[00:00] I can push in the stuff for non-semi-mt devices [00:00] It also depends on how bad the bad behaviour is. [00:00] though I guess I don't know that synaptics non-semi-mt don't do the same [00:00] it's not that bad of behavior [00:00] xorg-edgers? [00:00] it's just a little faster than expected [00:00] if I had to guess, somewhere between 1.5 and 2 times faster [00:00] and only when you hold the button down [00:00] so click+drag [00:01] hmm, that ratio sounds familiar [00:01] in fact, it's hardly a regression since click+drag didn't really work at all before :) [00:01] before this, clickpads would try to scroll if you did click+drag [00:02] which isn't really what the user wants [00:02] oh, there was an input bug with touchpads when used on HD monitors that the X speed would be different from Y by the aspect ratio of the HD monitor [00:03] bryceh: Yeah, that's the intended behaviour :) [00:06] RAOF, no, there was a period where it was mis-calculated. [00:06] Oh. I didn't notice that :) [00:08] cnd: Given its not a regression, the main concern would be that changing to the correct behaviour might be susprising. [00:08] heh, I don't think that's really a concern [00:08] clickpads are just broken broken broken on ubuntu right now [00:08] and this makes them not be :) [00:14] so it doesn't have anything to do with the X server's acceleration profile [00:14] whew [00:15] phew* [00:18] aha, I found the culprit [00:19] the estimate_delta function in synaptics [00:20] tbh, I think there's way too many independent acceleration, dampening, etc. functionalities through synaptics + xserver [00:20] and commenting out estimate_delta makes things behave normally [00:23] I have an nvidia gtx 560m... i get video tearing in all video players (incl. flash, mplayer) with compositing on, flash still tears with compositing off, mplayer is also seemingly dropping frames with compositing on, which is a lot more noticable when on battery (gpu clocks down to 202MHz gpu/324MHz ram) [00:24] and i NEVER had problems with the 8800gts on my desktop. what an insane regression. [00:25] any help? i've spent a lot of time trying to fix this... at the moment i'm running xorg-edgers, nvidia-current 295.17. [00:29] jschall: You've got the relevant “sync to vblank” things flipped? (In Compiz, in nvidia-settings) [00:29] RAOF: kubuntu... i have vsync on in kwin, in nvidia-settings [00:31] wonder if noveau will ever be as good as the nvidia drivers [00:32] nouveau* [00:35] Depends on what you mean by “as good” [00:37] well, doesn't seem to even allow kwin compositing to work [00:37] but maybe i didn't have a package i needed or something [00:37] The 560m is, I think, too new to get acceleration support. [00:37] At least in an Ubuntu release. [00:38] well, anyway, i'm happy with this laptop apart from this stupid video tearing thing... love to get it fixed [00:38] RAOF: any ideas at all? [00:38] RAOF: other than vsync switches? [00:39] Not really, sorry. [00:40] maybe it'll be fixed in a new nvidia driver some day... [00:41] Possibly ;) [00:41] RAOF: the juddering thing just seems weird, though... why wouldn't a modern video card be able to decode some video without dropping frames? [00:41] RAOF: the cpu can do it no problem... [00:42] RAOF: even though it's avc [00:45] Shrug. [00:45] It's not like we have much insight into the driver ;) [00:49] RAOF isn't the one to ask because he's a nouveau partisan [00:49] * RAOF prefers drivers we can fix, sure :) [00:49] is it just fullscreen thats tearing? there was a change recently in kde to stop unredirecting fullscreen windows which might be why it changed for you that you might be able to undo with a setting somewhere but yeah KDE, dont use or know anything about it [00:49] I also don't spend much time using the nvidia binary driver, so I'm particularly good at fixing problems with it. [00:51] Sarvatt: it does tear in fullscreen with or without compositing on. [00:51] i was going to suggest reporting it but apparently you did and they told you to come here, it really sounds like a kde/kwin specific problem though [00:51] err reporting it at nvnews.net forums [00:51] i don't think the 560m is too new [00:51] if the blob supports it it should get compositing and whatnot [00:51] bjsnider: will it run games with nouveau? [00:52] probably not as good as you want if you cared enough to get a good GPU :) [00:53] Sarvatt: does it have support for all the opengl extensions that the proprietary drivers