**** BEGIN LOGGING AT Tue Oct 02 02:59:58 2012 Oct 02 11:52:22 for anyone with a real beaglexm: can you try today's daily quantal image? Oct 02 11:52:39 http://cdimage.ubuntu.com/ubuntu-server/daily/current/quantal-server-armhf+omap.img Oct 02 11:52:42 thanks Oct 02 12:08:24 ppisati: have you updated kernel config with tidspbridge support ? Oct 02 12:57:25 bizulk: nope, busy dpoing other things Oct 02 12:57:45 bizulk: which kernel version are you running? Oct 02 12:57:54 bizulk: 3.2 or 3.5? Oct 02 12:58:00 ppisati: 3.2.30 Oct 02 12:58:09 bizulk: uhm ok Oct 02 12:58:21 willing to test quantal? Oct 02 12:59:05 bizulk: beagle right? Oct 02 12:59:11 http://cdimage.ubuntu.com/ubuntu-server/daily/current/quantal-server-armhf+omap.img Oct 02 12:59:19 ppisati: is quantal 12.10 ? Oct 02 12:59:25 bizulk: yep Oct 02 12:59:44 ppisati: yes if the tidspbridge is installed. Cause that would be the only reason I would be a"allowed" to. Oct 02 13:00:58 ppisati: I am also experiencing some network init issue on the 12.04 : nm-applet does not apply configured profile Oct 02 13:02:25 bizulk: well, nm is userspace Oct 02 13:02:35 bizulk: if ifconfig shows the interface, we are good Oct 02 13:02:48 bizulk: actually i was looking for someone willing to do some testing on real hw Oct 02 13:02:55 bizulk: since we are facing a problem with usb Oct 02 13:03:39 ppisati: I have a BB xM. But as I am in the office I 'can' only work on the dspbridge stuff, u know Oct 02 13:04:22 ppisati: what's your PB with USB ? it seemed to work well (using a powered usb hub in my case) Oct 02 13:04:24 bizulk: ack Oct 02 13:04:39 bizulk: it's broken since 3.5 Oct 02 13:04:45 bizulk: still broken in 3.6 Oct 02 13:04:57 but i wanted someone else to double check it Oct 02 13:05:08 actually booting a precise 3.2 kernel everthing is ok Oct 02 13:05:09 so... Oct 02 13:05:30 bizulk: anyway, remind me your lp bug Oct 02 13:05:36 bizulk: i'll do the config changes now Oct 02 13:05:58 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1058022 Oct 02 13:05:59 Launchpad bug 1058022 in linux "no tidspbridge support in kernel." [Medium,Triaged] Oct 02 13:07:30 ppisati: I experience USB pb on standalone 3.6 kernels, but I thoughed It was because of my power supply (usb always restarting, until the CPU itstelf resets) Oct 02 13:13:40 bizulk: does it generate a /dev entry? which one? Oct 02 13:14:10 /dev/DspBridge Oct 02 13:14:48 ppisati: crw-rw-rw- 1 1000 1000 251, 0 2011-08-31 13:24 rootfs/dev/DspBridge Oct 02 15:06:05 janimo: is the "tegra hw video decoder config bug" already fixed in some ac100 kernel image? Oct 02 15:08:24 ah, just saw http://kernel.ubuntu.com/git?p=jani/ubuntu-ac100.git;a=shortlog;h=refs/heads/packaging-3.1 Oct 02 15:08:27 sorry Oct 02 15:08:58 but still kinda wrong - did this even compile? Oct 02 15:09:09 marvin24, I think the deb built fine Oct 02 15:09:19 marvin24, I was told later that this change is not enough Oct 02 15:09:20 mmh, I think we also need CONFIG_TEGRA_AVP_KERNEL_ON_MMU Oct 02 15:09:24 ok Oct 02 15:09:38 I tested a kernel with both options enabled only Oct 02 15:09:45 can someone build the package and provide the right config option changes? Oct 02 15:10:07 I will ... Oct 02 15:10:12 just a few secs ;-) Oct 02 15:10:19 cloning the repo, fdr clean, fdr editconfigs, debuild Oct 02 15:10:38 where fdr is fakeroot debian/rules and debuild the cross-build line Oct 02 15:10:46 * ogra_ would if it wouldnt take ages to pull the source package on my loaded line Oct 02 15:10:51 or use a script