| === superm1|away is now known as superm1 | ||
| === superm1 is now known as superm1|away | ||
| soren | tjaalton: Ok, I've read up on the subject of the kvm brokenness due to evdev, and it almost makes me cry. | 08:55 |
|---|---|---|
| tjaalton | soren: ok, so the guest is broken? | 08:56 |
| soren | No. | 08:56 |
| soren | It's... well.. | 08:56 |
| soren | Let me explain the issue: | 08:56 |
| soren | kvm does full virtualisation, so it emulates a keyboard of some flavour. | 08:57 |
| soren | This means that it has a virtual keyboard that sends scan codes to the guest os like a regular keyboard would do. | 08:57 |
| soren | This is key. | 08:58 |
| soren | it doesn't send "f", if you press the f button. It sends keycode 41. | 08:58 |
| soren | Let's rewind about a year.. | 08:59 |
| soren | Back then, the way this was accomplished in kvm (and qemu), you had to pass "-k fi" to kvm if you had a Finnish keymapping. | 08:59 |
| soren | That way, kvm could look at the keysym(!) look it up in the finnish keymap it had, figure out to which keycode that corresponded, and sent that to the guest. | 09:00 |
| soren | This was annoying for many reasons. | 09:00 |
| soren | First of all, it was a bit flaky from time to time. | 09:01 |
| soren | E.g. azerty keyboards need you to hold down shift to type numbers, so kvm couldn't send the shift key to the guest until it knew what you had pressed. | 09:02 |
| soren | Also, you set this in kvm, so when you're using vnc, you have to be using the same mapping as the one you told kvm you'd be using. | 09:02 |
| soren | this sucks big time if you want to give different users access to kvm instances and they don't share a keymap. | 09:03 |
| soren | ...or if you simply don't know ahead of time which keymap the client will be using (in hosted environments for instance). | 09:03 |
| soren | Fast forward to... hmm.. February this year. | 09:03 |
| soren | I got fed up with this and started looking for a way to properly fix this.. Anthony Liguori (vnc and kvm hacker (very, very convenient combination)) extended the VNC protocol to be able to send keycodes instead of keysyms. | 09:04 |
| soren | Er... I should have said scancodes almost every time I said keycodes, I think. | 09:05 |
| soren | Anyhow.. | 09:05 |
| soren | It turned out that the key code X reported corresponded exactly to the scan code from the keyboard. Great success. So we took those, sent them untranslated over the wire to kvm, and kvm passed them directly, untranslated to the guest. | 09:06 |
| soren | This was fantastic. | 09:06 |
| soren | No more passing "-k <lang code>" to kvm to make things work (somewhat). The world was a better place. | 09:07 |
| soren | Fast forward a few months.. | 09:07 |
| soren | Now, evdev comes along. | 09:07 |
| soren | evdev is not doing anything wrong per se.. There's nothing in the X protocol that dictates that keycodes should correspond to the hardware scan codes from the keyboard. | 09:07 |
| soren | and it just so happens, that with evdev, they don't. | 09:08 |
| tjaalton | Ok, I see your point | 09:08 |
| soren | ...so we no longer have the raw scan codes. | 09:08 |
| soren | There are a few different solutions.. | 09:08 |
| soren | We could ditch evdev. | 09:09 |
| soren | We could tell the evdev developers to "fix" it. (as I said, they're not doing anything wrong per se, but it's annoying to us) | 09:09 |
| tjaalton | that's possible | 09:09 |
| tjaalton | since fedora should be seeing the same bug by now | 09:10 |
| soren | We could come up with a way of a) detecting the use of evdev, and b) if detected, translate evdev keycodes to scancodes again. | 09:10 |
| soren | According to my sources, Fedora are not switching to evdev. | 09:10 |
| soren | Because of this very bug. | 09:10 |
| soren | s/bug/issue/ | 09:10 |
| tjaalton | anyway, maybe it would be best if you could mail xorg@lists.freedesktop.org about this problem | 09:10 |
| tjaalton | fedora rawhide is already using it | 09:10 |
| soren | I think someone already did. I'll need to check up on it. | 09:11 |
| tjaalton | they didn't switch to it for F9 | 09:11 |
| soren | Oh. | 09:11 |
| soren | Er.. | 09:11 |
| tjaalton | but there were other bugs too | 09:11 |
| soren | Right, that might be true. | 09:11 |
| soren | What they're *not* doing is shipping the patch to kvm that enables the raw keycode thing to happen. | 09:11 |
| tjaalton | ah | 09:12 |
| tjaalton | so they like old bugs | 09:12 |
