| RAOF | robert_ancell: Yo! | 00:02 |
|---|---|---|
| robert_ancell | RAOF, howdy | 00:02 |
| robert_ancell | RAOF, regarding bug 1583624, do you know why a Mir session would be attempting to work out what VT it is on? | 00:03 |
| ubot5 | bug 1583624 in Light Display Manager "Mir cannot open a tty when started by lightdm" [Undecided,Incomplete] https://launchpad.net/bugs/1583624 | 00:03 |
| RAOF | Hm. Because for some reason it's trying to run on bare KMS. | 00:05 |
| RAOF | Is unity8 being passed the --host-socket option? | 00:05 |
| robert_ancell | RAOF, not by LightDM | 00:06 |
| RAOF | Well, something should be. | 00:06 |
| robert_ancell | That was my guess, some change has made it think it's no longer running under u-s-c | 00:06 |
| RAOF | Something needs to tell the greeter's Mir to connect to usc rather than try to drive the hardware. | 00:07 |
| robert_ancell | LightDM sets MIR_SOCKET - is that no longer sufficient? | 00:07 |
| RAOF | I don't think that was ever sufficient? | 00:08 |
| robert_ancell | That's how Unity 8 sessions work... | 00:08 |
| robert_ancell | LightDM picks a socket name, starts U-S-C with that then runs the sessions with MIR_SOCKET set (the greeter is just a special case of a session). | 00:09 |
| RAOF | robert_ancell: Hah. https://bugs.launchpad.net/bugs/1290345 | 00:11 |
| ubot5 | Launchpad bug 1290345 in Mir "The server-side use of $MIR_SOCKET is confusing" [Medium,Fix released] | 00:11 |
| RAOF | So, Mir 0.21 changed the behaviour to require MIR_SERVER_HOST_SOCKET (or, equivalently, passing --host-socket=). | 00:12 |
| robert_ancell | That looks like it | 00:13 |
| robert_ancell | So, I'm still confused. A Mir client reads MIR_SERVER_HOST_SOCKET for the parent Mir server to connect to? | 00:14 |
| robert_ancell | And MIR_SOCKET is the "socket you should open for your children to connect to" | 00:15 |
| RAOF | No; a Mir client (which uses mir_connect_sync(nullptr)) reads MIR_SOCKET. | 00:15 |
| robert_ancell | But a shell (which is a Mir client in this case) reads MIR_SERVER_HOST_SOCKET? | 00:16 |
| RAOF | For a Mir *server* it checks if MIR_SERVER_HOST_SOCKET is set (or, equivalently, --host-socket), and if so uses that host to nest under. | 00:16 |
| robert_ancell | So it sounds like I just need to replace MIR_SOCKET with MIR_SERVER_HOST_SOCKET and it should still work. | 00:16 |
| robert_ancell | I will however need to keep setting MIR_SOCKET for backwards compatibility | 00:17 |
| RAOF | Yes. | 00:17 |
| robert_ancell | Yay for API stability ;) | 00:17 |
| RAOF | Of course, there's a good chance this will change again in the not-too-distant future :) | 00:19 |
| robert_ancell | ahahahahah. *sigh* | 00:19 |
| robert_ancell | RAOF, that change in 0.21 was just to remove the backwards compatibility right? | 00:19 |
| RAOF | Yes. | 00:19 |
| robert_ancell | Do you happen to know when the behaviour changed? I might just not bother keeping LightDM backwards compatible because it's too hard with Mir and practically it probably wont matter. | 00:20 |
| RAOF | Hm, actually, I think that MIR_SERVER_HOST_SOCKET has *always* been the correct thing to do. | 00:20 |
| RAOF | And we just automatically set MIR_SERVER_HOST_SOCKET if we found MIR_SOCKET was set. | 00:20 |
| RAOF | So, just set MIR_SERVER_HOST_SOCKET, and I'll tell you if the nested-platform-becomes-an-actual-platform rework changes things :) | 00:21 |
| tsdgeos | greyback: it finished compiling :D | 08:36 |
| greyback | tsdgeos: yay! | 08:44 |
| tsdgeos | ltinkl: are you going to do https://code.launchpad.net/~aacid/unity8/fix_qinputdeviceinfo_leaks/+merge/295799 too or want lpotter to? | 08:46 |
| ltinkl | tsdgeos, yeah, I'd prefer someone else more knowledgeable of the code (lpotter or mzanetti) | 08:46 |
| tsdgeos | oki | 08:46 |
| tsdgeos | lpotter: if you're still around can you have a look at it? ↑ | 08:47 |
| tsdgeos | ltinkl: how do you feel about https://code.launchpad.net/~aacid/unity8/fix_uninitialized_use/+merge/295805 ? | 08:50 |
| ltinkl | tsdgeos, the first 2 are obvious, the 3rd one... need some context, sec :) | 08:52 |
| mterry | tsdgeos, in silo 59, is there anything that would cause the CardTool tests to be slow? | 14:05 |
| tsdgeos | yes | 14:05 |
| tsdgeos | maybe | 14:05 |
| tsdgeos | :D | 14:05 |
| tsdgeos | i mean there was a branch that made things slower | 14:05 |
| tsdgeos | but i thought i had fixed cardTool to be fast again | 14:05 |
| mterry | tsdgeos, we failed autopkg tests because qmluitests timed out. And running them manually, it seems slow | 14:05 |
| tsdgeos | maybe i only made it on card | 14:05 |
| * tsdgeos tries to remember the name of the branch | 14:06 | |
| tsdgeos | E_TOO_MANY_BRANCHES | 14:06 |
| tsdgeos | yes https://code.launchpad.net/~aacid/unity8/make_dash_test_more_stable/+merge/294127 | 14:07 |
| tsdgeos | probably needs the same fix of tst_card on test_cardtool | 14:07 |
| tsdgeos | let me have a look | 14:07 |
| tsdgeos | there's only one such construct in cardTool | 14:09 |
| tsdgeos | mterry: is it noticeably slower than trunk | 14:09 |
| tsdgeos | ? | 14:09 |
| mterry | tsdgeos, I will test... but first I wanted to let the current run finish... it's been a long time. like 20 min at least | 14:09 |
| tsdgeos | locally? | 14:10 |
| mterry | tsdgeos, yeah | 14:10 |
| tsdgeos | that's bad :D | 14:10 |
| mterry | tsdgeos, lp:~ci-train-bot/unity8/unity8-ubuntu-yakkety-landing-059 for testing | 14:10 |
| mterry | tsdgeos, aha, the test finished :) | 14:16 |
| tsdgeos | yeah it's another of those goddam parentless findChild | 14:19 |
| tsdgeos | let me go over the test that have those and make sure they didn't get slow too | 14:20 |
| tsdgeos | mterry: ok, pushed to the branch, should be better now | 14:41 |
| mterry | tsdgeos, oh thanks! | 14:41 |
| * mterry will rebuild | 14:41 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!