[08:21] is there a patch in X to try drivers in a specific order for a given PCI? [08:21] like, try nouveau, then nvidia, then nv? [08:23] pwnguin: no [08:23] the selection only works if there's no xorg.conf [08:24] that's why there's no patch to add those [08:24] obviously [08:24] i mean, obviously it only auto selects if you dont have an xorg.conf [08:25] it's a different codepath [08:25] what is? [08:25] and likely fixable to apply also to the case where you have an xorg.conf but no driver specified [08:26] the auto selection uses the ids files in /usr/share/xserver-xorg/pci [08:28] look at hw/xfree86/common/xf86AutoConfig.c :) [08:53] hrm, the intel vblank freezes are annoying [09:29] sigh, braindead check for xserver 1.6 in wacom [09:43] grr, I hate this drier [09:43] -ver [09:51] after poking configure it does recognize the headers are from xserver 1.6 but still leaves the variable undefined [10:13] Alexia_Death: I wonder how you managed to build wacom on jaunty :) [10:14] it took some time to make 0.8.2.2 build [10:14] tjaalton: I havent built 0.8.2-2... [10:14] ok [10:14] well someone has posted the patches there to make it build [10:14] tjaalton: I have selfpatched 0.8.2-1 [10:14] and if fails [10:14] *it [10:14] why it fails? [10:15] You got it building? [10:15] * Alexia_Death is trying now [10:15] first because it only checks for xserver 1.6.0 (which doesn't exist), and when I make it to match 1.5.99 it still doesn't generate a correct src/include/xdrv-config.h [10:16] WCM_XORG_XSERVER_1_6 is left undefined [10:17] so how do I use your daemon? [10:19] http://a.death.pri.ee/watahod-0.3.2.tar.gz [10:19] I've got that [10:19] tjaalton: thats the latest version [10:19] some bugs fixed. [10:20] it contains an udev rules file. [10:20] that tho not mandatory should be used instead of the current wacom one. [10:21] tjaalton: it depends on python-gtk2 and python-notify. [10:21] then run watahod and plug in your tablet. [10:23] eraser starts with wrong bottomX bottomY, bug in wacom driver. You can set it to correct values with xsetwacom and then save defaults and every time you plug it in they are set correctly [10:24] I need to restart udev first? [10:25] reload, yes [10:25] usbParse: Exceeded channel count; ignoring the events. [10:25] It uses device names to remember prefs so if you change your rules later it wont pick up the prefs again [10:25] that's all I get in the logfile, maybe the driver is just buggy for me [10:26] thats a bug in the driver yes. [10:26] it was debugged in the list... [10:26] and there was even a patch for it... [10:26] but it wasnt passed publicly, ping just let someone try it out... [10:26] IRC [10:27] IIRC* [10:27] it should be included in 0.8.2.2 [10:27] what tablet do you have? [10:27] can you pass me a patch for your build fix? [10:28] I have a bamboo and an old inuos one serial tablet. neither have exhibited this behaviour... [10:28] aiptek 10000U, a waltop OEM model [10:29] and it works with wacom driver? [10:29] used to [10:30] It was recognized and loaded by the daemon? [10:30] only the first hunk of the patch is included in 0.8.2.2 .... [10:30] your daemon? yes [10:31] Ping hasnt been that good with this release... [10:32] Perhaps patching 0.8.2 with the crashfix patch and the other one gives better results. [10:32] the crashfix is in [10:32] doesn't crash X when I unplug it [10:32] Thats the broken define [10:33] that you fixed. [10:33] ah [10:34] It basically removes a double free. X insists now freeing somehing that the driver has already freed and assigning NULL to it did not get where it should have gotten to. [10:34] jot the driver just wont free it for now [10:36] doesn't work even with the other hunk [10:37] it worked with 0.8.2? [10:37] no, with 0.8.1.6 [10:38] hmm [10:38] I'll restart X to see if it helps [10:42] nope [10:42] hmm [10:42] Can I see your x log? [10:43] hmm, now even the previous version fails to work [10:43] and Id still like to have a build patch for the 0.8.2-2 if you can make one... [10:44] ? [10:44] I'll try to make one [10:45] but all you need to do is change the VERSION in configure