with it. [04:38:28] <^d> We already have a file_sha1 in AbuseFilter [04:38:33] <^d> Adding a file_size shouldn't be too hard [04:38:57] I have a patch, just haven't had the chance to test it yet. [04:39:56] <^d> Ah, I was going to say I could have one pretty shortly. [04:40:54] <^d> It's like a 3 line patch :p [04:41:36] there's i18n stuff to add as well [04:41:45] <^d> I was counting that :) [04:41:52] 4 lines :P [04:42:06] <^d> I didn't add qqq because sha1 didn't have qqq :) [04:42:15] trying to break osmium again [04:44:17] when all local ports are exhausted, connect() fails with EADDRNOTAVAIL [04:46:37] I got confused for a second and thought the file_sha1 field mentioned was a database column name. Wouldn't that be nice... [04:51:06] and that does generate a "CONNECTION FAILURE" error message due to the default case being hit in the error handling switch [04:51:11] <^d> legoktm: While we're here...do we have a file_mime and file_ext on hand we could use? [04:52:51] ^d: I don't see a public accessor, UploadBase has a protected $this->mFinalExtension [04:53:10] <^d> Meh, if there's no accessor don't worry about it. [04:53:23] <^d> size + ip range should do it anyway [04:54:55] if they're logged in AF can't filter on IP range [04:56:02] and then if you use the same Memcached object again, it will give "SERVER HAS FAILED AND IS DISABLED UNTIL TIMED RETRY" [04:56:23] I think the reason I failed to reproduce it yesterday was because I had strace running [04:56:51] no, i just tried reproducing it a little while ago on mw1225 (which happens to be repooled) and couldn't [04:57:11] ^d: https://gerrit.wikimedia.org/r/183430 [04:57:32] well, I reproduced it on osmium [04:57:42] I'll reproduce it on mw1225 as well if you like [04:57:58] <^d> legoktm: Considering you wrote the exact same patch as me, +2 and merged :) [04:58:03] yeah, could you? i'm curious now [04:58:27] it has to do at least 28232 connections in a minute [04:58:31] :P [04:58:38] I'll put it in tomorrow morning's SWAT [04:58:41] if the server is slow, then it could drop to less than that rate [04:59:15] <^d> legoktm: sounds good to me [04:59:41] <^d> cc robla: patch merged for file_size to master, will swat in AM. [04:59:41] on a probably-unrelated note: it's strange that this small example results in a CONNECTION FAILURE: addServer('localhost', 11211); $c->get('foo'); echo $c->getResultMessage() . "\n"; [05:00:02] getResultMessage gets the status of the last operation, so calling it after a get is not crazy [05:00:51] if you use "localhost" as the hostname then it alternates between 127.0.0.1 and ::1 [05:01:04] right, i noticed the tcp6 connections in netstat [05:01:21] which will also dilute the time_wait contention since it is per-IP [05:01:42] the MW configuration uses 127.0.0.1 so that is a realistic scenario [05:02:54] tstarling@mw1225:~$ telnet localhost 11211 [05:02:54] Trying ::1... [05:02:54] Trying 127.0.0.1... [05:02:54] telnet: Unable to connect to remote host: Connection refused [05:03:01] yeah there is also that [05:03:30] let's just pretend that didn't happen [05:03:38] everyone clear your buffer [05:04:03] it's 11212 [05:04:37] right... [05:04:44] /clear [05:05:46] so, what do we do about it? because the load is only 30% percent higher (15 vs. 20 pybal weights), setting the address to localhost would probably resolve it, because it would distribute connections over 127.0.0.1 and ::1, but that seems like a gross solution [05:06:05] the connection reaping sysctl param seems like a good solution, but i don't know what side effects that would have [05:07:18] we could distribute across multiple endpoints in less silly ways [05:07:21] unix domain sockets are attractive too and nominally faster, but you lose some debugability (is there an easy way to tap into data flowing