**** BEGIN LOGGING AT Mon Feb 28 02:59:57 2011 Feb 28 04:41:05 the one I use with the tab is still an android butchered kern Feb 28 04:41:13 but too much is broke Feb 28 04:41:25 and I have to start rolling back some of the android patches Feb 28 04:41:35 which is proving rather troublesome Feb 28 04:51:23 it all depends on the quality of the changes Feb 28 04:51:50 if it's not breaking stuff at least, you can still use a config compatible with ubuntu Feb 28 04:51:59 with the needed modules and stuff Feb 28 09:50:20 hrm Feb 28 09:50:38 no images :( Feb 28 09:51:15 bug 721118 Feb 28 09:51:17 Launchpad bug 721118 in unity "Unity FTBFS on armel due to Nux" [High,Fix committed] https://launchpad.net/bugs/721118 Feb 28 10:04:43 morning Feb 28 10:46:44 hello Feb 28 10:47:04 I'm using the natty version on BeagleBoard xM and it's slow Feb 28 10:47:15 is the proc frequency 600 MHz only ? Feb 28 10:47:55 second, looks like there is no 3D acceleration. Is there something possible, or .. Feb 28 10:47:58 thanks in advance :) Feb 28 10:52:43 i think there is some cmdline option you can set to make it operate on full speed Feb 28 10:54:13 ogra: I believed cpufreq-selector could help, but the frequency management is not handled (if I searched correctly) Feb 28 10:54:16 and the sgx drivers should be in multiverse somewhere Feb 28 10:54:30 ogra: ah, thanks. was the name missing me Feb 28 10:54:32 sgx Feb 28 11:03:39 I think you do mpurate= to get itr faster Feb 28 11:03:51 XorA_: at boot time ? Feb 28 11:04:00 ericb2: yes Feb 28 11:04:13 XorA_: ok. and what number are possibles ? Feb 28 11:04:33 numbers Feb 28 11:04:35 i think it takes MHz values Feb 28 11:04:48 so 1000 should get you a GHz Feb 28 11:04:50 ok, I'll try 800 then Feb 28 11:05:07 800 certainly safe if thats the right arg, Im working from memory Feb 28 11:05:24 * XorA_ only got an XM yesterday Feb 28 11:05:28 * ogra imagines 1000 should work too, buut 800 is surely safer Feb 28 11:05:29 ok. anyway, we'll see Feb 28 11:06:20 is there a key to hit at boot ? Feb 28 11:06:44 I think koen was running XM happilly at 1.2Ghz Feb 28 11:06:48 no, you edit /boot/boot.script and run sudo flash-kernel Feb 28 11:06:59 on the running system Feb 28 11:07:24 that will change your boot args Feb 28 11:08:23 ogra: thanks Feb 28 11:08:45 * XorA_ shall Natty his XM tonight Feb 28 11:08:50 heh Feb 28 11:09:32 i think we shopuld probably set mpurate at image buildtime Feb 28 11:10:14 ogra, ping Feb 28 11:10:22 the prob is that these images also run on B and C series beagles and we cant really determine if we are on a XM Feb 28 11:10:40 janimo, yo Feb 28 11:10:52 the cpufreq patches should be mainstream now, and they know what type of omap you have Feb 28 11:11:24 janimo, in bug 724615 there is a "fix committed" set, could you please link the branch in the future ? (desktop team asked) Feb 28 11:11:25 Launchpad bug 724615 in unity "unity FTBFS on armel" [High,Fix committed] https://launchpad.net/bugs/724615 Feb 28 11:11:37 but last I saw there were some bugs in cpufreq framework preventing going to full speed as it assume unique voltages for each frequency Feb 28 11:11:46 that way its easier to find what the patch was :) Feb 28 11:11:52 XorA_ I got an xM rev 1 if I'm not wrong Feb 28 11:13:12 dont know which rev I got Feb 28 11:13:22 no sticker ? Feb 28 11:13:34 ogra: its at home, my eyesight not that good Feb 28 11:13:44 train it ! Feb 28 11:13:49 someone sold it to me cheap over the weekend Feb 28 11:14:15 * XorA_ never persuaded TI to part with any :-D Feb 28 11:15:34 shit I tried to install libgl1-sgx-omap3 libgles2-sgx-omap3 and it fails. Feb 28 11:15:52 the reason is : FATAL: module omaplfb not found Feb 28 11:16:31 you need the powervr package first i think Feb 28 11:16:54 powervr-omap3-dkms Feb 28 11:17:04 * ericb2 tries Feb 28 11:17:48 it is already installed Feb 28 11:18:36 Building ... Feb 28 11:18:42 is it done compiling already ? Feb 28 11:18:44 aha Feb 28 11:19:07 Error! Bad return status for module build on kernel: 2.6.38-1-omap (armel) Feb 28 11:19:18 I'll consult the make.log Feb 28 11:19:43 is that your own kernel ? Feb 28 11:19:53 ogra: no it isn't Feb 28 11:20:13 error : unknown fiels 'ioctl' specified in initializer Feb 28 11:20:15 hmm, strange Feb 28 11:20:27 s/fiels/field/ Feb 28 11:20:49 there might be a ppa where rsalveti provides an update package, wait until he is around Feb 28 11:21:16 in build/services4/3rdparty/bufferclass_ti/bc_cat.c Feb 28 11:21:18 though i'm not sure he rolled a new one for omap3 ... i know there is one for omap4 somewhere Feb 28 11:21:32 ogra, there was no branch I committed to trunk Feb 28 11:22:06 there's a ton of ubnity email since I am on the team, so I would not be surprised commit messages to be overlooked Feb 28 11:22:19 janimo, then make a branch first and link it or attach a patch or some such Feb 28 11:22:37 its just very hard to find out about the fix if nothing is linked Feb 28 11:22:57 you can even just copy the