I wrote some time ago ;-) Oct 02 15:11:08 ogra_, that is why you should keep the git tree checked out and updated frequently :) Oct 02 15:11:20 you have a beefy machine, no excuse for not building kernels anymore Oct 02 15:11:22 pfft, i still stay away from git if i can Oct 02 15:11:54 ogra_, in this case you only need git clone, if you just want to check new stuff or do custom builds Oct 02 15:11:56 and i have a local mirror ... so i can just use source packages usually ... but not for univers Oct 02 15:11:58 e Oct 02 15:12:39 if marc has a local branch he will surely be faster Oct 02 15:12:50 ogra_: I have one Oct 02 15:13:00 but building without ccache is hell Oct 02 15:13:12 janimo: did you found out how to enable it? Oct 02 15:13:15 what are you building on ? Oct 02 15:13:21 x86_64 Oct 02 15:13:37 well, same here Oct 02 15:13:55 marvin24, I have a not-too fast core duo, which builds in maybe 30 minutes Oct 02 15:14:05 same here :-( Oct 02 15:14:07 i use to do heavy builds in ramdisks though, that speeds up things a lot Oct 02 15:14:09 not sure if that qualifies as 'hell' for you though :) Oct 02 15:14:25 with ccache it takes less than 2 minutes Oct 02 15:14:42 without ccache it takes around 10 for me ... Oct 02 15:14:53 marvin24, I believe you. Not sure why I never looked at why ccache did not work Oct 02 15:15:09 must have something to do with fakeroot Oct 02 15:29:12 ogra_: picking up AC100 tomorrow 112GBP (139 EUR) Oct 02 15:29:25 hah, cool ! Oct 02 15:41:13 ogra_: did akon reached you regarding the library name problem? Oct 02 15:41:38 marvin24, nope, i was one at 1:30 when he pinged me Oct 02 15:41:42 (sorry I'm just catching up with the backlog because of short time holidays) Oct 02 15:41:53 waiting for him to re-appear thugh Oct 02 15:42:06 you took a long weekend ? Oct 02 15:46:11 infinity, mind helping to explain the nvidia ld issue to srwarren (from nvidia) ? Oct 02 15:46:17 infinity, could you detail the technical issues that the incorrect soname in the NVIDIA Tegra R16 libs? Oct 02 15:46:29 i fear i'm not as accurate as you can be :) Oct 02 15:46:42 I know the names are wrong and what needs to be changed; I'm just trying to understand the exact implication of the current incorrect names Oct 02 15:47:02 it boils down to "that it currently works if you put the libs into /usr/lib is sheer luck) Oct 02 15:47:52 ogra_: yep, I measured the circumference of the Edersee Oct 02 15:47:55 if you put the libs into any different path thatrs not hardcoded in ld only ld.so.cache will be used ... in which we have the wrong SONAME Oct 02 15:47:56 For ldconfig, I think what will happen is that if libfoo.so's soname is libfoo.so, presumably ldconfig would simply not create any symlink since the file is already present under the expected name and move on Oct 02 15:48:13 srwarren: The libraries end up being uncacheable by ldconfig because the filenames and SONAMEs can't match. Oct 02 15:48:24 if they live in /lib or /usr/lib ld falls back to walk the path and actually look at the links too Oct 02 15:48:34 srwarren: This is a problem given that ld.so uses the cache to find libraries. Oct 02 15:48:50 janimo: kernel without CONFIG_TEGRA_AVP_KERNEL_ON_MMU crashed hard on video decode here Oct 02 15:48:51 srwarren: And yes, if things are in the "built-in" paths, then they get found the slow (cache-missed) way. Oct 02 15:49:26 srwarren: So, the best way to look at it is that it's a performance hit. The worst way to look at it is that all the SONAMEs are wrong, and that's