in and out of a unix domain socket a la tcpdump/wireshark)? [05:07:33] like using a port range, or an address range in 127.0.0.1/24 [05:07:52] nutcracker can listen on as many sockets as you like, right? [05:08:45] i don't think so [05:09:26] we could just not use nutcracker, that would distribute the connections across the backends [05:10:15] domain sockets seem cleanest [05:10:33] reproduced on mw1225 [05:10:44] in fact, i'm pretty sure ryan lane ended up doing that at his new gig, and i don't think he ran into any problems [05:10:57] cool [05:11:30] I monitored TIME_WAIT count in /proc/net/sockstat and confirmed that it stopped when it hit ~28000 [05:11:58] domain sockets? [05:12:37] having nutcracker listen on a UNIX domain socket [05:15:14] <^d> legoktm: I didn't notice your wmf cherry-picks and made PS2 for both :p [05:15:54] I think the socket domains are unix, TCP, IPX, etc. [05:16:06] i.e. the unix domain is a particular domain that sockets can be in [05:16:20] hence my confusion at "domain socket" [05:17:18] sorry, i was being imprecise [05:19:56] for debugging, there is always strace [05:20:31] yeah, true [05:22:24] <^d> ori: I'm calling it a night actually, I'll pick your brain about forking tomorrow :) [05:22:37] ^d: cool. i haven't forgotten, i'll look at it tonight probably. [05:22:57] <^d> Sweet. G'night all [05:22:57] _joe_'s hiera work makes it easy to override a setting for a subset of the application servers. we could do just a couple of app servers for starters. [05:23:33] ok [05:24:26] want to write a patch or should i? i know where to look so i can do it pretty quickly [05:24:54] sure, you can do it [05:32:47] I am just finishing reading the relevant kernel code [05:33:34] there are definitely no statistics or logs that would tell you when this happens, not in the kernel anyway [05:35:02] just monitoring the "tw" field in /proc/net/sockstat is not good enough either, since you don't know how many addresses are represented [05:35:54] so it is really up to the client to check for EADDRNOTAVAIL and log a critical error message [07:41:10] 3ContentTranslation-Deployments, MediaWiki-extensions-ContentTranslation, MediaWiki-Core-Team: Content Translation Beta Feature security review - https://phabricator.wikimedia.org/T85686#962007 (10Nikerabbit) >>! In T85686#960146, @csteipp wrote: > Unfortunately, there also should be a requirement there that jav... [08:59:07] 3MediaWiki-extensions-CentralAuth, MediaWiki-Core-Team, SUL-Finalization: Investigate and fix OOMs caused during account globalization - https://phabricator.wikimedia.org/T78727#962146 (10Nemo_bis) a:3Joe [09:40:46] anyone here who knows how to fix https://gerrit.wikimedia.org/r/#/c/166705/16 quickly? [09:53:19] bd808|BUFFER: *rage* ;) [12:34:24] 3MediaWiki-Core-Team, MediaWiki-General-or-Unknown: Exception if no accelerator (e.g. APC) is available - https://phabricator.wikimedia.org/T86162#962522 (10Anomie) 3NEW [13:10:57] 3MediaWiki-Core-Team, MediaWiki-General-or-Unknown: Exception if no accelerator (e.g. APC) is available - https://phabricator.wikimedia.org/T86162#962588 (10Anomie) 5Open>3Resolved [15:19:28] 3operations, MediaWiki-Core-Team, MediaWiki-ResourceLoader: Bad cache stuck due to race condition with scap between different web servers - https://phabricator.wikimedia.org/T47877#963015 (10chasemp) p:5High>3Normal Seems like no meaningful update in a month, no assignee, and no direct actionable that I can... [15:53:08] <_joe_> if I want to apply some config to a single server via mediawiki-config, what is the standard way to do it? Check the hostname via gethostname()? [15:59:58] probably [16:00:16] I don't think we have a standard way of doing it, as it's not something we usually do [16:06:31] _joe_: ^ [16:07:38] <_joe_> Reedy: the fact we don't tells me