janimo | ogra_, you really should change that 30min estimation to 5min in the tarball unpacking message :) | 12:20 |
---|---|---|
* janimo is installing today's ac100 image | 12:21 | |
=== hrw is now known as hrw- | ||
ogra_ | janimo, heh, ok | 12:22 |
=== hrw- is now known as hrw | ||
janimo | getting into X in the installer is noticably slower than in precise | 12:23 |
janimo | at one poits it's at console login as if no X is gonna start | 12:23 |
ogra_ | the initrd somewhat takes longer | 12:23 |
ogra_ | ubiquity-dm was always slow though | 12:23 |
janimo | this was after starting userland on root, not sure it was the initrd | 12:23 |
ogra_ | lightdm is fine | 12:23 |
janimo | X seemed slow to paint | 12:24 |
ogra_ | well, the time it takes from booting to the fsck messages | 12:24 |
ogra_ | thats significantly longer ... | 12:24 |
ogra_ | and i fear on all arches | 12:24 |
janimo | getting to fsck took longer a bit, but then another 2 min until X | 12:25 |
janimo | installer enocuntered unrecoverable error :( | 12:25 |
* janimo checks logs | 12:25 | |
janimo | I wish there was a way of logging in/starting a shell to debug when this installer crashes | 12:26 |
janimo | 'a desktop session will be run so that you may investigate' but that too crashes | 12:27 |
ogra_ | wow, my install worked fine OOTB here | 12:30 |
ogra_ | which image did you use ? todays ? | 12:30 |
janimo | ogra_, today's yes | 12:30 |
ogra_ | weird | 12:35 |
ogra_ | no installer issues at all here | 12:35 |
ogra_ | the screen doesnt wake up properly after DPMS kicked in though | 12:36 |
ogra_ | (just noticed) | 12:36 |
=== awafaa_ is now known as awafaa | ||
ogra_ | janimo, bug 1026577 ? | 12:48 |
ubot2 | Launchpad bug 1026577 in ubiquity "ubiquity crashed with TypeError in _execute_child(): Can't convert 'list' object to str implicitly" [Critical,Fix released] https://launchpad.net/bugs/1026577 | 12:48 |
ogra_ | even though i didnt see it here, it is obviously in these images) | 12:48 |
ogra_ | ah, no, ubi-partman | 12:49 |
ogra_ | ignore me | 12:49 |
janimo | ogra_, how hacky/hard would it be to provide a login shell before the install is complete? Ubuntu has a live desktop with ubuntu user where you can look around at will | 12:58 |
ogra_ | pretty hackish and hard | 12:58 |
ogra_ | (and really nothing i want to invest time in) | 12:59 |
janimo | ogra_, what caused the crash last time was trying to use wifi, let's see what happens this time when I skip that | 12:59 |
ogra_ | weird, worked all fine here | 12:59 |
janimo | ogra_, I agree then, if there is time to invest it should be towards unifying with the ubuntu live installer | 12:59 |
janimo | I kept pushing connect and it did not | 13:00 |
ogra_ | the only issue i currently see is DPMS and plymouth | 13:00 |
janimo | if I connected from the NM icon in the panel it worked | 13:00 |
janimo | but then crashed | 13:00 |
ogra_ | weird | 13:00 |
ogra_ | the page doesnt automatically move forward after it connected | 13:00 |
janimo | oh now it went ahead did not even ask about wifi | 13:00 |
ogra_ | but the button changes | 13:00 |
marvin24 | ogra_: that's a tegrafb bug | 13:00 |
marvin24 | switch to X and back may help | 13:01 |
ogra_ | janimo, is that a freshly unpacked tarball or just the failed install rebooted | 13:01 |
janimo | ogra_, failed install rebooted | 13:01 |
ogra_ | if its the failed one, debconf definitely keeps your input | 13:01 |
ogra_ | so thats normal ... | 13:01 |
janimo | ok, now it installs | 13:02 |
janimo | still a bug but maybe it's in the logs somewhere | 13:02 |
ogra_ | marvin24, thx, it does indeed | 13:02 |
ogra_ | yeah | 13:02 |
ogra_ | might be HW though | 13:02 |