to be 1, 5, 99 [10:45] and then change src/include/xdrv-config.h.in to #define WCM_XORG_XSERVER_1_6 [10:45] then it should build, at least the package [10:46] OK. Will try [10:48] 0.8.1.6 does work after all, I just needed to kill X first [10:48] because the new module was loaded [10:53] does it work hotplug for you? [10:53] or does it crash on remove? [10:54] it should crash. [10:54] tjaalton: but look up where that define is used, you should see the bad free immediately. [10:55] tjaalton: hey you uploaded my patches, I thought you had taken a new release from debian instead. [10:55] tjaalton: which is good :-) [10:55] * Alexia_Death is going to have to kill x to test the new module [10:55] brb [10:56] Alexia_Death: it does crash [10:56] tseliot: new release? [10:57] tjaalton: of synaptics [10:57] I thought you did another merge [10:58] there is no new release [10:58] Alexia_Death: is the crash caused by that xfree(priv); in wcmConfig.c? [10:58] yes [10:58] debian has 0.99.3-1 [10:58] tjaalton: well, I didn't check [10:58] Xserver tries to free that again [10:59] Alexia_Death: why don't you add something like if (priv != NULL) [10:59] there is. [11:00] it does not wotk [11:00] tseliot: Is I understand the problem, it checks if the priv elements pointer in driver controll structure is NULL [11:00] however, the priv in driver is just a local variable [11:00] so doing priv=NULL means nothing to the server. [11:01] and it tries to free it again. [11:01] tseliot: I can find you the bug about this in X bugzialla [11:03] this one, right? http://bugs.freedesktop.org/show_bug.cgi?id=19176 [11:03] Freedesktop bug 19176 in Server/general "CheckMotion for uninitialized devices segfaults the server" [Critical,Resolved: fixed] [11:03] Alexia_Death: http://users.tkk.fi/~tjaalton/foo/Xorg.0.log [11:05] ok, so we have a patch for X and a patch for the wacom driver [11:06] tseliot: in the end of that is the discussion yes [11:06] Alexia_Death: so doing "local->private =NULL;" should fix it in the driver [11:06] possible [11:07] I havent had time to investigate it much further [11:07] tjaalton: do we have the patch in comment 18 of the bug report? [11:08] if we do, then I can experiment with the wacom driver and fix this problem [11:09] tseliot: checkmotion patch is in [11:09] good [11:09] the buttons patch isnt and it does not apply right either :( [11:10] why? [11:10] Not on top of other patches in the source package anyway... [11:10] something conflicts.. [11:11] ok, I'll fix that too [11:11] I dont know why yet. [11:11] :) that would be great. [11:12] Alexia_Death: did you touch the udev rules to make sure that multiple devices are hotplugged? [11:12] ? [11:13] e.g. for the pen, pad, eraser, etc. [11:13] you said that you created a script which uses dbus [11:13] tseliot: the pack contains a new udev rules mainly to make sure that the wacom devices have sane disdinguishable names. [11:13] to make sure that not only the pen but also the other components work when you plug in the tablet [11:14] not directly. indirectly I need to know what devices a particulat tablet has. So currently theres some guessing based on name [11:14] did you create a package? [11:14] a deb package [11:14] tseliot: no, right now theres just a tar archive. [11:15] Its not stable enough to package yet [11:15] can you give me the link to it? [11:15] tseliot: http://a.death.pri.ee/watahod-0.3.2.tar.gz [11:16] tested yesterday on some arch users. [11:16] should more or less work. [11:17] tseliot: if you are already debugging the wacom driver, perhaps you can also fix the erasers bottomX and bottomY not being initzialized right. [11:18] on hotplug that is. its done right when added in conf AFAIK. [11:18] doesn't that depend on the specific tablet? [11:18] the values yes, but the tablet is capable of guessing them [11:18] driver that is. [11:18] it's likely "updated wcmUSB.c to ignore unparsed data" which broke it for me [11:19] yay for version control [11:19] it does it right on coldplug and for stylus in hotplug [11:19] or the nonexistence [11:19] *of it [11:19] ? [11:19] I've no idea what kind of a