something :P [16:07:48] "should you really be doing this?" [16:07:49] <_joe_> but I found ori did the same as I was doing now [16:07:54] What're you trying to do? [16:08:09] <_joe_> Reedy: no, quite the opposite "we never test features on a subset of hosts" [16:08:48] <_joe_> Reedy: https://gerrit.wikimedia.org/r/#/c/183436/ [16:09:00] <_joe_> we need the corresponding switch of config in mediawiki-config [16:10:31] https://noc.wikimedia.org/conf/highlight.php?file=mc.php [16:10:35] where do you want your if? ;) [16:11:39] <_joe_> Reedy: I am writing it there [16:41:38] ^d: I'm in richmond! [16:41:50] its 16! [16:43:23] <_joe_> ? [16:43:30] <_joe_> 16 what? [16:43:54] -8.8C I presume [16:43:57] in proper money [16:44:50] silly Europeans with your silly units. ;) [16:45:14] <_joe_> silly Europeans with your international standard units [16:45:49] 16F [16:45:50] yeah [16:46:02] unpleasantly cold [16:46:04] Jimmy Carter decided in the 1970's that Americans were too dumb to learn SI [16:46:14] <_joe_> bd808: this reminds me when my host in the UK in 1990 told me "in two years, europe will join england, and you'll eventually have to concede and drive on the correct side of the street" [16:46:24] heh [16:50:15] temperature in C has some big advantages, not least of which that you have to teach children about negative numbers real quick [16:55:29] <_joe_> manybubbles: not where yuvi is born, or in sicily :P [16:55:55] <_joe_> I was shocked when I noticed in sicily almost no public thermometer has the leds for the sign [16:57:12] crazy [17:02:50] <^d> manybubbles: My mom was complaining about the cold :p [17:03:01] <^d> I told her it was 65F here yesterday with the same forecasted for today. [17:03:49] <_joe_> manybubbles: well it *snowed* in Palermo once when I was living there. the city basically grinded to an immediate halt. No one is equipped against snow, as it usually snows once every 50/100 years [17:04:11] <_joe_> it's almost the same here in rome, everyone complains it's very cold and today we have 4 C [17:08:30] <^d> AaronS: https://gerrit.wikimedia.org/r/#/c/183434/ [17:12:46] * bd808 is all hopped up on cold meds and not being very productive [17:33:06] ^d: https://gerrit.wikimedia.org/r/#/q/owner:%22Aaron+Schulz%22+status:open,n,z lots of profiler stuff [17:34:07] 3MediaWiki-extensions-CentralAuth, MediaWiki-Core-Team, SUL-Finalization: Investigate and fix OOMs caused during account globalization - https://phabricator.wikimedia.org/T78727#963434 (10Joe) Actually, if the bug is related to an OOM I doubt the new package would fix that. The fix would be for cases where the... [17:34:17] 3MediaWiki-extensions-CentralAuth, MediaWiki-Core-Team, SUL-Finalization: Investigate and fix OOMs caused during account globalization - https://phabricator.wikimedia.org/T78727#963445 (10Joe) a:5Joe>3None [17:34:31] <_joe_> legoktm: ping [17:34:54] hi [17:35:07] I guess hoo was thinking about something else then? :/ [17:36:07] huh? [17:36:13] context? [17:36:40] hoo: https://phabricator.wikimedia.org/T78727 [17:37:30] <^d> AaronS: Yeah I was looking. [17:37:31] legoktm: How do you know it's OOM? [17:37:33] <_joe_> legoktm: no it was me thinking the issues of a specific user could be due to something else [17:37:53] <_joe_> legoktm: also, we should try to reproduce this OOM I guess [17:38:05] <_joe_> it's not like it is an HHVM internal bug [17:38:32] hoo: because at the exact time of the centralauth logs of them trying to merge their account there's an hhvm log entry for OOM on the same host [17:38:58] Ok [17:39:09] <_joe_> it's just counting memory used by php better than zend, probably [17:39:09] but I couldn't see that for any of the Special:MergeAccount cases [17:39:53] I found the entry for the case I linked on the bug [17:44:17] legoktm: Can you paste