| alan_g | dednick: this might well affect you - https://code.launchpad.net/~alan-griffiths/miral/fix-initialization-order/+merge/318353 (I can't see how this ever worked) | 12:44 |
|---|---|---|
| === hikiko is now known as hikiko|ln | ||
| === hikiko|ln is now known as hikiko | ||
| tjaalton | bregma: hi, I still need this patch to build the xserver, maybe you have glamor disabled on your build? http://pastebin.com/YpDyBkqP | 16:45 |
| bregma | tjaalton, I just slap my refreshed patch into your git packaging branch and submit to a zesty pbuilder to build, but I don't believe I have -proposed in my pbuilder environment, maybe that's the difference? | 16:48 |
| tjaalton | bregma: no that's not required | 16:49 |
| tjaalton | ah | 16:50 |
| tjaalton | well | 16:50 |
| tjaalton | if you use the packaging branch then you get this patch :) | 16:50 |
| tjaalton | but it doesn't build without | 16:50 |
| tjaalton | "xmir-fixes.diff" | 16:51 |
| bregma | tjaalton, I can import that patch upstream if you prefer | 16:58 |
| tjaalton | bregma: yep, please | 17:03 |
| tjaalton | had it separate since 1.18 :) | 17:04 |
| bregma | tjaalton, OK, give me a little while to verify | 17:09 |
| tjaalton | you can check with the packaging branch, it's uptodate now | 17:11 |
| mterry | Hello! Can I get some help with some odd unity8/mir behavior I'm seeing that's interrupting some snappy work? Let me lay out scenario... | 17:45 |
| mterry | I'm trying to make the unity8 snap use the mir-libs snap | 17:45 |
| mterry | And so I'm pointing it at the server-platform bits there | 17:45 |
| mterry | But Mir is trying to load library from system... | 17:46 |
| mterry | Here's a pastebin where I've printed out each dlopen as we intercept it in some ld-preload trickery we have | 17:46 |
| mterry | http://pastebin.ubuntu.com/24079316/ | 17:46 |
| mterry | AlbertA: ^ ? | 17:46 |
| mterry | You can see it finds the server platform in mir-libs | 17:46 |
| mterry | But later on it tries to load from system, gets shunted into snap by our ld-preload hacks, and then isn't found | 17:47 |
| mterry | It's loading the client platform just fine thoug | 17:48 |
| alan_g | mterry: that sounds like something that Saviq was asking about. So: you're trying to connect to a host server (USC) that's using drivers from system? | 17:50 |
| mterry | alan_g: yes -- and that is a ticking time bomb, but I would expect it to work in this case, where the versions match | 17:50 |
| Saviq | bug #1665123 | 17:50 |
| ubot5 | bug 1665123 in Unity8 Session Snap "Mir graphics platform determined based on host server outside of snap" [High,Triaged] https://launchpad.net/bugs/1665123 | 17:50 |
| alan_g | mterry: agreed, this isn't a supported scenario. | 17:51 |
| Saviq | mterry, is /snap/unity8-session/x7/usr/lib/x86_64-linux-gnu/mir/server-platform/graphics-mesa-kms.so.12 there? should it be there? | 17:52 |
| mterry | Saviq: it's not there in this testing snap I'm making where I'm trying to use mir-libs instead | 17:52 |
| mterry | Saviq: I had thought that bug was just about version mismatch, but now I see it dives deeper | 17:52 |
| mterry | alan_g: ... well Mir team won't support a snap that doesn't use mir-libs. But now the Mir team won't support a unity8 that does? | 17:53 |
| alan_g | mterry: there are *a lot* of unanswered questions about how we can support Mir on snaps. | 17:54 |
| Saviq | mterry, MIR_SERVER_PLATFORM_PATH doesn't help? | 17:55 |
| alan_g | So: Mir team can't support snaps (yet) | 17:55 |
| alan_g | Saviq: no | 17:55 |
| mterry | alan_g: understood... but but we've had several meetings where the Mir team said mir-libs was the answer | 17:55 |
| mterry | Sounds like the guidance is that the Personal team is blocked on Mir team | 17:55 |
| alan_g | mterry: Mir team says "both server and client should use mir-libs" - but that's not what USC is doing. | 17:55 |
| mterry | I mean, not *blocked* blocked. We can run something on the screen, I'll just stop using mir-libs | 17:56 |
| mterry | alan_g: there was talk of Mir team providing an implicit :mir-libs interface on Ubuntu 16.04 that pointed at USC libraries to solve this | 17:57 |
| Saviq | yeah, we need more meets like that, it doesn't seem to be the answer AFAICT... there's just too much ABI dependencies there still | 17:57 |
| mterry | alan_g: that way same guidance (use mir-libs interface) would work in all places | 17:57 |
| mterry | But I had hoped in mean time, we could use it as long as versions matched | 17:57 |
| mterry | Sounds like I'll avoid it for now until it works better | 17:57 |
| alan_g | mterry: well... I don't think nesting will work that way either. | 17:58 |
| alan_g | We have plans that resolve a lot of these issues. But not in the 0.26 series. | 17:58 |
| * alan_g knows that doesn't help you now | 17:59 | |
| mterry | Yeah, this is disappointing after fighting with the Mir team to avoid the long-term problems of mir-libs, but settling on it because it was the only solution that seemed to work in short term. But now it doesn't even help with the short term problems. So... | 18:00 |
| alan_g | mterry: the way we get nested working is https://trello.com/c/TzcWYzxM/498-remove-guest-platforms | 18:01 |
| mterry | I don't have access to that, but no biggie, understood there's work in progress | 18:01 |
| mterry | Hopefully mir-libs will eventually solve our middle-term problems | 18:02 |
| alan_g | IMO it is the least painful option, but far from ideal. | 18:03 |
| alan_g | And not pain free | 18:03 |
| AlbertA | mterry: so this is unity8 snap on top of classic? | 18:09 |
| AlbertA | with usc started by classic? | 18:09 |
| mterry | AlbertA: unity8 snap on top of 16.04 yes, with deb-based USC | 18:10 |
| mterry | i.e. the only currently supported way to run unity8 snap | 18:10 |
| mterry | Hmm, I can use a symlink trick to solve this in short term... | 18:12 |
| AlbertA | mterry: so, isn't this the issue in your particular example? | 18:13 |
| AlbertA | "MIKE dlopen /usr/lib/x86_64-linux-gnu/mir/server-platform/graphics-mesa-kms.so.12 -> /snap/unity8-session/x7/usr/lib/x86_64-linux-gnu/mir/server-platform/graphics-mesa-kms.so.12" | 18:13 |
| AlbertA | maybe the shim preload should redirect mir libs as well... | 18:14 |
| mterry | AlbertA: yeah I'll just set up a symlink in our snap to point at mir-libs version -- that can get us working in short temr | 18:14 |
| AlbertA | right | 18:14 |
| mterry | AlbertA: symlink will do same. I'd prefer to avoid adding more code to our preload | 18:14 |
| mterry | Trying to drop that hack :) | 18:14 |
| AlbertA | mterry: yep | 18:14 |
| bregma | tjaalton, merged in that patch, build-tested and verified on zesty | 18:39 |
| tjaalton | bregma: great, thanks | 18:40 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!