link to the revision into a comment Feb 28 11:23:44 ok, I was hoping that bzr commit would do something as I inlucde dLP :#XXX in the commit msg Feb 28 11:23:54 hmm Feb 28 11:24:17 is this on ayatana or u-desktop? Feb 28 11:24:19 i think it should ... i rarely cimmit to upstream branches directly though, i usually link merge branches Feb 28 11:24:33 the discussion was on u-desktop Feb 28 12:24:35 ogra: is it possible to compile by hand ? Feb 28 12:25:03 look up the dkms docs, i think you can trigger it by hand Feb 28 12:25:42 ogra: I'm currently in /var/lib/dkms/pwervr-omap3/3.01.00.07/build/services4/3rdparty/bufferclass_ti, but make leads me to undefined $KERNEL_DIR Feb 28 12:26:09 * ericb2 looks Feb 28 12:26:44 hmm, really looks like you dont use the official kernel Feb 28 12:27:11 ogra: no idea : Im running the Ubuntu image I found the link on the Ubuntu wiki Feb 28 12:27:27 ogra: this is natty, could be related ? Feb 28 12:27:47 linux-image-2.6.38-5-omap Feb 28 12:27:55 thats the version you should have installed Feb 28 12:28:26 ogra: uname -a tells : Linux beagle 2.6.38-1-omap Feb 28 12:28:35 ogra: uname -a tells : Linux beagle 2.6.38-1-omap #28-Ubuntu Feb 28 12:28:58 that sounds outdated Feb 28 12:29:28 indeed, apt-cache tells me about 2.6.38-5 Feb 28 12:32:08 shit .. no network Feb 28 12:32:31 is /etc/init.d/networking restart wrong ? Feb 28 12:33:10 that only restarts low-level bits iirc Feb 28 12:33:24 network manager cares for the higher level bits Feb 28 12:33:43 ogra: and I need to got X.org running for that ? Feb 28 12:34:24 you can also set it up in /etc/network/interfaces instead Feb 28 12:34:36 then NM will use the config from there without X Feb 28 12:41:30 I added two lines in /etc/network interfaces : allow-hotplug usb0 and iface usb0 inet dhcp Feb 28 12:41:44 and ifup usb0 fails Feb 28 12:41:45 and auto usb0 ? Feb 28 12:41:57 you shouldnt need allow-hotplug Feb 28 12:42:12 cat /proc/net/dev|grep usb ? Feb 28 12:43:52 * ericb2 removed allow-hotplug an retries ... Feb 28 12:45:34 ogra: got : usb0: 64009 677 .... and so on Feb 28 12:45:44 looks fine Feb 28 12:46:17 ok, works now. Suddenly ... Feb 28 12:46:53 its magic :) Feb 28 12:48:10 ogra: ufff ... loosk like I'm in debug mode ... got zillions of [ xxxx.xxxxxx] smsc95xx 1-2.1:1.0: usb0: kevent 2 may have been dropped Feb 28 12:48:48 * ericb2 installing kernel 2.6.38-5 Feb 28 12:49:48 (omap version) Feb 28 12:53:28 * ericb2 hopes to be able to work a bit today Feb 28 13:03:48 morning Feb 28 13:06:00 rsalveti: hi Feb 28 13:06:16 rsalveti: got a problem installing powervr on omap3 Feb 28 13:06:21 rsalveti: X.org is gone ... Feb 28 13:06:30 rsalveti: + the build fails Feb 28 13:06:45 ericb2: could be that sgx is broken with latest updates Feb 28 13:06:57 ericb2: can you open a bug against it? Feb 28 13:07:02 will take a look then Feb 28 13:07:06 rsalveti: where ? Feb 28 13:07:13 rsalveti: do you have a link I meant Feb 28 13:07:18 maybe by updating the package or by fixing the current code Feb 28 13:07:34 ericb2: just open a bug agains the sgx package for omap3 Feb 28 13:07:44 putting your dkms logs Feb 28 13:07:58 ericb2: there's a workaround to avoid the smsc95xx messages Feb 28 13:08:07 let me get the link Feb 28 13:08:53 XorA: ogra: and for omap I'd like to merge the patch that sets the mhz higher for most beagle boards Feb 28 13:09:45 also on my todo, was waiting more discussion and see if it'd go for 39 Feb 28 13:09:54 than I can easily backport for our current 38 Feb 28 13:10:36 does anyone know how long i can expect to wait for the image writing to complete? I'm following instructions here for a beagle board install, https://wiki.ubuntu.com/ARM/OMAPMaverickInstall Feb 28 13:12:26 ericb2: for smsc messages: Add to /etc/sysctl.conf: vm.min_free_kbytes = 8192 Feb 28 13:12:47 rsalveti: ok, I'll do once I'll have installed kernel-sources for 2.6.38-5 Feb 28 13:12:56 ericb2: thanks Feb 28 13:14:41 btw, mpurate=800 seems to work well. mpurate=1000 does nothing, and looks like 600 MHz is the fallback Feb 28 13:15:56 * ericb2 doing everything on a simple 16GB micro SD card Feb 28 13:16:23 k Feb 28 13:16:33 would be fantastic to find 32 or even 64GB a day, using such "size" :) Feb 28 13:16:57 1cm square hard disk is a bit of fun Feb 28 13:17:16 crap just reminded me I dont have any high speed microSDs Feb 28 13:35:25 rsalveti: done. ( /etc/sysctl.conf ..etc) Feb 28 13:46:36 rsalveti: https://bugs.launchpad.net/ubuntu/+source/powervr-omap3/+bug/726541 Feb 28 13:46:38 Launchpad bug 726541 in powervr-omap3 "powervp installation broken on aremel (kernel 2.6.38-5)" [Undecided,New] Feb 28 13:48:34 ericb2: cool, thanks Feb 28 13:49:38 rsalveti: I added one comment Feb 28 13:50:54 rsalveti: wiating, what can I do to repair my X.org (currently broken) Feb 28 13:50:54 ? Feb 28 13:51:02 s/wiating/waiting/ Feb 28 13:51:24 ericb2: what happened to your x.org? Feb 28 13:51:51 something like "module omaplfb not found" or something close Feb 28 