it? [17:44:53] * legoktm looks [17:49:08] 3Wikidata, MediaWiki-Core-Team, wikidata-query-service: Investigate Titan supernode issues - https://phabricator.wikimedia.org/T85173#963600 (10bd808) p:5Triage>3Normal [17:49:18] 3Wikimedia-Blog, MediaWiki-Core-Team: Finish up blog post for Cirrus - https://phabricator.wikimedia.org/T85176#963602 (10bd808) p:5Triage>3Normal [17:49:25] 3Wikidata, MediaWiki-Core-Team, wikidata-query-service: Experiment with Wikidata Query Engine - https://phabricator.wikimedia.org/T1028#963604 (10bd808) p:5Triage>3Normal [17:49:54] 3MediaWiki-Core-Team, Librarization, Continuous-Integration: Set up composer validate job for operations/mediawiki-config - https://phabricator.wikimedia.org/T76621#963607 (10bd808) p:5Normal>3Low [17:50:38] 3MediaWiki-Core-Team, Librarization, Continuous-Integration: Set up composer validate job for operations/mediawiki-config - https://phabricator.wikimedia.org/T76621#807514 (10bd808) p:5Low>3Normal [17:50:56] centralauth.log-20150106.gz:2015-01-06 02:32:21 mw1072 enwiki: password Hoary@ruwiki (and a bunch more) [17:50:56] Jan 6 02:32:22 mw1072: #012Fatal error: request has exceeded memory limit [17:51:02] hoo: ^ [17:52:01] <^d> AaronS: What do you think about https://gerrit.wikimedia.org/r/#/c/167655/? [17:52:01] without any info about where that happened? [17:52:04] argh [17:52:50] The merge process should be totally lightweight memory wise... might be that we do something that makes HHVM go crazy, though [17:53:07] hoo: only think I can think of is the creation of all the Password objects. [17:54:17] but that's less than 2k no matter what happens [17:54:29] yeah.... [17:54:57] 3MediaWiki-Core-Team, ContentTranslation-Deployments, MediaWiki-extensions-ContentTranslation: Content Translation Beta Feature security review - https://phabricator.wikimedia.org/T85686#963641 (10csteipp) >>! In T85686#962007, @Nikerabbit wrote: > Is there an existing whitelisting tool or even just a whitelist... [17:55:39] legoktm: Actually no... that thing does some kind of stupid things [17:56:09] So if one has many passwords... [17:56:37] I'll try to come up with a patch that doesn't open up any memory leaks [17:57:18] <^d> anomie: came across this last night and I thought of you and api output formats.... [17:57:31] <^d> http://www-01.ibm.com/support/knowledgecenter/SS9H2Y_6.0.0/com.ibm.dp.xi.doc/json_jsonxconversionexample.html [17:58:24] ^d: It's like wddx, but different! [17:58:44] <^d> haha [18:02:45] removing fields from sqlite sucks so much [18:03:01] Anyone want to give https://gerrit.wikimedia.org/r/#/c/182858/ (specifically ApiResult.php and ApiFormat*.php) a once-over? Basically to make sure it's not "-2, awful design" before I submit a pile of extension changes based on it. [18:03:20] legoktm: ^ Will you be my victim? [18:03:48] heh [18:03:50] I suppose [18:03:56] Thanks! [18:07:35] <^d> AaronS: I know right? [18:09:32] anomie: what does kvp stand for? [18:09:36] <^d> AaronS: I only have one unreviewed patch now! [18:09:39] legoktm: key-value pair [18:10:20] <^d> Katholieke Volkspartij, a Dutch political party [18:10:29] kunal's verbal proficiency [18:10:29] <^d> Kalamazoo Vegetable Parchment, a defunct papermill once headquartered in Parchment, Michigan, USA [18:10:49] <^d> Peak kilovoltage (kVp), a unit of radiation used in x-ray imaging [18:12:16] kindred venereal pathology [18:13:28] ori: https://gerrit.wikimedia.org/r/#/c/181248/ [18:13:48] AaronS: keep verifying patches! [18:15:20] <^d> AaronS: For 183455, did you grep to make sure there's no extension callers to profileIn|Out? [18:18:30] <_joe_> ori: I merged the nutcracker patch in puppet, and prepared one for mediawiki [18:18:40] <_joe_> but I wanted you to take a look [18:18:53] \o/, will do [18:21:18] anomie: It looks fine, is there any