patch that was [11:19] or is there a public VCS repo somewhere? [11:20] with history and all [11:20] wacom has a a git repro somewhere [11:20] Ive built it once... [11:20] on git.debian.org, but it's not what ping uses [11:20] Alexia_Death: I would like to modify xsetwacom so that it can read settings from a configuration file, so that we can apply the settings once the device is hotplugged [11:21] http://git.debian.org/?p=collab-maint/linux-wacom.git;a=summary [11:21] tseliot: it works now, but having conf dump&load would be good. [11:21] ok [11:22] tseliot: look in the python code, its very simple. [11:22] tseliot: the daemon even mainains a conf directory under .wacom [11:22] maybe better use .config/wacom [11:23] tjaalton: why? [11:23] to not pollute ~/ [11:23] right about that... [11:23] the .xxx mess is annoying. [11:24] its just a path anyway. easy to change. but Im reluctant to do so untill there is no conf tool. [11:24] decent conftool. [11:24] wacomcpl does not count. [11:26] my tool will be written in Python and will write a configuration file which the C program will parse and apply [11:26] tjaalton: I have the url of the git that ping uses in my mail archives somewhere, but since kmail is not installable right now its abit hard to acccess. [11:27] tseliot: why not use the same format Ive used in my tool? [11:27] tseliot: lets just collaborate and make one thing? [11:27] daemon+conf tool. [11:27] Ive requested a sf project for it. [11:28] sf? [11:28] freedesktop? [11:28] but perhaps laouncpad would be better. [11:28] sourceforge [11:28] aah [11:28] sf is crap :) [11:28] * Alexia_Death shrugs [11:28] I like SVN:P [11:28] svn is too :) [11:28] yeah. git lovers. [11:28] dvcs ftw [11:29] launchpad should be good enough [11:29] and so is bzr (IMO) [11:29] Alexia_Death: the daemon should be written in C though [11:29] tseliot: why? [11:30] why is not python good enough? [11:30] not many people would like to see another daemon written in python [11:30] it does not do anything much [11:30] because they want something which doesn't slow down startup [11:30] it does not [11:31] and they are concerned about wasting resources [11:31] I love Python ;) [11:31] its too trivial a task to complicate it with C. [11:31] it needs to be flexible. [11:31] where should this daemon be run? in the user session? [11:31] tjaalton: yes [11:32] it needs to run in user rights. [11:33] this daemon only listends to dbus and reacts to some device add events... does nothing else. [11:33] but maybe not on every desktop [11:33] tjaalton: ? [11:34] I mean that the users should install it themselvers [11:34] -r [11:34] it can't be a dependency of wacom, since it's installed on every machine [11:34] Alexia_Death: people were not happy at the UDS when I talked about daemons in Python [11:34] tjaalton: not for wacom, but for wacom tools perhaps [11:34] tjaalton: or may be even completely different. [11:35] independent* [11:35] right [11:35] it does need xsetwacom [11:35] the driver only Suggests -tools, so it's not installed by default [11:35] xsetwacom is in tools? [11:36] yes [11:36] then tools depend on this daemon. [11:36] err [11:36] saemon depends on tools [11:36] * Alexia_Death is wee bit hung over today... [11:37] my sympathies :) [11:37] tseliot: Id agree about system daemons. this is a user daemon and very much optional. [11:37] hmm, maybe I'll take a break and do something else for a change -> [11:39] we already have a printer applet and the update notifier [11:39] which run as daemons and are written in Python [11:39] those are much more common. this one is just for wacom users. [11:40] tseliot: I do have a command line utility with all the same fucntions for people who do not want to run the daemon. [11:41] uses the same "lib" python file as the daemon. [11:41] Alexia_Death: do you suggest that we enable the daemon only after users configure their tablet in the UI? [11:41] yes [11:42] the daemon has no deep point unless theres a a tablet configured. It should be just a wizard that ends with registering the daemon for startup [11:43] and in the configuration utility