13:52:12 rsalveti: happened when I tried to install powervr Feb 28 13:53:09 and apt-cache search omaplfb ... is empty :/ Feb 28 13:53:28 that's because this module is loaded during the sgx initscript Feb 28 13:53:36 but it shouldn't affect your xorg Feb 28 13:53:58 as the omap 3 driver doesn't properly support the xorg Feb 28 13:54:11 rsalveti: where is the config file located ? Maybe I can reconfigure X.org ? Feb 28 13:55:50 ericb2: there's no xorg conf by default Feb 28 13:56:13 rsalveti: sorry, how does it work then ? Feb 28 13:56:19 ericb2: at /usr/share/X11/xorg.conf.d are the basic confs, but not something that sets your video driver Feb 28 13:56:43 you can still create your own config, but not needed Feb 28 13:56:59 as you're basically using xorg with framebuffer Feb 28 13:58:06 rsalveti: ok, and how can I figure out what happens ? The only info I got is : ddxSigGiveUp : Closing log Feb 28 13:58:21 in /var/log/Xorg.0.log Feb 28 13:58:26 ericb2: can you post your x.org log for me? Feb 28 14:00:16 rsalveti: yes, sure. Let me try to launch it manually Feb 28 14:01:08 (EE} HID 04f3:013: failed to initialize for relative axes. Feb 28 14:02:25 could the mpurate=800 be the reason ? Feb 28 14:06:47 and servce gdm restart does notthing ... tested on CTRL + ALT + F6 -> F10 Feb 28 14:09:10 if xorg is failing, then gdm restart will not work Feb 28 14:09:26 rsalveti: I know. Was to obtain a track, somewhere to search Feb 28 14:19:02 rsalveti: the only track I got, is when I installed the libegl1-sgx-omap3 lib, a lot of libs have been uninstalled Feb 28 14:19:45 rsalveti: bingo Feb 28 14:19:53 rsalveti: broken dependencies Feb 28 14:21:00 it's normal to remove some mesa libraries and dependencies, as the sgx one will replace them Feb 28 14:21:19 but it could be that something was remove accidentally Feb 28 14:21:23 *removed Feb 28 14:21:33 I'm setting up my beagleboard to test the package again Feb 28 14:21:38 and try to fix it Feb 28 14:21:43 rsalveti: you mean everyting removed + omaplfb module should replace everything ? Feb 28 14:22:08 * ericb2 will redo .. not clear Feb 28 14:22:13 no, the sgx packages should replace the libegl/libgles from mesa Feb 28 14:22:26 and when the mesa packages are removed, some others are also removed, as not used anymore Feb 28 14:22:38 rsalveti: ok, but sgx package won't work without the module ? Feb 28 14:22:48 ericb2: nops Feb 28 14:23:00 rsalveti: ah .. sorry looks I'm plain wrong Feb 28 14:23:01 but your xorg should still work Feb 28 14:24:16 rsalveti: it's not the case Feb 28 14:31:12 rsalveti: Did you manage to give that rootfs a stab? Feb 28 14:31:43 lag: argh, sorry, it took so long to download that I completely forgot Feb 28 14:31:45 let me dig it here Feb 28 14:32:45 :) Feb 28 14:32:47 np Feb 28 15:18:51 lag: did you try this rootfs with kernel upstream? Feb 28 15:19:03 or how are you planning to use it? Feb 28 15:19:49 rsalveti: I have various kernels I can try to use it with Feb 28 15:20:01 rsalveti: The main one is on git.linaro.org Feb 28 15:20:08 lag: so what exactly you want me to test with it? Feb 28 15:20:09 rsalveti: I'm guessing you don't have HW though Feb 28 15:20:23 Anything? Feb 28 15:20:26 just to try it with beagle? Feb 28 15:20:28 ok Feb 28 15:20:29 rsalveti: I think it's just a generic kernel Feb 28 15:20:32 Sure Feb 28 15:39:45 hey prpplague Feb 28 15:39:56 XorA_: hey bud Feb 28 15:41:17 prpplague: what happens if I plug a zippy into beagle XM? Feb 28 15:41:45 XorA: it goes *poof* Feb 28 15:41:52 * prpplague jokes with XorA Feb 28 15:41:53 :-( Feb 28 15:42:14 XorA: zippy and zippy2 have been tested with xm with no issues Feb 28 15:42:34 prpplague: sweet, even bigger rootfs with LVM then :-D Feb 28 17:54:58 lag: image is working fine Feb 28 17:55:06 lag: with our 38 for omap Feb 28 17:55:16 linaro image but still "Welcome to Ubuntu Natty" :-) Feb 28 18:17:51 rsalveti: It's the Linaro image of Ubuntu Feb 28 18:18:25 heh Feb 28 18:18:34 "the linaro image of ubuntu" Feb 28 18:18:57 sounds like "the duke of earl" Feb 28 18:21:59 rsalveti, do TI plan a .38 upload now that it seems to be close to good enough? Feb 28 18:23:28 TI ... upload ... ? Feb 28 18:23:51 janimo, cooloney does our kernel, whenever he is ready we should get a package Feb 28 18:24:16 given that there are only a few hours until freeze i wouldnt expect it before alpha3 though Feb 28 18:24:27 ogra, ah great. Well, green light for upload, pull request whatever :) Feb 28 18:24:45 it's a soft freeze :) Feb 28 18:24:52 ogra talking bad about TI *g* Feb 28 18:24:55 k, i'll try to ping him tomorrow morning Feb 28 18:24:56 and it does not go into images so should not destabilize A3 Feb 28 18:25:20 we wont have parallel kernel packages for omap4 Feb 28 18:25:34 ah, so it's a straigh update Feb 28 18:25:35 so it has to be good enough to replace the existing one Feb 28 18:25:39 ok, then Feb 28 18:25:39 right Feb 28 18:25:52 kernel team refused to maintain two packages Feb 28 18:26:00 makes