marvin24 | the nv guy seems to have given upon the console problems as it seem | 13:02 |
marvin24 | I got no reply from him for a few days | 13:02 |
* ogra_ just built an lzma compressed kernel ... curious if that will work | 13:03 | |
ogra_ | marvin24, do you know of any source for a README or HOWTO what needs to be switched on in the kernel to make the binary driver work ? | 13:03 |
ogra_ | i suspect our kernel still doesnt have everything needed and would like to find out if its a config prob | 13:04 |
marvin24 | nv added a lot of options, yes | 13:05 |
marvin24 | I guess some tegra_defconfig should be sufficient | 13:05 |
ogra_ | is that documented anywhere ? | 13:05 |
ogra_ | hmm, k | 13:05 |
marvin24 | arch/arm/configs/... | 13:05 |
janimo | ogra_, maybe getting the kernel image from the L4T package and seeing what config it has enabled would help? | 13:05 |
ogra_ | janimo, i would, if i could :) | 13:05 |
marvin24 | is there a new l4t release? | 13:06 |
ogra_ | nvidia downloads are completely shut down since the hack | 13:06 |
ogra_ | marvin24, only R15 final | 13:06 |
janimo | marvin24, L4T 15 final of a week or so ago | 13:06 |
ogra_ | i think you know about it | 13:06 |
janimo | lilstevie, said the actual download links still work, just the site frontend/logins is stopped | 13:06 |
ogra_ | we already have it in quantal, but x fails | 13:06 |
janimo | it's just that we don't know the link URLs :) | 13:06 |
ogra_ | janimo, well, doesnt help if i want to lok for a readme | 13:07 |
ogra_ | i know the url | 13:07 |
janimo | ogra_, if there's a kernel image in the URL that may have configs already | 13:07 |
ogra_ | but the README contained doesnt talk about kernels at all | 13:07 |
janimo | which we can compare with ours | 13:07 |
lilstevie | ogra_: I have most of the documentation on hand | 13:07 |
lilstevie | what is it about the kernel that you are interested in | 13:07 |
ogra_ | yes, it might be in the softfloat package | 13:07 |
lilstevie | yeah | 13:08 |
lilstevie | itttt is | 13:08 |
lilstevie | but I also have the documentation package that is just the docs | 13:08 |
ogra_ | lilstevie, on ac100 lightdm gots in a crash loop ... killing lightdm and running startx gets me a desktop but as soon as i click something it crashes | 13:08 |
ogra_ | *goes | 13:08 |
lilstevie | quantal only? | 13:08 |
ogra_ | yep | 13:08 |
ogra_ | ABI12 | 13:09 |
marvin24 | ogra_: so we are left with tegra_android_defconfig | 13:09 |
lilstevie | hm | 13:09 |
ogra_ | marvin24, hmm, android doesnt use the xerver, do they ? | 13:09 |
lilstevie | well as far as the docs say as long as you are using l4t-r15-rc tag it should be fine | 13:09 |
marvin24 | no, but that's the only tegra2 config I found | 13:09 |
ogra_ | is our kernel at that tag ? | 13:10 |
lilstevie | as of r15 they rolled the DC code to be a common base | 13:10 |
marvin24 | ogra_: no, we are further in the future ;-) | 13:11 |
marvin24 | rel-15r7 | 13:12 |
marvin24 | (branch) | 13:12 |
=== satellit-arm is now known as satellit | ||
marvin24 | tegra-15r7.1-android-4.0 | 13:12 |
marvin24 | tag | 13:12 |
janimo | marvin24, lilstevie actually with the last update we are at origin/l4t/l4t-r15 | 13:13 |
janimo | I rebased on what I learned was very latest and recommended with L4T drivers by nvidia | 13:14 |
marvin24 | janimo: so you stepped back? | 13:14 |
janimo | marvin24, is that back?? | 13:14 |
marvin24 | rel15r7 is 4 weeks old | 13:14 |
marvin24 | l4t/l4t-r15 is 7 weeks old | 13:14 |
janimo | I hate nvidia tag/branch naming scheme | 13:14 |
marvin24 | http://nv-tegra.nvidia.com/gitweb/?p=linux-2.6.git | 13:15 |
marvin24 | janimo: yes, this is very confusing | 13:15 |
janimo | marvin24, so 15r7 is the one that goes with L4T final? | 13:15 |