reason for ApiResult to stay a ContextSource? it only gets used for request values in continuation data, messages for warnings/errors, and checking one config setting [18:23:22] legoktm: Messages for warnings/errors is the main reason. I suppose it could just have an internal context source it uses for that. [18:23:41] ^d: they all use wf* ones [18:23:56] Although then I'll have to dig through all the extensions again and make sure they're not using any ContextSource methods on the ApiResult... [18:27:11] If any extension is doing that, that would be silly, since in most cases it'll be inside an API module and that is a contextsource itself [18:28:45] 3MediaWiki-Core-Team: Enable ProfilerSectionOnly - https://phabricator.wikimedia.org/T86215#963796 (10aaron) 3NEW a:3aaron [18:29:05] 3MediaWiki-Core-Team: Support strategy #1 in PoolCounter classes - https://phabricator.wikimedia.org/T85026#963804 (10aaron) p:5Normal>3Low [18:29:54] TIL that MediaWiki could have been named "Pediawiki" -- https://lists.wikimedia.org/pipermail/wikipedia-l/2001-August/000390.html [18:30:10] And here I was thinking that MediaWiki was a horrible name... [18:30:13] <^d> AaronS: I figured that was mostly the case but thanks for double checking [18:30:26] * AaronS already looked [18:30:54] <^d> kk [18:31:02] SMW is wrapping profiler though with some wierd thing [18:31:30] <^d> bd808: Or Encyclode haha [18:32:15] that's the only thing I saw [18:41:08] legoktm: Took a stab at it [18:41:56] !( $password instanceof Pbkdf2Password ) why? [18:42:17] legoktm: New passwords are UTF-8 all the time ;) [18:42:25] and new passwords are pbkdf2 [18:42:56] Could be that all type B passwords are also utf8, but for that I would need to check the history to when we converted wikis and when type b came [18:43:00] so I took the easy way [18:44:07] also hash_equals gets slower on longer hashes... and pbkdf2 hashes are the longer [18:44:11] * longest we have [18:49:34] legoktm: Feel free to undo that [18:49:37] I don't really mind [18:49:45] just thought it would be an opportunity to save [18:52:34] legoktm: Extensions do silly things sometimes. For example, https://gerrit.wikimedia.org/r/#/c/171895/ or https://gerrit.wikimedia.org/r/#/c/171893/ [18:56:03] legoktm: Code reviews of the stuff I have in gerrit would be appreciated ;) [18:56:09] food now [18:59:19] anomie: haha there's a good chance the second one was my fault >.< [19:05:05] legoktm: https://git.wikimedia.org/blob/mediawiki%2Fextensions%2FZeroBanner.git/d7e7b1da58918f968e19417817566cb577839782/includes%2FApiZeroBanner.php#L70 :/ [19:05:24] ugh [19:11:39] OTOH, that seems like the only one, unless I'm forgetting something when I think all local variables for ApiResult instances have names ending (case-insensitively) in "result", "api", or are "$res" or "$r". [19:32:24] anomie, you're using commit id as a topic? https://gerrit.wikimedia.org/r/#/q/I7b37295e8862b188d1f3b0cd07f66ac34629678e,n,z :P [19:34:42] MaxSem: I get lazy when using the exact same commit message for a pile of extension updates [19:36:02] as a result, I can't even get to the core commit [19:36:31] * anomie grumbles that whoever added a unit test for "all classes are in $wgAutoLoadWhatever" didn't fix the extensions that broke from it [19:37:06] MaxSem: Oh, ha. Core commit is I7b37295e8862b188d1f3b0cd07f66ac34629678f. [19:37:36] Which is different, just not in the shortened version... [19:38:47] <^d> AaronS: Are you going to clean up all the extensions using profiling now? :p [19:38:55] <^d> Or hoping people will take the lead we've done in core [19:40:32] you could always do it ;) [19:40:58] <^d> I was probably going to take a pass at deployed extensions soon. [19:55:39] 3MediaWiki-Core-Team, MediaWiki-API: Clean up ApiResult and ApiFormatXml, create format=json2 - https://phabricator.wikimedia.org/T76728#964062 (10Anomie) Necessary extension changes are [[https://gerrit.wikimedia.org/r/#/q/I7b37295e8862b188d1f3b0cd07f66ac34629678e+status:open,n,z|here]] [19:56:07] 3MediaWiki-extensions-ContentTranslation, MediaWiki-Core-Team, ContentTranslation-Deployments: Content Translation Beta Feature security review - https://phabricator.wikimedia.org/T85686#964064 (10Nikerabbit) htmlpurifier seems to be written in PHP, which is not ideal given that cxserver and cx extension are wri... [20:18:48] 3MediaWiki-Core-Team: Add support for svg in logo - https://phabricator.wikimedia.org/T86229#964128 (10Paladox) 3NEW [20:19:07] 3MediaWiki-Core-Team, MediaWiki-skins-Vector: Add support for svg in logo - https://phabricator.wikimedia.org/T86229#964128 (10Paladox) [20:19:19] wut [20:30:36] hoo|away: do you mind if I amend your patch? [20:35:17] anomie: use thunderbird! [20:35:27] the gmail web ui filtering options are pretty terrible [20:47:14] legoktm: As long as you don't break it and do reasonable changes, surely not ;) [20:51:52] robla: I got your email to services@ and it's in the archive: https://lists.wikimedia.org/pipermail/services/2015-January/000002.html [20:52:15] hoo: I uploaded my changes, now trying to figure out why they fail in zend but pass in hhvm [20:52:21] oh, there it is. I wonder if I just had a cached archive page [20:53:13] * robla hits shift reload on https://lists.wikimedia.org/pipermail/services/ and January magically appears [20:53:44] legoktm: Oh that :P I only added that support back in last minute [20:53:57] because I thought some callers should not change [20:54:11] why? [20:54:26] convenience [20:54:38] having two ways to call a function is messy :/ [20:55:02] yeah... stupid PHP ... [20:55:15] in a reasonable language you could have both signatures under the same name [20:55:36] lol [20:55:37] # [20:56:57] also I noticed that if I set $wgPasswordDefault = 'B'; the CA tests are 3 times faster [20:57:05] hahaha [20:57:12] That's what I did to TestUser, remember? [20:59:47] yeah [21:00:01] ok, I think I'm happy with the patch now [21:00:46] also, in hhvm if you use ReflectionMethod to access a protected/private function, it ignores whether you've called ReflectionMethod::setAccessible() [21:01:10] hoo: +1 my changes? [21:01:34] need to look again [21:03:40] done [21:04:16] AaronS: TimStarling: waiting in the channel for y'all [21:12:40] 3Security-Reviews, MediaWiki-Core-Team, Phabricator: Aphlict security review - https://phabricator.wikimedia.org/T1286#964251 (10Qgil) https://secure.phabricator.com/T6559 has been resolved as Fixed. In principle, Aphlict has no dependency on Flash anymore. Being Aphlict and dependencies fully open source now,... [21:24:14] legoktm, wait what? [21:24:19] it ignores setaccessible? [21:32:03] Krenair: http://3v4l.org/WqnUZ [21:32:53] <^d> manybubbles: I'm thinking of just dropping the ES thing @ dev summit. [21:33:04] <^d> Like only 3 people were interested and we're both going to be hella busy as it is [21:33:39] legoktm, so it either crashes or ignores the accessibility? [21:36:00] ^d: fair [21:36:04] apparently, I only tested with hhvm 3.3.1 [22:20:20] <^d> bd808: I bumped the keybase issue asking for MW identities :p [22:20:59] <^d> https://github.com/keybase/keybase-issues/issues/704 [22:21:19] * bd808 looks [22:21:48] <^d> They want you to make an edit and then copy the permalink to the revision [22:21:51] <^d> I said use the API :p [22:23:38] neat. I'm still not quite sure what keybase is good for. :) I got an invite out of the blue and signed up but haven't really played with it [22:23:57] extending WoT via twitter seems weird [22:24:29] But the kids these days like their social media (or something) [22:24:36] bd808: Can you sign up and send me an