perfect sense Feb 28 18:26:03 specially since we would get naming probs Feb 28 18:26:16 and cross-grading would become a pain Feb 28 18:26:43 ogra, what were the takeaways from the emdebian meeting? Feb 28 18:26:48 anything new to us? Feb 28 18:27:02 janimo: ogra: new .38 upload should be done when cooloney work on it Feb 28 18:27:13 but, still not tested Feb 28 18:27:15 rsalveti, right Feb 28 18:27:27 the pull request is done from TI side already Feb 28 18:27:31 i think janimo just said it was good for him above Feb 28 18:27:35 ok Feb 28 18:27:36 yup, i saw the mail Feb 28 18:27:49 the sound stuff doesnt look convincing though Feb 28 18:28:02 but we have time to fix that this time (i hope) Feb 28 18:28:08 yeah Feb 28 18:28:25 I believe we'll go a3 with the kernel we have in hands Feb 28 18:28:35 this new on a ppa and after a3 we can do the switch Feb 28 18:28:44 yeah Feb 28 18:28:45 as the soft freeze for a3 is today Feb 28 18:28:52 wasn't expecting that Feb 28 18:28:57 freeze is at 23:00 UTC Feb 28 18:29:01 where is that .38 kernel coming from? Feb 28 18:29:02 yeah Feb 28 18:29:07 armin76: from TI Feb 28 18:29:13 its always on mondy evening/tuesday morning Feb 28 18:29:16 armin76: but basically the same one linaro currently delivers Feb 28 18:29:28 ogra: yeah, was expecting it to be tomorrow, but fine :-) Feb 28 18:29:37 nothing urgent Feb 28 18:29:42 well, looks like unity didnt make it yet Feb 28 18:29:42 rsalveti: link for it? Feb 28 18:29:48 armin76: getting for you Feb 28 18:29:50 so that looks like a delay Feb 28 18:29:56 thanks Feb 28 18:30:03 (for the freeze) Feb 28 18:30:06 ogra: what you mean? we may need some bugfixing Feb 28 18:30:14 but that's fine if we can do that tomorrow Feb 28 18:30:16 rsalveti, i know ... Feb 28 18:30:16 I believe Feb 28 18:30:24 we wont have images tomorrow Feb 28 18:30:33 ogra: didn't you say you fixed it? Feb 28 18:30:34 unity will be uploaded tomorrow morning only Feb 28 18:30:35 :-) Feb 28 18:30:40 finally KDE packages are out of the FTBFS queue, now the chart looks familiar again Feb 28 18:30:44 no,. janimo fixed it Feb 28 18:30:51 although I wish Libo built on arm for a change Feb 28 18:30:59 but the fix was sitting idle in upstreams bzr for three days Feb 28 18:31:01 oh, will still be pushed... Feb 28 18:31:08 it will be pushed Feb 28 18:31:18 argh Feb 28 18:31:23 but the desktop team just announced they wont make it before tomorrow morning Feb 28 18:31:41 which means we will have no images tomorrow morning Feb 28 18:31:47 yeah Feb 28 18:31:50 with luck we'll have them in the evening Feb 28 18:31:51 maybe I should have made a package upload as well. Dunno, I was trying not to mess with too many bzr branches at once Feb 28 18:32:00 yeah Feb 28 18:32:14 I thought you also did a package upload Feb 28 18:32:20 no Feb 28 18:32:25 janimo, given treh experience we now have with nux and unity slowness, i would actually recommed package fixes in the future Feb 28 18:32:28 I saw they upload quite frequently Feb 28 18:32:31 so I leave it to them Feb 28 18:32:32 it's taking too much time for the fixes to be in the archive Feb 28 18:32:41 nux took ages Feb 28 18:32:52 yeah Feb 28 18:32:55 because they kept back a new upstream relese until FF Feb 28 18:32:58 true, but I did not know how long it takes them. I will consider package uploads from now on though Feb 28 18:33:02 seems they don't care much about arm Feb 28 18:33:31 given the silly casting errors that cause our build failures i would agree with that :) Feb 28 18:33:32 if it's breaking our images, then a package upload would be better I guess Feb 28 18:33:42 rsalveti, I can't blame them. The flood of bugmails I am getting since on the unity team has * a lot* of crash reports Feb 28 18:33:55 they ned to make that stabl eon x86 before caring for arm Feb 28 18:33:57 crash or ftbfs? Feb 28 18:34:05 crashes Feb 28 18:34:11 crash is fine :P Feb 28 18:34:15 yeah Feb 28 18:34:17 I still canoot run unity on my two x86 laptops I test on Feb 28 18:34:18 ftbfs blocks images Feb 28 18:34:21 just goes away Feb 28 18:34:22 ftbfs is blocking Feb 28 18:34:55 ogra, rsalveti I know, what I am saying they are swamped by bugfix work so probably let us handle arm and only do it when explicitly pinged Feb 28 18:34:58 i'm really curious if unity-2d will run at all Feb 28 18:35:25 janimo, right, do package fixes and additionally dup a merge request in place for next upstream Feb 28 18:35:30 armin76: https://github.com/sebjan/linux-2.6/tree/int-2.6.38-rc6-iv3 Feb 28 18:35:31 *dump Feb 28 18:35:49 yeah Feb 28 18:36:03 otherwise we're not making a3 Feb 28 18:41:16 * armin76 wonders why is it hidden Feb 28 18:41:20 rsalveti: thanks Feb 28 18:41:33 armin76: it's not hidden, heavy dev Feb 28 18:41:45 armin76: if you follow linaro tree you'll probably get the same patch set Feb 28 18:42:07 arm guys! Feb 28 18:42:12 I have a quandry Feb 28 18:42:22 I want to install from ubuntu-standard a full gnome desktop Feb 28 18:42:34 but ubuntu-desktop