marvin24 | I think they have different l4t and android branches | 13:15 |
lilstevie | they seem to maintain these things in parallel and not talk to eachother | 13:15 |
janimo | I thought from the most recent discussion with lilstevie that the branch I used was | 13:15 |
lilstevie | janimo: technically no | 13:15 |
janimo | oh well | 13:15 |
marvin24 | l4t and android may have different branches, yes | 13:16 |
marvin24 | I used the android one because it seemed to me that it has more fixes | 13:16 |
janimo | I tried avoiding the android branch for that reason | 13:16 |
janimo | by that reason being - it goes with L4T | 13:16 |
janimo | no idea what the diffs are | 13:16 |
marvin24 | works fine here ... | 13:16 |
janimo | marvin24, this branch of yours is newer that what we had last week in ubuntu? | 13:17 |
lilstevie | I haven't seen the new hdmi stuff in the newer rel-15r7 code from android | 13:17 |
marvin24 | mine is based on tegra-15r7.1-android-4.0 | 13:17 |
marvin24 | while you seem to have used tegra-l4t-r15-rc | 13:18 |
janimo | ogra_, so ac100 install finished just as usual the first thing it does is crashes and invokes apport | 13:23 |
ogra_ | lol | 13:30 |
ogra_ | well, if you had a ubiquity crash there is a core dump in /var/crash on first start | 13:30 |
ogra_ | apport just picks that up | 13:30 |
ogra_ | WOW !!!!!! | 13:44 |
=== ahs3` is now known as ahs3 | ||
ogra_ | building the kernel lzma compressed saves 1MB ! | 13:44 |
ogra_ | if that actually works, i could leave plymouth ion the initrd and drop all the ugly diversion hackery | 13:45 |
ogra_ | and it does ! | 13:45 |
lilstevie | ogra_: I have been building my kernel lzma'd for a while now for size reasons | 13:46 |
ogra_ | lilstevie, right, it was never clear if the ac100 fastboot can handle that | 13:46 |
marvin24 | kernel decompresses the initrd | 13:47 |
ogra_ | i had mixed results in the past (though testing initrds, not kernels) | 13:47 |
lilstevie | well the kernel decompressor code does it | 13:47 |
marvin24 | fastboot only has the 2MB limit | 13:47 |
marvin24 | so you may have ran into this problem | 13:47 |
ogra_ | i had initrds failing with the old kernels when trying out other methods than gzip | 13:47 |
ogra_ | (2.6.29 etc) | 13:47 |
marvin24 | you need to select the decompressor in the kernel config | 13:48 |
ogra_ | for the initrd it just needs to be enabled | 13:48 |
marvin24 | older kernels didn't support so many (only gz and bz2 I think) | 13:48 |
ogra_ | and we always had lzma available as inbitrd compression method | 13:49 |
ogra_ | even in the kernels where iot didnt work | 13:49 |
ogra_ | lilstevie, do you use any special cmdline options on your tablets to work with the r15 driver ? | 13:50 |
ogra_ | vram or something alike | 13:50 |
lilstevie | no more than usual | 13:52 |
ogra_ | hmm, k | 13:52 |
lilstevie | http://paste.ubuntu.com/1100098 | 13:52 |
ogra_ | vmalloc=128M video=tegrafb | 13:52 |
ogra_ | do you use these normally ? | 13:52 |
lilstevie | but that is what the bootloader passes up | 13:53 |
lilstevie | yes | 13:53 |
lilstevie | almost the entire commandline is from the bootloader | 13:53 |
ogra_ | i assume that doesnt fix the DPMS issues i see here "no_console_suspend=1" ... | 13:54 |
lilstevie | DPMS issues? | 13:56 |
ogra_ | i need to switch to console after waking up from DPMS to get X back | 13:56 |
ogra_ | els i end up with a black screen | 13:57 |
lilstevie | oh | 13:57 |
lilstevie | no I don't have that | 13:57 |
lilstevie | in fact switching to any tty means I don't come back :/ | 13:58 |
ogra_ | ah | 13:58 |
ogra_ | janimo, oh, i get a crash massage as well, seems its plymouthd | 13:59 |
lilstevie | but I wake from DPMS fine | 13:59 |