just plain, well, wrong. ;) Oct 02 15:49:29 janimo: it is also enabled in tegra_defconfig and I want to stay as close as possible to downstream Oct 02 15:49:37 With the current sonames, don't the filenames and sonames always match? Oh, I guess you're renaming the files in the Ubuntu package so that apps with DT_NEEDED=libfoo.so.1 can actually find the library? Oct 02 15:49:58 srwarren: Yeah, everything with a NEEDED it looking for the correct SONAME, which isn't in the library. Oct 02 15:50:10 srwarren: So, we get a cache miss, then start walking the filesystem. Oct 02 15:50:40 srwarren, no, we put the libs into /usr/lib/nvidia-tegra and the SONAMEs end up in the cache ... i.e. libEGL.so ... GLES apps are built to look for libEGL.so.1 Oct 02 15:50:52 srwarren: It's not about us renaming them. It's about the fact that they need to exist by those names. :P Oct 02 15:50:52 OK, so the entries in /etc/ld.so.conf.d (or whatever the file is) only get used to build the cache, and not as part of the fallback searching Oct 02 15:51:09 right Oct 02 15:51:15 srwarren: Right, because parsing a conf.d directory when loading every single binary on your system would be, well, dumb. Oct 02 15:51:51 srwarren, but even if that wouldnt be an issue ... the first GLES app you would build on a system with the drivers installed would have completely broken linking Oct 02 15:52:04 right, that's the part I already understood Oct 02 15:52:10 srwarren: Anyhow, I'm trying to decide which bit you're asking me to explain. If you want to know why the SONAMEs are wrong, or why attempts to work around it suck? Oct 02 15:52:32 afaik for libEGL.so as well as for libGLESv2.so the sonames are actually standadized Oct 02 15:52:42 rsalveti, might know :) Oct 02 15:52:52 I know exactly why the SONAMEs are wrong; I was trying to understand what practical impact that had. I'd only deduced the application-compilation issue so far, not the searching issue Oct 02 15:53:33 Ahh, yes. Oct 02 15:53:53 So, yeah. If we ship everything in /usr/lib, is kinda works due to the cache-miss->directory-walk thing. Oct 02 15:53:55 infinity, so when the files aren't found in the cache, and ld.so falls back to searching e.g. /usr/lib, how does it find the files even then, if they still have the wrong soname and filename? Oct 02 15:54:01 marvin24, I agree with staying close to defconfig downstream. Just that I not always sync up with defconfig in the package, at least some bits are not needed or incorrect in ubuntu (lzo) so I tend to drop more Oct 02 15:54:13 srwarren, it doesnt :) Oct 02 15:54:14 srwarren: Because we symlink the correct SONAME to them. Oct 02 15:54:21 srwarren: And then it find them by filename. Oct 02 15:54:33 ok, that makes sense - there's a workaround to make it work Oct 02 15:54:38 ogra_: Don't ask me to explain things and then jump in with contradictory statements. ;) Oct 02 15:54:40 oh, i misread Oct 02 15:54:48 lol, sorry Oct 02 15:55:00 didnt mean to, i just read something completely different Oct 02 15:55:06 srwarren: Yes, our workaround for now is to symlink stuff from /usr/lib (or something else on the path) Oct 02 15:55:26 but thats nothing we can do in a package Oct 02 15:55:28 srwarren: But that's still pretty wildly less than ideal, if you guys can actually fix the SONAMEs. Oct 02 15:55:49 ogra_: Well, we *can*... We really shouldn't. Oct 02 15:56:15 yes Oct 02 15:56:21 (And I probably won't accept it in the archive...) Oct 02 15:56:25 But, y'know. You can upload it. Oct 02 15:56:35 janimo: that's fine, but you can check my paz00_defconfig against the latest you used for the last package Oct 02 15:56:47 I guess this is because the multi-driver co-existence stuff is based on putting entries into the ld.so cache-building path list, rather than using the alternatives system on the .so filenames themselves Oct 02 15:56:53 janimo: that should be a pretty short diff Oct 02 15:57:22 srwarren: Using alternatives on .so doesn't make much sense, since we don't install .so files except with -dev packages... Oct 02 15:57:37 Well, *.so.1 Oct 02 15:57:51 marvin24, janimo - https://bugs.launchpad.net/ubuntu/+source/linux-ac100/+bug/1059866 Oct 02 15:57:52 Launchpad bug 1059866 in linux-ac100 "video hw acceleration still dont work" [Undecided,Confirmed] Oct 02 15:58:03 srwarren: And yeah, we use alternatives on the ld.so.conf instead of on the files. Correct. Oct 02 15:58:21 srwarren: Which is actually much more manageable. If all the libraries work. ;) Oct 02 15:58:39 yes, I can see that scales a lot better with multiple libs Oct 02 15:58:53 Just a lot less error-prone, really. Oct 02 15:59:13 Adding and removing alternatives and slaves for a ton of stuff makes people go cross-eyes, no matter how awesome the syntax-hilighting. Oct 02 15:59:24 s/cross-eyes/cross-eyed/ Oct 02 16:02:16 ppisati: hi. I saw my bug update. As soon as possible I take a look at this Oct 02 16:07:54 marvin24, stuw ok I'll have a look. It's just that yesterday's suggestion was a one line diff as well but was not enough Oct 02 16:08:32 it's just that I do not use the ac100 and have testcases to check various features so I will mostly blindly do whatever I am asked by others who actually use the machine :) Oct 02 16:09:15 ideally those people would take care of kernel packaging too but I am asking too much (hint hint) Oct 02 16:09:19 ;) Oct 02 16:10:14 janimo, well, it was discussed on and off in #ac100 what options need to be on :) you could have fished it out of your backlog Oct 02 16:10:17 janimo, http://paste.ubuntu.com/1256243/ - .config after make ARCH=arm paz00_defconfig Oct 02 16:11:02 ogra_, I haven't logged in ac100 in many months, also as a cosequence of it mostly being low signal to noise for what I was interested in back then Oct 02 16:11:56 well, that changed Oct 02 16:12:12 there is still a lot of noise but currently thats all ac100 noise Oct 02 16:13:34 ogra_: not noice, hifi sound! Oct 02 16:13:41 *noise Oct 02 16:13:55 oh, yeah, compared to the last 6 months this current hype is hifi Oct 02 16:14:05 Hahaha. Oct 02 16:15:27 * marvin24 still wonders why people are still interested in 2 years old machines Oct 02 16:16:12 Likely due to the lack of decent ARM netbooks. Oct 02 16:16:26 yeah Oct 02 16:16:29 I just want a reasonably speedy one with a non-Android en_US keyboard. Some day... Oct 02 16:16:33 you can still buy it and its still cheap Oct 02 16:16:59 I like my transformer for that reason, running ubuntu it makes a nice ARM netbook Oct 02 16:17:05 and with the 1280x720 display and the internal UDB disk its now gotten really uasble Oct 02 16:17:10 *USB Oct 02 16:17:18 lilstevie: Yeah, but the transformer keyboard makes me die a little inside. Oct 02 16:17:25 oh yeah Oct 02 16:17:31 not just a little Oct 02 16:17:31 infinity, why? Oct 02 16:17:49 lilstevie: Well, (a) Android layout, and (b) it's just not a nice keyboard to type on. Oct 02 16:17:52 * ogra_ also doesnt like the shape ... since i have my zatab i use that more than my transformer Oct 02 16:18:03 less sharp corners