requires unity Feb 28 18:42:41 and jockey-gtk Feb 28 18:42:43 rsalveti: actually, if you talk about my tag, it does not contain the Linaro patches, only TI ones on top of mainline kernel Feb 28 18:42:45 but unity depends on compiz Feb 28 18:42:54 and jockey-gtk depends on nvidia-common Feb 28 18:43:01 *what* is going on? Feb 28 18:43:17 (btw unity-2d doesn't provides: unity either) Feb 28 18:43:29 sebjan: true, I mean, if he gets the andy linaro tree he'll get everything Feb 28 18:43:47 rsalveti: correct Feb 28 18:43:48 as he's also merging your tree Feb 28 18:44:02 Neko unity is the default ubuntu-desktop in natty Feb 28 18:44:09 Neko: we're fixing this Feb 28 18:44:23 Neko, and unity-2d doesnt provide unity since it uses bits and pieces of unity Feb 28 18:44:32 -2d needs unity installed Feb 28 18:44:39 Neko. Packages are being rebuilt frm massive uploads last week. Expect a little turbulence in the pool, it should settle by the end of the week. Feb 28 18:44:44 right and I tried that but unity depends on compiz-abisomething Feb 28 18:44:46 jockey is fixed since half a day Feb 28 18:44:57 ah okay so I am just in the middle of an update? Feb 28 18:44:58 phew.. Feb 28 18:45:02 yeah Feb 28 18:45:03 Yes. Feb 28 18:45:04 the rest will be fixed tomorrow or so Feb 28 18:45:04 okay I am satisfied then Feb 28 18:45:12 Neko: we still don't have images Feb 28 18:45:15 since feb 16th Feb 28 18:45:21 depending how fast the desktop team works Feb 28 18:45:24 we expect this to be fixed this week Feb 28 18:45:36 will there be a way to install a maverick-ish gnome desktop without unity? Feb 28 18:45:39 we're waiting too Feb 28 18:45:51 not when using ubuntu-desktop Feb 28 18:46:12 but there's no new meta package to be back to the old behavior? Feb 28 18:46:30 ogra, I am not sure u-2d need unity installed Feb 28 18:46:31 I dunno. I like the IDEA of unity, just not the way it soaks up all the screen space with it's little menu bar Feb 28 18:46:37 what pieces it uses from it? Feb 28 18:46:54 janimo, the dash, icons etc Feb 28 18:47:23 I don't see those in the depends Feb 28 18:47:32 and iirc it build deps on libunity Feb 28 18:47:46 I think there is some common C++ helper indeed Feb 28 18:47:46 which should add it to debs through shlibs Feb 28 18:47:52 but not unity (the apps ) itself Feb 28 18:48:15 the point is that you can install all bits and pieces independently Feb 28 18:48:29 i.e. i used the panel on a normal gnome desktop for a while here ;( Feb 28 18:48:31 err Feb 28 18:48:32 ;) Feb 28 18:48:55 the unity package should probably have more recommends though Feb 28 18:49:14 err Feb 28 18:49:18 unity-2d indeed Feb 28 18:52:27 GrueMaster, if you have some spare cycles today it would be nice to know if unity-2d still runs ... nowing that in advance before we have images would be helpful Feb 28 18:52:53 I'm already downloading the updates now. Feb 28 18:53:07 * ogra hugs GrueMaster Feb 28 18:56:24 I'm kind of impressed with how fast natty is just as a running system... I hacked in gnome and ambience theme and the backdrops and it's like lightning Feb 28 18:56:36 unity seems to run okay it's just I had to spend 2 hours getting it to even install Feb 28 18:57:12 but I understand, we had this problem during Maverick... trying to do work during the day before and the day after an alpha release = eek :D Feb 28 18:59:13 Neko, when booting, you can select classic desktop when logging in. not sure if there are plans for a meta package. Feb 28 18:59:23 it's no problem Feb 28 18:59:37 it just scared me that I couldn't install ubuntu-desktop because it depends on unity and unity didn't work Feb 28 18:59:49 which means no desktop ata ll Feb 28 18:59:52 at all Feb 28 19:00:13 I'll try it again mid-March and it will hopefully Just Work (tm) :) Feb 28 19:00:25 you should try on thu. Feb 28 19:00:39 release day? I doubt I will have good luck then :D Feb 28 19:00:50 ?? Feb 28 19:01:02 we usually have frightening problems with bandwdith to the ubuntu servers Feb 28 19:01:03 this thursday Feb 28 19:01:12 Alpha 3 release date is Thursday Feb 28 19:01:16 yes Feb 28 19:01:26 hundreds of geeks downloading new ISOs, pulling packages for their alpha 2 systems... Feb 28 19:01:27 that shouldnt affext ports.ubuntu.com though Feb 28 19:02:04 in theory :D Feb 28 19:03:11 Neko: I have my own mirror of ports so I don't get hit by bandwidth issues. Feb 28 19:03:31 I use approx so I have a good buffer but it doesn't help when 90% of stuff got rebuilt Feb 28 19:04:24 my mirror updates every 4 hours. real slow for the first few weeks after UDS, but good overall. Feb 28 19:33:08 janimo: ping ? Feb 28 19:37:30 ericb2, hello Feb 28 19:37:46 janimo: hi. Are you Jani Monoses ? Feb 28 19:37:54 ericb2, yes Feb 28 19:38:25 janimo: maybe you'll be interested : I implemente the interlock part in arm assembler for armv7 + in OOo4Kids Feb 28 19:38:47 ericb2, regarding the recent bug I filed? Feb 28 19:38:48 janimo: in sal/osl/unx/interlck.c Feb 28 19:39:04 