invite? [22:24:50] csteipp-ish: Sure [22:24:51] I've been meaning to play with it... [22:26:26] csteipp-ish: it's in the tubes [22:29:46] I've got 3 more invites and the window open if anyone else wants a keybase.io account [22:34:12] bd808: hit me (why not) greg@grossmeier.net [22:34:46] greg-g: {{done}} [22:38:14] bd808: ty [22:41:05] I've got 3 too [22:42:00] legoktm: The patch you merged today... shall we SWAT it tonight? [22:42:14] DerHexer would like to continue asking people to merge their accounts [22:46:30] <^d> bbl. [23:01:47] hoo: yeah that sounds like a good idea [23:04:56] legoktm: Ok, I'll sign us up, then [23:11:12] ori: If you still feel like merging https://gerrit.wikimedia.org/r/#/c/182323/ I think the discussion on the patch has died down to nothing. [23:14:46] 3MediaWiki-Core-Team: Parsoid performance analysis - https://phabricator.wikimedia.org/T85870#964679 (10tstarling) api.log already existed, it is the "api" log channel, generated by ApiMain::logRequest(). [23:16:58] bd808: needs manual rebase :) [23:17:01] rebase and i'll +2 [23:17:18] can do [23:21:00] 3Security-Reviews, MediaWiki-Core-Team, Phabricator: Aphlict security review - https://phabricator.wikimedia.org/T1286#964692 (10csteipp) >>! In T1286#964251, @Qgil wrote: > https://secure.phabricator.com/T6559 has been resolved as Fixed. In principle, Aphlict has no dependency on Flash anymore. > > Being Aphli... [23:31:59] 3Librarization, MediaWiki-Core-Team: xhprof for MW - https://phabricator.wikimedia.org/T759#964712 (10tstarling) 5Open>3Resolved [23:38:17] what is the status of the memcached issue? [23:38:50] TimStarling: mw1230 and 1231 are about to be switched over [23:41:05] 3MediaWiki-Installer, MediaWiki-Core-Team, Librarization: composer.json should be useable by WMF, core and extensions - https://phabricator.wikimedia.org/T67188#964747 (10bd808) Does the addition of my merge plugin resolve this issue entirely or are there additional steps that need to be taken? `composer require... [23:42:02] TimStarling: {{done}}. _joe_ volunteered to deploy to the rest of the cluster if those two hosts look OK. [23:42:58] bd808: on #hhvm: " ori: if you have a moment, I was also inspired by the wikipedia post to look into getting the remote hhvm debugger to work, but I couldn't get the daemon to even list on port 8089 https://gist.github.com/jazzdan/71eb4a3aa5d4922a542f . Were there any gotchas there?" [23:43:18] I never really looked at the remote debugging thing too closely, but IIRC you did -- any tips to share? [23:43:58] yeah there were a bunch of weird things as I recall. Let me look at the puppet stuff [23:44:24] but the client is not updated? [23:46:15] mc.php still looks the same [23:46:25] i still need to sync a change [23:51:41] synced now, but i don't think it's working correctly on mw1230/31 [23:52:42] yeah, just revert the client change for now [23:52:58] I am testing it with hhvm -md and it looks like it might not be working [23:53:08] well, they're depooled [23:53:22] so we don't have to revert immediately [23:53:31] what's hhvm -md? [23:53:56] addServer('unix:///var/run/nutcracker/nutcracker.sock', 0); $res = $c->getResultMessage(); echo "$res\n"; $c->set( 'ori.test', 12 ); $res = $c->get( 'ori.test' ); echo "$res\n"; [23:54:02] this outputs SUCCESS followed by a blank line [23:55:04] the permissions on the socket are wrong [23:55:24] hhvm -md is roughly equivalent to php -a [23:56:06] whose responsibility is it to set the right permissions -- the init script or nutcracker itself? [23:56:22] i think faidon wrote the init script [23:57:01] it should be g+w, yeah? [23:57:31] the init script executes nutcracker using start-stop-daemon, which accepts a --umask arg [23:57:57] i suppose g+w won't be good enough either if apache is not part of the nutcracker group