provide? [00:53] No. [00:53] nowhere near it :) [00:53] Sarvatt: honestly the only game i really ever run any more is heroes of newerth... that's going to change when d3 comes out. [00:53] that's a laugh [00:54] you really dont want nouveau [00:54] hmm. k [00:54] i just don't want screen tearing [00:54] have you tried gnome? [00:54] there is an option in kwin to unredirect fullscreen [00:54] it's pretty good [00:54] * RAOF still contends that it's probably too new for acceleration under nouveau. [00:54] RAOF: it should be fine in precise [00:55] bjsnider: you know, i haven't tried gnome but i really just don't want to use it. i did run a livecd to check for tearing [00:55] did that have tearing too? [00:55] then again that was nouveau :) [00:55] yes but i think it would be interesting to see if there was tearing in gnome [00:55] it didn't seem to have tearing [00:56] but yeah, i guess it would've been nouveau [00:56] there's no tearing by design in mutter, so i would try gnome-shell [00:56] i can't remember if i installed the nvidia drivers [00:56] i guess i couldn't have in a live usb [00:57] Of course, there's no tearing by design in compiz either; that doesn't prevent tearing :) [00:57] sure there's tearing in compiz [00:58] you can turn off vsync and screw up the frame rate [00:59] but i'm an evil gnome-shell partisan so of course i'm going to recommend that [00:59] jschall: have you tried disabling gpu acceleration to see if its any different? [01:00] in whatever playback apps you're using [01:00] Sarvatt: well, mainly its flash [01:00] bjsnider: I just meant that vsyncing in the compositor is insufficient to ensure tear-free. (Indeed, I don't think it's possible to ensure tear-free) [01:01] Sarvatt: and it does have that option but it makes no difference [01:01] Sarvatt: i don't even think it does anything [01:01] ah gotcha [01:01] jschall, in which playre? [01:01] how about in smplayer? [01:02] bjsnider: that's flash [01:02] switching to gl, or xv, or x11 [01:02] jschall, are you talking about right-clicking? [01:02] Sarvatt: let me try it again [01:02] bjsnider: yes [01:02] it doesn't do anything [01:02] ah [01:03] that has been disabled in linux because it causes minor aggravations like taking down x [01:03] smplayer was abandoned [01:04] bjsnider: with xv if there's any tearing its unnoticeable in smplayer [01:04] bjsnider: only issue is decoding avc EATS cpu [01:04] xv is useless [01:04] bjsnider: why? [01:04] which version of ubuntu is this? [01:05] gl has tearing [01:05] and is kinda slowish [01:05] juddery [01:05] its not a lot of tearing though [01:06] gl (fast) seems better [01:06] go ask uau in #mplayer2 which output driver you should use [01:06] gl2 also juddery [01:06] what version of ubuntu? [01:06] and tearing [01:06] kubuntu 11.10 [01:07] -vo caca is perfect. [01:07] ok, well, there shouldn't be more than 30% cpu usage decoding avc, because we have ffmpeg-mt now [01:07] unless you're doing that 3d crap [01:08] 3d crap? [01:10] bjsnider: mplayer using 75% cpu decoding 1080p avc. cpu clocking to 2000mhz ish [01:10] bjsnider: with kwin compositing ("3d crap") disabled [01:10] bjsnider: that's with xv out [01:10] no, 3d video i meant [01:10] 10-bit video [01:10] bjsnider: oh [01:10] i'm assuming this is standard 1080p [01:11] what cpu? [01:11] bjsnider: vdpau out clocks the cpu down to 1200 and mplayer uses 10-25% [01:11] bjsnider: i7 2670qm [01:11] are you kidding me? [01:11] bjsnider: no [01:11] this is absurd [01:11] these numbers can't be correct [01:12] bjsnider: they are [01:12] you should get 5% with vdpau, and around 30% with ffmpeg-mt [01:12] bjsnider: as reported by i7z and top [01:12] is this just kde suckage? [01:12] bjsnider: no. [01:12] bjsnider: that's how much cpu it takes to decode avc [01:12] i've got an nvidia card that's crap compared to yours and i do much better [01:13] what about 1080p x264? [01:13] bjsnider: dunno if i have any [01:13] bjsnider: i can run mplayer with no output i assume [01:14] that's alright, just search for and download big buck bunny [01:14] what is the mplayer command you're using? [01:16] bjsnider: mplayer -vo null filename [01:16] bjsnider: 80% usage at 1.8-2.4ghz [01:16] bjsnider: kde not being used at all [01:16] bjsnider: compositing still off [01:17] mplayer -vo vdpau -vc ffvc1 -ao pulse file [01:18] Cannot find codec matching selected -vo and video format 0x31637661. [01:18] and then just the audio plays [01:18] mplayer -vo vdpau -vc