janimo: I'm not aware Feb 28 19:39:14 I was under the impression we need to move away from asm Feb 28 19:39:30 I sent a patch to libo to use gcc atomic builtins for that Feb 28 19:39:30 janimo: it woarks really wel on OOo4Kids Feb 28 19:39:44 I think gcc does a good job for atomic ops nowadays Feb 28 19:40:05 and imo the less ifdefs and assembly code the better :) Feb 28 19:40:15 janimo: looks like I did the same Feb 28 19:40:16 ericb2, is that of fork of OO? Feb 28 19:40:18 janimo: http://eric.bachard.free.fr/patches/OOo4Kids/linux_arm/arm_DEV300_m93.diff Feb 28 19:40:31 janimo: yes, OOo4Kids (and OOoLight ) is a fork of OOo Feb 28 19:40:46 janimo: only the atomic part might interest you Feb 28 19:40:56 ah I see, so you created the original arm optimization patch for OO? Feb 28 19:41:02 that is currently in natty Feb 28 19:41:23 janimo: I wrote this patch, but I don't know what other people dd with it :) Feb 28 19:41:58 janimo: on my BeagleBoard, the change is really visible Feb 28 19:42:06 janimo: and faster Feb 28 19:42:25 janimo: so far, no crash yet, but maybe the patch needs to be tested intensively Feb 28 19:43:17 ericb2, it is a coincidence then, as I only filed a bug related to this against ubuntu libo today Feb 28 19:43:30 I bet it is faster, the alternative used pthread mutexes Feb 28 19:44:09 janimo: yes. Some people like rene cry because I don't care armv6 or prior Feb 28 19:44:13 still I think you'd get the same improvement and much cleaner code - no configure foo and no inline asm, using only the bits you used for arm <7 in your code Feb 28 19:44:17 janimo: what is the problem Feb 28 19:44:32 janimo: I tested both, and the inline asm is faster Feb 28 19:44:34 ericb2, well you care for that too, by letting gcc do the right thing there Feb 28 19:44:47 janimo: since I added other fixes Feb 28 19:45:22 I did not perf test but looking at the code generated by gcc I see the same sequence of instructions so I was hopiong the performance is more or less the same Feb 28 19:45:24 janimo: and the build is fine on Debian, on Ubuntu and on OE distributions (e.g. the one sakoman provides) Feb 28 19:45:35 one fix needed is meory barriers Feb 28 19:45:43 on beagle there is no need Feb 28 19:45:59 but on SMP systems like the panda not using dmb can lead to bugs I guess Feb 28 19:46:36 ericb2, https://bugs.launchpad.net/ubuntu/+source/libreoffice/+bug/726529 Feb 28 19:46:37 Launchpad bug 726529 in libreoffice "use arm assembly bits only for gcc < 4.6 on ARM > 6" [Undecided,In progress] Feb 28 19:47:01 janimo: you're probably right. I filed the code thinking people could be interested, and nobody can imagine all cases Feb 28 19:47:21 janimo: so any improvement is good to hear Feb 28 19:47:33 ericb2, I think they were interested as long as your patch is part of the current LibO and former OO builds in Ubuntu Feb 28 19:47:54 * ericb2 discovering the patch is part of current Libo Feb 28 19:48:06 LibO in Ubuntu Feb 28 19:48:10 it is a patch Ubuntu carries Feb 28 19:48:12 janimo: I know Feb 28 19:48:20 but it was commited to LibO as well today Feb 28 19:48:28 as it makes sense for older gcc build Feb 28 19:48:48 janimo: I was not aware. I don't follow Libo Feb 28 19:49:06 janimo: when Libo was decided, I was even not invited. Now, they can die Feb 28 19:49:10 it was mentioned in the bugs filed. Feb 28 19:49:20 janimo: no problem Feb 28 19:49:26 ericb2, I am not sure it was invitation based. I do not know the details Feb 28 19:49:41 I am sure they have a better future than OO, it is a shame you parted ways Feb 28 19:50:00 janimo: well, I think I contribute since a long while to OOo, and not inform me was a choice. No problem for me anyway Feb 28 19:50:02 they are responsive, I doubt you'll get your ARM patches in OO Feb 28 19:50:27 ericb2, I am not sure every contributor was made aware of the fork before the announcement Feb 28 19:50:38 if so it must have been an oversight in your case Feb 28 19:50:43 janimo: it was thought : use it if you consider it usefull. If not well that's no problem Feb 28 19:51:02 janimo: no, I was very close to people organising everything, and not inform me was a choice Feb 28 19:51:03 I only got involved (slightly) since the fork and they come across as the most friendly project I have seen in a while Feb 28 19:51:33 ericb2, sorry to hear that. HAve you asked them afterwads, did they clarify anything? Feb 28 19:52:00 ericb2, is this ARM patch originally authored by you? Feb 28 19:52:12 janimo: I wrote it, yes Feb 28 19:52:21 janimo: but I didn't follow the story Feb 28 19:52:39 janimo: precisely, I wrote the one I provided you the link http://eric.bachard.free.fr/patches/OOo4Kids/linux_arm/arm_DEV300_m93.diff Feb 28 19:52:44 hmm, sadly there is no attribution on the patch in Ubuntu Feb 28 19:52:53 the one in ubuntu looks very similar Feb 28 19:53:09 janimo: as you can see, I backported it to OOo, making it more simpler for other forks, people Feb 28 19:53:22 janimo: who commited