janimo | ogra_, no coming back from suspend either here, can't seem to switch to any VT | 14:04 |
ogra_ | yep | 14:07 |
janimo | ogra_, my initial ubi crash was indeed due to failing to connect NM, which crashed before that | 14:16 |
janimo | so this is clearly a bug unrelated to ac100 installer | 14:16 |
ogra_ | yeah | 14:18 |
janimo | unfortunately not enough info in the coredump | 14:19 |
* ogra_ is more intrested in the plyouth segfaults though | 14:19 | |
ogra_ | though it seems that my diversions are so good that i cant properly revert them :( | 14:20 |
janimo | marvin24, I was just about to tell you to skip moving the framebuffer as it causes the toshiba logo to scroll uglily up | 15:59 |
janimo | but I see you have a change already there | 15:59 |
janimo | would be good to skip it even if not using uboot for above aesthetic issue | 16:00 |
ogra_ | uglily eh ? | 16:00 |
ogra_ | :) | 16:00 |
janimo | well, it does not scroll smoothly enough | 16:00 |
ogra_ | ++ btw | 16:00 |
janimo | either we fix that or remove the scroll :) | 16:00 |
ogra_ | yeah, it shouldnt move at all | 16:00 |
ogra_ | it didnt in the past | 16:00 |
janimo | it's a new change in 3.1 kernels only, needed for nvidia devboards apparently, but maybe not at all for us | 16:01 |
ogra_ | yeah | 16:01 |
LetoThe2nd | howdy! is there some "official" ubuntu arm kernel git tree? like what goes into the omap4 images? | 16:23 |
marvin24 | janimo: I don't know why they did it | 16:55 |
marvin24 | maybe to see the bootloader output | 16:55 |
marvin24 | I can disable it | 16:56 |
marvin24 | and re-enable when we have u-boot *and* some bl framebuffer addresss | 16:56 |
janimo | marvin24, probably disabling it completely makes sense for us indeed | 16:58 |
=== zyga is now known as zyga-afk | ||
* janimo is wondering how hard it is reflashing the ac100 to have the original layout where it had an SOS partition | 18:02 | |
janimo | debugging kernels would be easier that way | 18:02 |
infinity | janimo: Mine still has an SOS partition. I guess you could clone mine? | 18:05 |
janimo | infinity, I even had the full factory dump of the ssd on my disk, but the process itself seemed convoluted | 18:05 |
janimo | involving lots of nvflash calls | 18:06 |
infinity | Everything involves lots of nvflash calls... | 18:06 |
janimo | or hadn editing of some files | 18:06 |
janimo | yes, but this with weired argas thatn usual | 18:06 |
janimo | of course it may have just been the regular Ncommander fearmongering at work and all is easy-peasy | 18:06 |
* janimo just remembers that ad-ridden wiki where the process was confusingly written down | 18:08 | |
janimo | also ramconsole is great, that prompted me to think of having it available easily on the ac100 | 18:08 |
janimo | infinity, also I'd probably need the exact model I have for the factory images | 18:09 |
janimo | infinity, which is the easiest out of the archive solution for creating arm rootfs from c86? Anything better than deprecated rootstock? | 18:10 |
janimo | but higher level than debootstrap | 18:10 |
janimo | I don't seem to see an easy (as in linaro-media-create) tool for saying make an armhf root tarball corresponding to xubuntu-desktop optionally including other packages or kernel packages | 18:12 |
infinity | janimo: live-build, though we have no nice wrappers to make it brain-dead. | 18:12 |
janimo | infinity, oh does it do cross-builds too? | 18:13 |
infinity | It can. I've not tested it much, but I know Linaro uses it. | 18:13 |
janimo | hmm, I should probably ask them what they use to create the l-m-c tarballs | 18:13 |
infinity | Then again, you can just untar ubuntu-core, toss in /usr/bin/qemu-arm-static, and tailor it yourself quickly too. :P | 18:14 |