on the case etc Oct 02 16:18:23 even though the zatab is classes slower Oct 02 16:18:28 infinity, old style or new though, android layout is a pity, but I don't really look at the keys, and I have altered the keymap Oct 02 16:18:39 (these A10's are really not made for multitasking) Oct 02 16:18:50 typing though I find it a bit better than my macs bt keyboard Oct 02 16:19:02 the tf101 keyboard was nowhere near as nice though Oct 02 16:19:13 * ogra_ really likes the ac100 kbd Oct 02 16:19:40 the low resolution bothered me for actual work, but i fixed that :) Oct 02 16:20:17 I wish the tf201 was a bit higher resolution Oct 02 16:20:27 1280*800 is nice, but it could be better Oct 02 16:20:38 bizulk: fix was committed, when the next kernel is but i'll be in Oct 02 16:20:41 * ogra_ is happy with 1280x720 Oct 02 16:20:47 I'd probably pay the Thinkpad brand premium for an ARM netbook with a Lenovo keyboard. Oct 02 16:20:49 bizulk: *kernel is cut Oct 02 16:21:42 ppisati: sorry you mean "next kernel is built I keep you informed" ? Oct 02 16:21:48 infinity, I wonder whatever happened to that Lenovo transformer like T3 tablet Oct 02 16:23:27 bizulk: no, it means next Precise kernel upload will contain the fix Oct 02 16:23:57 ppisati: How can I know when it's done ? Oct 02 16:24:13 bizulk: sudo apt-get update upgrade Oct 02 16:24:21 opps sorry Oct 02 16:25:17 with update-alternative cmd I can select the kernel release I want (including the unofficial one) ? Oct 02 16:26:07 kernels don't use alternatives, no. Oct 02 18:59:36 janimo, do you actually look over the linux-ac100 buglist sometimes ? Oct 02 19:00:15 bug 961302 seems valuable if you dont want to upload for a single config change :) Oct 02 19:00:15 Launchpad bug 961302 in linux-ac100 "[AC100] Request HID Waltop kernel module for Waltop tablet" [Undecided,New] https://launchpad.net/bugs/961302 Oct 02 19:21:51 ogra_, no I did not look at the buglist recently Oct 02 19:22:05 I'll look into them Oct 02 19:22:27 I added some more modules in a recent upload but still far from what stock ubuntu kernel has Oct 02 19:22:38 ogra_, git has a bright future Oct 02 19:22:45 just sayin ;) Oct 02 21:42:01 janimo: I fixed the fuse cannot be loaded bug Oct 02 21:42:02 https://bugs.launchpad.net/ubuntu/+source/linux-ac100/+bug/1060050 Oct 02 21:42:03 Launchpad bug 1060050 in linux-ac100 "Can't mount ntfs volume" [Undecided,New] Oct 02 21:42:16 you may pull my tree again Oct 02 21:42:25 \o/ Oct 02 21:42:27 marvin24, ok Oct 02 21:42:32 such a wondeful bug Oct 02 21:42:41 really deserves a printout and a frame :) Oct 02 21:42:57 (not the LP bug, the code issue indeed) Oct 02 21:43:48 in fact, renaming arch/arm/mach-tegra/fuse.c fixed it Oct 02 21:44:09 haha Oct 02 21:44:18 if a kernel parameter is created Oct 02 21:44:38 a file in /sys/module//parameters/... is created Oct 02 21:44:53 where filename is "fuse" in this case Oct 02 21:45:16 so the filesystem fuse driver cannot register anymore, because the sysfs entry is used already Oct 02 21:45:25 took some time to find this ... Oct 02 21:47:11 yeah, great catch Oct 02 22:00:56 could someone aid me in porting ubuntu to the tf700 Oct 02 22:05:25 anyone used linux target image builder here? Oct 02 22:05:30 or LTIB as it's called Oct 02 22:05:47 I'm wondering how I can set the --build parameter there Oct 02 22:05:52 or mach type Oct 02 22:06:02 on the make Oct 02 22:06:12 * ogra_ never heard of it Oct 02 22:06:24 Freescale uses it for their CPUs **** ENDING LOGGING AT Wed Oct 03 02:59:58 2012