it ? Feb 28 19:53:51 janimo: http://cia.vc/stats/project/OOo4Kids Feb 28 19:53:58 janimo: revision r1152 Feb 28 19:54:02 today it was Bjoern, the UBuntu Libo maintainer. It was part of the ubuntu tree for a long time so he probably did not know whom to attribute Feb 28 19:54:38 janimo: svn diff -cr1152 svn://svn.adullact.net/svnroot/ooo4kids1/trunk Feb 28 19:54:45 what other changes does OO4kids have in general? Feb 28 19:55:05 janimo: no Java, OOo - 40% Feb 28 19:55:18 janimo: no Basic, just Python works Feb 28 19:55:24 this 1152 commit is a few weeks old, is the patch older? Feb 28 19:55:28 janimo: and new UI, working with students Feb 28 19:56:32 janimo: the commit is the first time I published my code. After, I backported to DEV300_m93, and I uploaded the patch I shown you Feb 28 19:56:53 janimo: other patch, is attached to OOo issue zilla Feb 28 19:57:04 janimo: let me retrieve the link Feb 28 19:57:21 ericb2, ok, for some reason I was under the impression the patch in Ubuntu is older, but I do not follow Ubuntu OO, it was just a coincidence I spotted this yesteday Feb 28 19:57:23 janimo: http://www.openoffice.org/issues/show_bug.cgi?id=117017 Feb 28 19:57:34 ericb2: Error: Could not parse XML returned by OpenOffice.org: timed out (http://openoffice.org/issues/xml.cgi?id=117017) Feb 28 19:57:53 janimo: I really wrote it Feb 28 19:58:06 janimo: I learned everything, and spent several nights to that Feb 28 19:58:42 janimo: and I ignored there was one similar patch on Ubuntu repo Feb 28 19:58:53 janimo: do you have the link ? Feb 28 19:58:59 * ericb2 curious Feb 28 19:59:35 checking Feb 28 20:00:17 this went into LibO today http://cgit.freedesktop.org/libreoffice/ure/commit/?id=83f2c071758ae7d74669d992e272e50057b895ed Feb 28 20:00:33 I am not sure there's a plain Ubuntu patch except in a tgz somewhere Feb 28 20:03:11 janimo: the patch I attached to IZ and I commited is dated 19 february. I'm pretty sure I completed it one day before. This leads to 18 february. Here is the story I know Feb 28 20:03:30 janimo: what happens with gcc >= 4.6 ? Feb 28 20:04:41 ericb2, it generated good enough code that it does not need to be manually written Feb 28 20:04:52 I mean for the _sync_XXX stuff Feb 28 20:05:17 janimo: __sync_add_and_fecth, ___sync_sub_and_fetch and al ? Feb 28 20:05:31 yes Feb 28 20:05:32 janimo: I see Feb 28 20:05:41 that was part of your patch as well right? Feb 28 20:06:16 janimo: yes. Ben ( a guy building on qemu + Debian arm ), had an issue with that. Was my fault Feb 28 20:06:23 janimo: and I discoverd the stuff Feb 28 20:06:57 janimo: more precisely, he had an undefined __sync_add_and_fetch_4 or something Feb 28 20:07:20 janimo: and after some grep's I discovered the thing, and had the idea to learn arm asm :) Feb 28 20:07:58 janimo: some years ago, I worked on m68k asm , and a bit on x86 asm too Feb 28 20:08:08 so was not much surprised Feb 28 20:09:16 janimo: as I mentionned in the log, Simon Guinot, helped me (he explained me the change after armv5 Feb 28 20:09:17 ) Feb 28 20:09:40 janimo: e.g. swap vs ldrex strex Feb 28 20:09:45 right Feb 28 20:10:32 janimo: I'm about to propose a subject for GSoC, about performance issues on arm Feb 28 20:10:50 ericb2, hmm so that arm patch went into debian via Rene?I see it in the debian pkg changelog Feb 28 20:10:52 janimo: OOo4Kids is an elephant, and we can improve a lot Feb 28 20:11:04 janimo: probably Feb 28 20:11:15 ericb2, I know OO is an elephant but I think much better gains could be had by doing higher level optimizations Feb 28 20:11:16 janimo: rene was crying a lot after me Feb 28 20:11:20 before going low level Feb 28 20:11:28 janimo: he didn't read correctly the issue and he was red :) Feb 28 20:11:28 like doing much less disk IO Feb 28 20:11:39 janimo: yes Feb 28 20:11:46 and pruning code and duplication and 20 year iolkd cruft Feb 28 20:11:48 janimo: and remove lot of useless stuff Feb 28 20:11:56 janimo: and more if affinity ^^ Feb 28 20:12:20 at this point I think the code needs cleanup much more that asm level magic that only a ahdnful of devs understand well Feb 28 20:12:31 just time is missing me : I got a real life, a real job, a family, doing a lot of sport :) Feb 28 20:12:59 janimo: I'm in OOo code since 2004, and I worked a lot on the native Mac OS X port Feb 28 20:13:14 janimo: including the MAc Intel one (we had to dive into the bridge too) Feb 28 20:13:42 ericb2: this is what Sweetshark did commit. http://libreoffice.pastebin.com/etTyZc5R GCC-4.6 inlines the _sync_* primitives Feb 28 20:14:03 doko_: probably Feb 28 20:14:28 doko_: as I told, I didnt follow what happened since I donated the patch Feb 28 20:22:43 doko_: that's true I'd have appreciated to see my name mentionned somewhere though ... Feb 28 20:24:13 ericb2: I didn't see any name within the patch, nor do I know where it was submitted ... Feb 28 20:24:57 doko_: it was originaly submitted to OOo IZ, to simplify the backports for all forks and so on. Feb 