| soren | We're the only ones who ship it, and it really makes a massive difference. | 09:12 |
| tjaalton | ok.. | 09:13 |
| tjaalton | in that case the third solution sounds cleaner | 09:16 |
| soren | Detect evdev and try to do the mapping? | 09:21 |
| tjaalton | right.. | 09:22 |
| tjaalton | we can't avoid evdev forever, so better fix/work around any issues that come up :) | 09:24 |
| soren | Do you know how to detect evdev? | 09:45 |
| tjaalton | how low level should it be? | 09:49 |
| tjaalton | the detection | 09:49 |
| tjaalton | setxkbmap -print is one way if it's just a script that configures kvm runtime | 09:49 |
| tjaalton | there should be better ways to get that | 09:49 |
| soren | ah, yes. I could look at setxkbmap and see what it does. | 09:50 |
| === superm1|away is now known as superm1 | ||
| === superm1 is now known as superm1|away | ||
| === superm1|away is now known as superm1 | ||
| bdmurray | bryce: since alpha4 is out did you want to do something with the -ati bugs? | 20:43 |
| bryce | yep | 20:45 |
| bdmurray | was tormod doing some of it manually though? | 20:46 |
| tseliot | bdmurray: can you renew my membership in the Ubuntu Bug Control (ubuntu-bugcontrol) Launchpad team, please? | 20:48 |
| superm1 | tseliot, can't you do that yourself? | 20:49 |
| tseliot | superm1: if I could I would do it myself ;) | 20:50 |
| tseliot | "To prevent this membership from expiring, you should get in touch with one of the team's administrators: etc." | 20:50 |
| superm1 | oh | 20:50 |
| bdmurray | tseliot: what is your launchpad username? | 20:50 |
| tseliot | bdmurray: albertomilone | 20:50 |
| bdmurray | tseliot: Oh sure, done! | 20:51 |
| tseliot | bdmurray: thanks a lot :-) | 20:51 |
| bdmurray | no problem | 20:53 |
| tjaalton | bryce: btw, got input-device properties working with evdev today. peter said he'd apply the backported patches to xserver 1.5 etc | 20:57 |
| bryce | tjaalton: excellent | 21:06 |
| bryce | bdmurray: he might have, but it doesn't hurt to repeat. | 21:07 |
| bdmurray | heh | 21:07 |
| bdmurray | bryce: How does http://pastebin.osuosl.org/21740 look for a comment? | 22:18 |
| LaserJock | anybody here know how to set up a touchpad in Intrepid? | 22:20 |
| bryce | bdmurray: sounds good | 22:20 |
| LaserJock | there used to be a tab in the Mouse settings but I don't see it anymore | 22:20 |
| bryce | LaserJock: I've not tested any touchpads, but I saw three user contributed docs which I've linked to from https://wiki.ubuntu.com/X/Config | 22:21 |
| LaserJock | bryce: you don't have any laptops? :-) | 22:22 |
| bryce | er, /not tested any touchpads that required manual configuration/ | 22:23 |
| LaserJock | ah | 22:26 |
| LaserJock | well, mine works, just too well ;-) | 22:26 |
| LaserJock | I gotta turn of the tap-to-click or I start typing everywhere | 22:26 |
| LaserJock | *off | 22:26 |
| bdmurray | bryce: hmm, I just realized I should subscribe to all of these | 22:53 |
| pwnguin | anyone wanna take a look at https://bugs.launchpad.net/bugs/234466 ? | 23:01 |
| ubottu | Launchpad bug 234466 in wacom-tools "[hardy] Pen crazy in tablet PC" [Undecided,Confirmed] | 23:01 |
| pwnguin | there's a debdiff | 23:02 |
| pwnguin | or not | 23:03 |
| * pwnguin reads harder | 23:03 | |
| tjaalton | I can update wacom tomorrow | 23:04 |
| pwnguin | cool | 23:05 |
| pwnguin | i just got a new job yesterday | 23:05 |
| tjaalton | besides, I'll try to make it support input-properties.. | 23:05 |
| tjaalton | pwnguin: congrats! | 23:05 |
| pwnguin | its a nice, easy part time job | 23:05 |
| tjaalton | serving fast-food?- | 23:05 |
| tjaalton | :) | 23:05 |
| bdmurray | I'm not sure that qualifies as easy | 23:06 |
| tjaalton | hehe | 23:06 |
| pwnguin | lab assistant for a graphic design program | 23:06 |
| pwnguin | its all osx, which is good... and bad | 23:06 |
| tjaalton | (there's an old joke in finland; what an arts grad asked the tech student at McDonald's? "want fries with that?") | 23:10 |
| tjaalton | pwnguin: sounds good | 23:10 |
| pwnguin | well, there's actually a lot of help wanted posters | 23:36 |
| pwnguin | presumably because they're a community college rather than a university, and are cheaper | 23:36 |
| bryce | congrats pwnguin | 23:40 |
| bryce | pwnguin: make sure to pimp inkscape ;-) | 23:40 |
| pwnguin | heh | 23:40 |
| pwnguin | illustrator isn't very popular | 23:40 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!