[07:08:01] MW is so slow and broken [07:08:42] for anyone other than WMF [19:40:47] Looks like something is broken in editing on mediawiki.org: T134194 [19:40:47] T134194: Section edit broken on www.mediawiki.org - https://phabricator.wikimedia.org/T134194 [19:40:57] anybody knows what's up with that? [20:02:17] it's probably translate crap [21:55:38] bd808: around? [21:56:22] SMalyshev: in meetings for a while [21:56:32] bd808: ah, ok, thanks [21:56:52] bd808: got a vagrant question when/if you have a minute sometime [21:57:14] cool. leave it here or I'll ping you when I get free [21:58:29] bd808: well, so I'm trying to use xdebug and for some reason port 9000 inside vagrant is not connecting to port 9000 on host. And it worked before, so I wonder what may be broken.... [22:00:17] SMalyshev: ugh. who is the xdebug person to talk to... maybe tgr? [22:00:36] I never have used xdebug myself [22:01:40] bd808: it's not really about xdebug. xdebug works as soon as port forwarding works. it's about connecting port on host with port inside vagrant [22:41:43] SMalyshev: remote_connect_back is way less hassle IMO [22:42:01] tgr: how that is working? [22:42:51] tgr: does it need special IDE support (e.g. eclipse)? [22:43:03] you use a browser extension or whatever to set XDEBUG_SESSION cookie, XDebug notices it and connects to the address the request is coming from [22:43:17] from the IDE point of view there is no difference [22:43:42] tgr: well, I'm not sure it would be able, since the adress is localhost... it's vagrant [22:44:14] vagrant uses an internal network unless you changed the vagrantfle [22:44:16] tgr: the whole problem is that port forward for 8080 works fine, but for 9000 vagrant->host does not [22:44:59] also vagrant generally does not do guest->host forwards AFAIK [22:45:11] SMalyshev: what sets up that port forward for you? Is is part of a role? [22:45:12] tgr: well, it worked before for me... [22:45:22] * bd808 is out of his meeting [22:45:50] bd808: I don't really know... I think 8080 is forwarded in standard vagrant config [22:45:56] *nod* [22:46:09] what about the port 9000 forward you want/need? [22:46:36] but that's host->vagrant, and I need 9000 to go in opposite direction. If I forward it manually with ssh it works, but I remember there was a way to make it work via vagrant [22:47:05] bd808: I need 9000 to go in opposite direction - when xdebug connects to 9000 inside vagrant, it should go to the host [22:47:14] oh! I bet I know what changed. Let me find the bug and work around [22:47:32] Vagrant started binding port tunnels to 127.0.0.1 at some point [22:48:14] although that should still work for a xdebug from a local IDE [22:48:22] SMalyshev: if you look in the Vagrantfile you'll see something like "vm.network :private" [22:48:46] that will create a virtual LAN with just your host and guest machine on it [22:49:08] so the vagrant box is fully able to address the host machine directly [22:49:54] and xdebug can figure out the right IP for that, because that's where the request comes from [22:50:20] tgr: ah! ok, let me try with that [22:50:30] so just delete any manually set remote address, make sure remote_connect_back is enabled, and things should work out of the box [22:50:53] you can connect back via a tunnel or whatever but it's unnecessarily complicated [22:51:26] you can enable xdebug.remote_log, it helps with figuring out problems sometimes [22:52:46] if things work, it should show something like [22:52:47] I: Checking remote connect back address. [22:52:48] I: Remote address found, connecting to 10.0.2.2:9000. [22:52:48] I: Connected to client. :-) [22:52:52] tgr: ok, this seems to work [22:52:56] thanks! [22:53:13] `vagrant config host_ip 0.0.0.0` was the thing I was thinking about. See T115139 for discussion of that change [22:53:13] T115139: Forwarded ports are bound only to localhost - https://phabricator.wikimedia.org/T115139 [22:53:29] now eclipse has path mapping messed up, but that's probably different issue [22:53:45] heh. SMalyshev you wrote that bug :) [22:53:59] bd808: my memory is very short :) [22:54:28] bd808: thanks! [23:56:24] oh, interesting [23:56:46] I was reviewing https://gerrit.wikimedia.org/r/#/c/285449 yesterday but now it is merged [23:57:30] still untested I guess? [23:57:35] anomie AaronSchulz ? [23:58:18] I was mostly wondering how this code is ever hit, especially in the non-opportunistic case