28 20:25:32 doko_: imagine it was submitted to LO, it would never have been reversed to OOo. So to respect everybody, the most simple was to attach it to OOo IZ Feb 28 20:26:18 ericb2: please tell Sweetshark on #ubuntu-desktop Feb 28 20:49:04 ogra: word up Feb 28 20:49:16 hohoho Feb 28 20:49:33 do you have any idea if we'll be getting alternate toolchains for arm? Feb 28 20:49:44 particularly: non-eabi and softfloat variants? Feb 28 20:49:46 alternate ? like LLVM ? Feb 28 20:49:50 ah Feb 28 20:49:57 not really important for the compiler and binutils as much as for the libgcc... Feb 28 20:49:57 not in natty Feb 28 20:50:20 this stuff is important for building u-boot :) Feb 28 20:50:38 we will get a hardfloat port in natty+1 which will run in parallel with the current one Feb 28 20:51:01 what for are you building u-boot ? Feb 28 20:51:07 the kobo Feb 28 20:51:11 i love this device Feb 28 20:51:15 it's cheap and hell and quite awesome Feb 28 20:51:30 ogra: is hard float reliable ? Feb 28 20:51:30 nice Feb 28 20:51:31 it has a somewhat hacked up firmware installed Feb 28 20:51:49 ericb2, according to markos_ who does the port in debian currently, it is, yeah Feb 28 20:51:55 they released the code as a bunch of tarballs against old linux/redboot releases Feb 28 20:52:01 and the patches are quite awful Feb 28 20:52:11 what architecture is that ? Feb 28 20:52:22 i've managed to get u-boot going on it and also have some patches against the kernel to make that work Feb 28 20:52:23 ogra: starting which proc ? Feb 28 20:52:49 ogra: for omap3 one told me it was not really. But maybe more recent is safe Feb 28 20:53:10 angstrom is fully built in hardfloat mode afaik Feb 28 20:53:14 since quite a while Feb 28 20:53:23 ogra: interesting Feb 28 20:56:15 ericb2, it depends on what you mean by reliable, usually packages just work, but there are a few who need special attention :) Feb 28 20:56:20 these are just a few though Feb 28 20:56:40 debian has reached 87% and we're hoping we can reach 95% soon Feb 28 20:57:04 markos_: what can bring hardfloat exactly ? Feb 28 20:57:16 er, a hardfloat abi? Feb 28 20:57:31 markos_: yes, I know, but in the real life: is it that faster ? Feb 28 20:57:39 * desrt thought that the change to eabi brought hardfloat with it Feb 28 20:58:00 markos_: just a question, because I didn't test yet Feb 28 20:58:01 it depends on what apps you depend on Feb 28 20:58:16 most apps benefit 5-30% depending on their use of fp Feb 28 20:58:57 some don't benefit at all, while apps heavy on fp might be 200% faster (like a raytracer, like pov, yes it was actually 200% faster) Feb 28 20:59:21 markos_: the one I have in mind is OOo4Kids (an OOo fork) Feb 28 20:59:40 markos_: I'm working on performances issues, and am interested by everything who cold help in this domain Feb 28 21:01:17 no idea how much better -if at all- OOo would be, though from my experience, anything that renders fonts is faster, ~20-25% Feb 28 21:01:29 but ymmv with OOo Feb 28 21:01:52 markos_: so it is interesting Feb 28 21:02:10 yes it is, that's what I've been trying to tell people all along :) Feb 28 21:04:01 markos_: do yo have a link, where I could read the flags being used, the cases .. and so on ? Feb 28 21:04:42 you need the whole distro being built for hardfloat Feb 28 21:05:00 the binaries wont run on a softfloat distro Feb 28 21:05:19 ericb2, working on getting armhf d-i these days Feb 28 21:05:28 ogra: isn't there a flag allowing both ? Feb 28 21:05:36 markos_: great :) Feb 28 21:05:37 ogra, well he coud use a chroot Feb 28 21:05:45 they are binary incompatible Feb 28 21:05:51 yeah. you could use a chroot Feb 28 21:05:54 or a vm Feb 28 21:05:56 markos_: I'm not a specialist (arm is a jungle), but extremely curious and interested Feb 28 21:06:00 both would work Feb 28 21:06:47 ericb2, give me a few weeks, I've commited some armhf stuff to d-i, but there remain a few irritating points still to fix Feb 28 21:07:04 anyway gotta go, have to tell a bedtime story :) Feb 28 21:07:04 markos_: I will. thanks :) Feb 28 21:07:14 ubuntu will have it with the next release Feb 28 21:07:18 markos_: see you later Feb 28 21:07:26 and slowly migrating 100% to it Feb 28 21:07:53 ogra: in fact, I'd like to provide an adapted version of OOo4Kids (or OOoLight) for Ubuntu users Feb 28 21:08:09 so watch this place ;) Feb 28 21:08:11 ogra: so I'll add what is mandatory, and stick your needs Feb 28 21:08:55 if you could get it running with debian hf, you will also get it running in trhe upcoming ubuntu port Feb 28 21:09:17 ogra: currently, I got OOo4Kids building and working on natty Feb 28 21:09:28 ogra: but I used soft fp Feb 28 21:09:40 right, thats fine for natty Feb 28 21:09:44 ogra: ok Feb 28 21:09:49 n+1 will have both Feb 28 21:09:54 soft and hard Feb 28 21:10:11 ogra: the only thing I'll need is a new micro SD card: and I'll install any experimental version on it Feb 28 21:53:07 ericb2: ping? **** ENDING LOGGING AT Tue Mar 01 02:59:57 2011