there needs to be a way to remove it from startup easy. [11:43] ok, this is better than I thought [11:44] and people can even chose to run it manually when they are using a tablet and not on startup at all. [11:44] something like pressing a "detect" button in the UI? [11:45] or "load settings" [11:45] More like just selecting the daemon from the menu as a common app and running it. [11:45] those functions you listed are present in the daemon now [11:45] as menu items from the icon [11:46] Mostly for serial tablet [11:46] I have one connected through USB serial adapter. [11:46] Currently serial tablets are configured in a conf file. [11:47] I agree that this is good as a 'proof-of-concept' daemon, but in the long run there should be maybe only one user session daemon to handle input devices. frontends can be whatever the desktop requires [11:47] on load serial it checks if listed device nodes exist and loads them, if they do. [11:47] it could just be the generic configuration daemon of the desktop [11:47] which means extending the current one(s) [11:47] So the daemon runs without GUI and is driven by DBUS calls? [11:48] something like that, I don't know how it all works [11:48] I would like to see one daemon to deal with input devices and tablets [11:48] integration is the hard part, applies to everything [11:48] All tablets and input devices? [11:49] Alexia_Death: in the case of tablets we could use dbus [11:49] xinput for the rest of input devices [11:49] Alexia_Death: I meant that integration in general is what's lacking from the linux "desktop" [11:49] and of course the same should be done for outputs [11:49] tseliot: the problem is, only wacoms are nicely configurable through xsetwacom. [11:50] Alexia_Death: doesn't a C library for wacoms exist? [11:50] tjaalton: what does integration mean in this context? [11:51] why couldn't wacom support X input properties [11:51] tseliot: I dont know what you mean by c library... [11:51] I dont know those things... [11:51] Alexia_Death: a library which we could use for the daemon instead of having to call xsetwacom [11:51] Alexia_Death: integration with the existing desktop tools.. especially when it's not clear what the ultimate solution is [11:51] tseliot: I dont think so. [11:52] tjaalton: right now, I see a doable wacom daemon/conf tool. and even that is hard to do right because I just dont have the hardware to test it. [11:53] Alexia_Death: what hw are you missing? [11:53] tjaalton: tablets. theres tonns of tablets. [11:54] why would you need all of them? [11:54] tjaalton: not all of them. But there are some types. [11:54] tjaalton: like screen tablets. [11:54] which are serial [11:55] not all of them [11:55] * tseliot reboots [11:55] no, but most [11:55] I have a serial tablet and an usb tablet. [11:55] screen tablets have something that is totally unhandled gui wise, thats the callibration. [11:55] but how does a serial screen tablet differ from a serial handheld tablet? [11:55] ok [11:56] calibration. [11:56] well that's what tseliot is doing [11:56] tseliot has one to test? [11:56] a screen tablet? dunno [11:57] thats hard to do right If you dont have one. [11:58] sure [11:58] but now some LittleBigPlanet -> [11:58] :) [11:59] * Alexia_Death takes her hangover elsewhere too. [12:04] tseliot: do you have a screen tablet? [12:05] Alexia_Death: no, only a USB tablet [12:05] :( [12:06] but I know someone who owns 20 tablets [12:06] he could help us testing things [12:06] * Alexia_Death has serial and usb, but somebody with screentablet needs to do calibration [12:06] tats good [12:09] * tseliot nods [12:16] Alexia_Death: do you know what kind of patch system does wacom-tool use? [12:16] s/does// [12:16] xsewacom? its maintained in GIT afaik.' [12:16] s/use/uses/ [12:17] the debian package [12:22] dunno... [12:31] hehe "checking XInput extension version... >= 2.0" [12:31] I would like to see how it uses Xinput [13:38] tjaalton: any ideas on how to apply patches to wacom-tools? [14:05] Alexia_Death: setting local->private to NULL did the trick [14:05] X doesn't freeze if I plug in and then unplug my tablet [14:13] tseliot: there is no patch system afaik [14:13] tseliot: send patch to ping? [14:13] tjaalton: yes, I applied the patch manually [14:13] Alexia_Death: sure [14:14] what patch is this [14:14] ? [14:14] 0.8.2.2 doesn't crash on unplug [14:15] tjaalton: and is 0.8.2 in jaunty? [14:15] no [14:15] because it doesn't work for me, and I think it might break for others too [14:16] tjaalton: its the proper fix instead of the build switch. [14:16] I've got it packaged though [14:16] Alexia_Death: ok [14:16] tjaalton: it should be enough to make 8.1-6 not crash too [14:16] so you should get your tablet working. [14:16] sure, maybe just use that for now [14:17] jep. [14:17] tjaalton: what do you mean by "it doesn't work for me"? [14:17] tseliot: pastebinify the patch? [14:17] yes, let's stick with 8.1.6 [14:17] Alexia_Death: http://pastebin.com/d3a66dcb7 [14:17] tseliot: tjaalton has a tablet that has a problem with 0.8.2 [14:18] ok [14:18] Alexia_Death: now I would like to make sure that your patches for X apply [14:20] Alexia_Death: can you document your patches, please? [14:21] I just get a bunch of those "usbParse: Exceeded channel count; ignoring" -errors when the pen is close to the tablet [14:21] tseliot: the patch set I gave you is no longer needed. [14:21] with 0.8.2.2 that it [14:21] *is [14:21] tseliot: X went up a version. [14:21] Alexia_Death: aah, right [14:21] The patch that does not apply cleantly and does not is this: [14:23] https://bugs.freedesktop.org/show_bug.cgi?id=19282 [14:23] Freedesktop bug 19282 in Input/Core "Events handled wrong when master button map differs from the originating device..." [Normal,Resolved: fixed] [14:24] It should apply on clean rc1 but it does not apply on top of the already included patch set at least. [14:25] Alexia_Death: so are you saying that upstream has it already but we don't or what? [14:25] tseliot: yes [14:26] ok, so I'll make sure that this patch applies: https://bugs.freedesktop.org/attachment.cgi?id=21775 [14:27] yep. that should fix the buttons issue. [14:27] perhaps it's in master but not 1.6-branch [14:27] Let me know when you are done. [14:28] tjaalton: AFAIK it was cherrypicked on 1.6 a little after rc1 tag. [14:28] ok [14:28] Alexia_Death: nope, there are no commits after rc1 was tagged [14:28] hmm [14:28] then im wrong [14:29] its just in master then. [14:29] but it's on the list of proposed patches [14:29] http://www.x.org/wiki/Server16Branch [14:29] Its somewhat important tho. the stylus buttons are mapped wrong without it. [14:30] we'll get it with the next rc [14:30] yeah, but tablets are serious pita to use as is. [14:30] so is it useless if I get it to work now? [14:30] yeah, crashing and all :) [14:31] when is next rc due? [14:31] I wouldnt say useless... [14:31] dunno [14:31] no schedule? [14:32] no [14:32] maybe next week though, since LCA is over [14:33] tseliot: untill next week we can make due with the hack if we must ;) [14:50] hehe, now phoronix has picked up the intel "regressions" [14:51] tseliot: I looked at your patch... you have removed the free. [14:51] Alexia_Death: ok, the patch applies now [14:52] the free should stay if the NULL trick works. [14:52] tseliot: Great. Share? [14:53] Alexia_Death: ah, right. so I should add just add that line instead of replacing the xfree one with it [14:53] yep [14:53] Alexia_Death: it applies but I'm trying to build X now [14:53] applies != works [14:53] ;) [14:53] yes:D [15:14] hmm... 8.1-6 wont build for me [15:14] ok, X compiled [15:14] I've rewritten the patch for wacom [15:17] ok, wacom driver built. Now, I'll restart X to try it. Then I'll try the new (patched) X [15:18] * tseliot restarts X [15:21] wacom doesn't freeze X [15:21] \o/ [15:22] what version of wacom driver you have? [15:22] the one in jaunty [15:22] 8.1.6 [15:22] hmm [15:22] It wont build for me. [15:22] what's the error? [15:22] xf86Version.h is missing. [15:23] and some interface error [15:23] weird [15:23] I think my env is borked somehoe. [15:23] it could be [15:23] no, you just need to include the right headers [15:23] but I dont have clue how to fix it. [15:23]