ffh264 -ao pulse file [01:19] that's the one it uses normally [01:19] maybe i have an h264 file and not an avc file [01:19] no, it's the same thing [01:19] what's the cpu at? [01:20] bjsnider: 80% [01:20] are these actual blurays? [01:21] bjsnider: no [01:21] bjsnider: probably ripped from them though [01:21] alright, download big buck bunny and play that [01:22] give me the url and i will play it too [01:22] then we can compare [01:22] bjsnider: mmm firefly bluray rips [01:23] big buck bunny won't be avc so won't eat cpu [01:23] which version of bbb should i download? [01:23] mp4, h264, ogg? [01:24] h264 1080p [01:26] yeah, 40-60% and 1500-1800mhz [01:26] which comes out to like 20% at 3400mhz [01:26] which is what it can clock up to if it wants [01:27] what i should really be doing is power consumption tests [01:27] because that's what i care about if i'm going to watch on battery [01:27] anyway, something i noticed with mplayer -vo vdpau is that it didn't tear with compositing on... [01:27] so maybe its an smplayer thing [01:28] ok, try -vc ffh264vdpau [01:28] oh [01:29] bjsnider: yeah usual slowness and tearing [01:29] bjsnider: but low cpu usage! [01:29] how low? [01:29] 5%? [01:31] bjsnider: 15 [01:31] bjsnider: oh and... [01:32] it should be around 5% maximum, so i expect this is a kwin issue [01:32] and watching it that way is the way to get around power issues [01:32] bjsnider: and like 1ghz [01:32] bjsnider: so at 3ghz it'd be 5% [01:32] ok [01:32] bjsnider: well lets find out watts up [01:33] bjsnider: =P [01:33] bjsnider: pulled battery, plugged into watt meter [01:33] bjsnider: 66.5 watts with this [01:33] with vdpau? [01:33] bjsnider: 78w with no vdpau on -vc [01:34] 66.5 with vdpau [01:34] with no vdpau can jump to 80 [01:34] to be fair you'd have to run the cpu to full throttle [01:34] bjsnider: 72 with -vo xv and no vdpau in the codec [01:36] bjsnider: vdpau is the way to go for watts [01:36] bjsnider: for sure [01:36] we've known that for years [01:36] it's no surprise [01:36] but playback should be smooth and relatively judder-free [01:37] what's your screen's refresh rate? [01:37] bjsnider: but wait, different results with compositing off: vdpau uses MORE power than xv [01:37] bjsnider: wait, no [01:37] vdpau is superior to xv though [01:37] bjsnider: ran the test again, didn't give the meter enough time i guess [01:38] bjsnider: and now akonadi is doing something that uses 40% cpu... [01:38] bjsnider: oop, it just stopped [01:38] bjsnider: time to put battery back so i don't forget it's not there [01:39] bjsnider: anyway, interesting experiment [01:39] bjsnider: vdpau just sucks with compositing [01:39] constantly throttling the cpu down to the level of an old pentium isn't going to help performance [01:39] bjsnider: and flash just always sucks no matter what [01:39] bjsnider: umm, it works great... [01:40] bjsnider: it throttles it when it needs to [01:40] bjsnider: its a laptop [01:40] keep in mind that vdpau and compositing both use the gpu, at the same time [01:40] bjsnider: if i turned throttling off it'd eat a solid 40 more watts [01:40] bjsnider: yeah, but so does running heroes of newerth (should use a LOT more gpu than vdpau) [01:40] bjsnider: and heroes of newerth runs like 200fps if i turn vsync off [01:41] i doubt it [01:41] h264 is awfully hard on any cpu/gpu [01:41] bjsnider: why can my phone play it then? [01:41] if we didn't have ffmpeg-mt your sandybridge cpu couldn't play it [01:41] bjsnider: my phone plays sintel on youtube with no tearing just fine, but my $1400 laptop can't, apparently [01:42] bjsnider: but then, my phone isn't playing it at 1080p [01:42] bjsnider: and my phone has hardware decoding [01:43] my q6600 kentsfield cannot play 1080p consistently without multithreading [01:43] bjsnider: my q6600/8800gts played all the content i wanted to consistently, but i had it clocked at 3.4ghz... [01:44] bjsnider: 2.8ghz was the lowest i ever had it clocked to over a long period of time... [01:44] bjsnider: and it played anything i wanted perfectly with kwin compositing on [01:45] bjsnider: and i don't think the 8800gts was doing vdpau... if so i wasn't using it because i always used vlc [01:46] anyway i have to do my precalc homework for tomorrow [01:47] wonder if i can sell a netbook with linux on it on craigslist [01:47] because there's basically no way to reload win7 starter on it... [02:25] bryceh, RAOF: I ran out of time to get to the X uploads today [02:26] I