janimo | yes, the qemu + taylor it yourself bits what I hate | 18:14 |
janimo | :) | 18:14 |
janimo | I wonder if I am such a niche person with uncommon needs when I keep complaining about our build tools in general :) | 18:15 |
infinity | janimo: Well, what you want can be done with live-build, I just don't find it "easy" to quickly tailor up something non-standard. | 18:18 |
infinity | janimo: But, then again, I don't think any tool is easier than "chroot in and make it how you want it". Scripting that it only worth it if you do the same thing over and over. | 18:18 |
infinity | s/it only/is only/ | 18:18 |
janimo | infinity, well not very non-standard mind you, getting a xubuntu-desktop is quite a common use case I'd say | 18:18 |
infinity | Well, then you also said "optionally including other packages". ;) | 18:19 |
infinity | But, anything that isn't a base image set is "custom" in my mind. | 18:19 |
janimo | infinity, well what I did today was a chroot indeed, via mk-sbuild just because that seems a nice enough wrapper that handles cross arch without me knowing what the hell is exactly going on behind the scenes | 18:19 |
janimo | the whole debootstrap --foreign and 2 stages stuff is not something I want to learn about right now | 18:20 |
infinity | The only problem with mk-sbuild is that it does a few schroot-specific things to your chroot, plus adding some buildd cruft you don't need. | 18:20 |
infinity | Otherwise, yeah, that works. | 18:20 |
infinity | (But you could have started with core, which would save you from the --foreign business) | 18:20 |
janimo | infinity, well yes, extra packages is custom but well within what an apt call can do and common enough to warrant a tool. But yes, I am whining | 18:20 |
infinity | I suppose I just don't see how "chroot foo/ apt-get -y install xubuntu-desktop^" is any harder than trying to convince a tool to do that for you. | 18:21 |
janimo | as always, I would be content with good docs if tools are lacking, but somehow those too are non-existent, too dense or hard to find | 18:21 |
janimo | infinity, the cross-build part is where I am lost as I never did that | 18:22 |
janimo | and having read of xapt, dpkg-cross and a few other tools that in my mind overlap in functionality I got non-the-wiser | 18:22 |
infinity | There's no cross involved in the above. | 18:23 |
infinity | Just emulation. | 18:23 |
infinity | Which is also how live-build and higher level tools work. | 18:23 |
janimo | infinity, where does qemu-arm-static get involved? | 18:23 |
infinity | And mk-sbuild, for that matter. | 18:23 |
infinity | janimo: You just copy it into the chroot to /usr/bin | 18:23 |
janimo | yes, I saw ps whosing qemu while running mk-sbuild | 18:23 |
janimo | hmm, and then everything just gets ran by it? | 18:24 |
* janimo must try | 18:24 | |
infinity | So, "untar core chroot/ && cp /usr/bin/qemu-arm-static chroot/usr/bin/ && chroot chroot su -" | 18:24 |
infinity | janimo: If you install qemu-user-static on the host, binfmt_misc gets magically configured to throw all your non-native binaries at emulators. | 18:25 |
infinity | janimo: So, as long as your current root has the emulator, you win. | 18:25 |
janimo | yes, it has it, let's see | 18:25 |
infinity | (since interpreters are found by path, whether it's an emulator, ld.so, /bin/sh, etc) | 18:25 |
janimo | infinity, ok, running now in the chroot. I needed to create and chroot into it as sudo since it would not untar /dev nodes otherwise | 18:28 |
infinity | janimo: Well, yes. The above assumed you were root. | 18:29 |
=== k1l_ is now known as k1l |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!