don't have anything big planned tomorrow, so I should be able to get to them [02:26] cnd, ok [02:27] heh, I had a system with two dual head monitors, hot swapped them for two new monitors (different brand but same resolution), rebooted and it came up mirrored. Obviously. [02:53] bryceh, RAOF, tjaalton, Sarvatt: I just sent a couple patch series out to xorg-devel [02:53] they enable proper clickpad support [02:54] I want to get this into precise [02:54] I would suggest putting everything up through the first series (the synaptics touch handling) into precise asap (i.e. tomorrow) [02:55] I can put the second series into a ppa and put out a CFT [02:55] cnd, you sound a lot more confident than earlier! :-) [02:55] bryceh, I was afraid the motion issue was unresolvable [02:55] but it turns out to just be a bad algo [02:56] * bryceh nods [02:56] before I put out a CFT, if you could smoke test it I would appreciate it [02:56] git://people.freedesktop.org/~cndougla/xf86-input-synaptics branch clickpad-synaptics-ubuntu [02:57] bryceh, it should help your netbook [02:57] heh, the one computer I *haven't* updated so far today [02:57] and I bet the four of you have a few clickpads around [02:57] heh [02:57] cnd, ok will do [02:58] seriously my fibre has been on fire today [02:58] note that you will need to manually set the "Synaptics ClickPad" property to 1 [02:58] in most cases [03:01] It's not possible to detect that easily? Bah. [03:02] RAOF, it is, if the kernel driver tells us [03:02] I need to add it to the magic trackpad driver [03:02] Fair enough. [03:02] but I don't know what the state of it all is for synaptics [03:02] the detection logic is in the synaptics patch set [03:03] it's just not likely that people will have drivers that tell us they're clickpads [03:04] RAOF, actually, maybe most synaptics clickpads will work OOTB [03:04] I see support for setting the clickpad property in the kernel driver [03:04] I think it's just my specific trackpad on my older dell netbook that doesn't appear to the driver to be a clickpad [03:05] ok, off to get dinner [03:14] cnd, make check fails [03:14] make[2]: Entering directory `/home/bryce/ubuntu/xserver-xorg-input-synaptics/clickpad-synaptics-ubuntu/test' [03:14] /bin/bash: line 5: 28107 Segmentation fault (core dumped) ${dir}$tst [03:14] FAIL: eventcomm-test [03:31] bryceh, hmm. interesting [03:32] That's a symbol-mismatch, right? [03:33] RAOF, bryceh, no, it's an issue with the new mechanism for instantiating SynapticsHwState [03:33] I missed that the test program needs to be updated [03:35] bryceh, thanks for catching it, I'll fix it tomorrow [03:35] it's not hard to fix, but not something I can fix in 30 seconds [03:36] no prob, I worked around it (just hacked out the test) [03:36] I've got it installed and will play with it this evening while watching Dexter [03:37] cnd, are there any lp#'s associated with this work? [03:38] well I already see a problem [03:39] cnd, with one finger moving, if a second finger happens to touch the touchpad, the pointer stops dead [03:40] hmm, including if second finger is just carefully touching the left mouse button. IOW drag and drop becomes impossible. [03:40] so.. no dexter for the netbook [03:41] bryceh, no lp#s [03:41] bryceh, drag and drop should work... [03:41] that's the point of the clickpad support [03:42] check to make sure the "Synaptics ClickPad" property is set to 1 [03:42] oh, right! [03:44] bryceh: testing to see if I'm connected on my iPad [03:45] do you see this? [03:45] Yes. [03:45] k [03:49] bryceh: did setting the property fix things? [03:52] cnd, no, mouse cursor doesn't work at all [03:52] cnd, after setting the property in xorg.conf.d [03:53] bryceh: how did you set the property? [03:54] I suggest using xinput as it's faster to test [03:54] ah [03:54] well, wife's calling me to dinner so I'm out of time [03:54] if you do xinput list-props it will show it [03:54] and set-prop to change it [03:55] thanks for trying though... I hope it works better when you try again [03:57] cnd, alright lets try [03:57] Synaptics ClickPad (266) 1 [03:57] still, same behavior [04:00] hmmm... [04:00] so when you put two touches down, the pointer stops moving [04:00] right [04:01] when you press a button, it still doesn't move? [04:01] I've put packages up of what I used at https://launchpad.net/~bryce/+archive/clickpad-synaptics-ubuntu [04:01] when two touches are down but the button isn't pressed, it will emit scroll events (if enabled) or do nothing [04:01] cnd, yes if I depress the button it will begin moving again [04:02] if I just touch the button, it stops [04:02] bryceh, ahh, so click and drag works, right? [04:02] anyway, gotta go [04:02] k === yofel_ is now known as yofel [07:27] syncing pixman [07:27] ..later [07:27] not possible yet it seems === Daviey_ is now known as Daviey === bryceh is now known as bryce === inetpro_ is now known as inetpro [08:07] whoa, xorg dev position open? [08:11] not one but two.. === bryceh_ is now known as bryce === horstle_ is now known as horstle === seb128_ is now known as seb128 === Daviey_ is now known as Daviey === inetpro_ is now known as inetpro [09:50] amazing what a piece of hair in front of the mouse laser does to it's accuracy.. [11:30] RAOF: btw, flash 11.2 beta fixed my screen tearing [12:32] checking for glXCreateContext in -lGL... no [12:33] why the #&%#¤ does it fail on sbuild but not pbuilder [12:45] don't care anymore, ship it [12:45] (gstreamer-vaapi) === stgraber_ is now known as stgraber [15:37] whee, patch to fix scrolling from whot.. will test asap [15:50] cnd: whot posted a patch to fix bug 925785 and I verified it works [15:50] Launchpad bug 925785 in xorg-server (Ubuntu) (and 1 other project) "Starting to scroll is erratic with edge scrolling on touchpad or mouse scrollwheels (affects: 10) (heat: 50)" [Critical,Triaged] https://launchpad.net/bugs/925785 [15:50] tjaalton, ok, I'll be sure it's folded into the packages I upload today [15:51] cnd: I can add the patch, have it locally [15:52] tjaalton, if you don't mind, I'd rather add it [15:52] ok [15:52] sure :) [15:52] so I can keep track of what are patches vs what are merged from upstream [15:52] yeah this is pretty fresh (posted 20min ago), so probably not upstream yet [15:53] hasn't come across xorg-devel yet even :) [15:54] nope, took it from bugzilla [15:55] now to see if a newer kernel fixes nouveau on my laptop.. [15:56] it's failing to use nouveau dri, falling back on swrast [15:56] promising, plymouth splash [15:57] nah, "error creating GPU channel: -19" [15:58] ok, accel disabled by default [15:59] tjaalton, do you have a clickpad? [15:59] Sarvatt, you too? [15:59] cnd: no, just the thinkpad [15:59] ok [15:59] Sarvatt does [15:59] macbook air [16:11] this isnt a clickpad [16:11] oh [16:11] broadcom BCM5974 touchpad [16:11] what's a clickpad then? [16:11] actual synaptics touchpads that click [16:12] this just uses the synaptics driver [16:12] ok [16:12] i dont even have any systems with real synaptics touchpads in them anymore, crazy [17:13] Sarvatt, no, I'm meaning clickpad in the functional sense [17:14] so macbook trackpad counts [17:14] not in the Synaptics ClickPad (R) sense [17:14] oh! let me read the scrollback and see what ya asked [17:14] Sarvatt, install the package from https://launchpad.net/~bryce/+archive/clickpad-synaptics-ubuntu [17:15] then, use xinput to set the "Synaptics ClickPad" property to 1 if it's not already [17:15] then see if you can drag and drop with two fingers [17:15] and if everything else seems sane [17:16] cnd, from last night, yes click and drag worked. Anything else need tested? [17:16] cnd: going to invalidate any testing if i use it against 1.12? [17:16] ah it doesn't even build against 1.12 [17:16] damn, gonna take me about 45 minutes to remove edgers [17:16] Sarvatt, you can fetch from git and build from sources [17:16] http://cgit.freedesktop.org/~cndougla/xf86-input-synaptics [17:17] i did last night and it failed the same way [17:17] branch clickpad-synaptics [17:17] oh clickpad-synaptics, i used the ubuntu one [17:17] yeah, the ubuntu one has two patches for the 1.11 backport [17:17] cnd, what's the cleanest way to install (and uninstall) it if you build from source? [17:17] bryce, I usually use --prefix=/usr [17:17] that way when I reinstall the dpkg everything is overwritten [17:18] bryce, does the trackpad now behave as you expect? [17:18] any issues? [17:18] my hope is this makes these devices much saner [17:19] cnd, well there's the issue that when both fingers are on the pad, the pointer does not move [17:19] bryce, that's expected [17:19] it will either perform scrolling or do nothing [17:19] depending on your scroll options [17:19] or send touch events [17:19] cnd, ah, there's a way to configure that? [17:20] like, a way to get it to just ignore the second mouse touch? [17:20] bryce, in gtk mouse settings you can enable two-touch scrolling [17:20]