**** BEGIN LOGGING AT Wed Mar 07 02:59:58 2012 Mar 07 06:29:57 infinity, one way of solving the ghc issue is by not touching ghc at all, instead making llvm default to float-abi=hard on armhf. I don't know what the downside of that is, besides being Ubuntu specific change Mar 07 06:30:52 llvm's llc being a cross-compiler always - native being a particular subset - maybe it is good to always have explicit flags for almost everythinh Mar 07 06:54:47 janimo`: you can ask Sylvestre from Debian Mar 07 06:55:00 micahg, ok thanks Mar 07 06:55:05 janimo`: also, I have another Chromium upload today, so let's hope armhf builds this time Mar 07 06:55:37 micahg, oh nice, let's hope it does :) Mar 07 08:22:10 This would be nice to get to precise: http://sourceware.org/ml/libc-ports/2012-02/msg00079.html Mar 07 08:41:57 https://launchpad.net/ubuntu/+source/chromium-browser/17.0.963.66~r124982-0ubuntu1/+build/3265949 will be the entertainment for this morning :) Mar 07 09:25:26 ogra_, infinity : What /is/ the expected behavior on Ubuntu (in general) if the RTC clock has been reset (due to empty battery etc.)? Mar 07 09:40:45 sveinse, i belive there are various userspace apps that will have issues once you drop back to the epoch Mar 07 09:42:14 but i cant tell for sure ... Mar 07 09:43:22 what i know is that for example gdm has probs with autologin once the epoch kicks in, but since we switched to lightdm thats not an issue anymore for us (and i doubt anyone will look into it with much motivation, now that gdm is in universe) Mar 07 09:43:39 ogra_, I mean, is it expected that the system should start up normally? Mar 07 09:44:15 E.g. since there was a need for fixrtc in the first place, I mean Mar 07 09:44:28 no idea, really, i havent tried it .... and on i.e. the pandaboatd the "epoch" was picked by TI and defaults to 1990 ... so we wouldnt have these issues there at all Mar 07 09:44:46 fixrtc stems from a time where there was a bug in fsck Mar 07 09:45:17 that bug was fixed post lucid, but we kept fixrtc for the potentially broken userspace issues Mar 07 09:45:34 its just still around for historical reasons Mar 07 09:45:52 the initial bug that caused its existance is long gone Mar 07 09:45:58 yeah, so my mountall (perhaps fsck) lockup is worth investigating -- even on natty Mar 07 09:46:21 right Mar 07 09:46:43 though i doubt anyone will invest time into natty unlerss someone sends a patch that can be easily included Mar 07 09:46:58 simply because its EOL so soon Mar 07 09:48:03 yes, because the reason I'm bringing this up, is that we have been running with protoypes for over a year now. And we've never seen this lockup before. I can't honestly belive that all of these units have had valid rtc contents. Mar 07 09:48:28 So it could be a fsck-kernel regression of sorts. Mar 07 09:48:43 I'll take a look at it since we have to fix it anyways. I'll keep you in the loop Mar 07 09:50:34 While I'm at it: Our hw is headless, we just have a serial console. Yet I see mountall connecting to plymouth. How can I direct the output (like from fsck) from plymouth out on /dev/ttyO2 ? Mar 07 09:58:34 you should be able to use the text theme, it should just display on the default console you set on the kernel cmdline Mar 07 09:58:56 thanks Mar 07 10:00:09 if that doesnt work, ask the guys in #ubuntu-server, they definitely have experience with headless servers that run fsck on serial Mar 07 12:33:30 Laney, for ghc I assume you want a clean and properly upstreamable patch for armhf ftbfs not just any fix until upstream fixes it? Mar 07 12:54:32 janimo`: I don't know how long the upstreamable patch would take. I'm quite keen to get it fixed in the distro ASAP so if we need to do it in two stages then that's OK with me. Mar 07 12:54:53 I tried some stuff including using cpp on the Debian porterbox but nothing has stuck yet Mar 07 12:55:03 and the builds take so long which makes it quite frustrating Mar 07 12:55:10 Laney, worst case is put a sed in debian/rules, guarded by ifdef armhf :) Mar 07 12:55:29 yeah Mar 07 12:55:30 to append -float-abi=hard in that one .hs file. Mar 07 12:55:39 evil but matches your ASAP requierement Mar 07 12:55:49 I looked at what it takes to make it proper(er) Mar 07 12:56:01 and it needs changes across many haskell files Mar 07 12:56:06 I don't mind that. Or you could have a .patch that is applied and unapplied conditionally Mar 07 12:56:43 Laney, wait I keep forgetting to ask this. Is there support for debian/patches/foobar.patch.armhf ? Mar 07 12:57:05 if so, much cleaner than running sed :) Mar 07 12:57:10 apply a single line patch if armhf Mar 07 12:57:12 no(t that I know of) Mar 07 12:57:19 ah, shame Mar 07 12:57:29 it would be ifeq(armhf,$(DEB_HOST_ARCH))\n\tpatch ... Mar 07 12:57:45 and that patch would not be listed in series? Mar 07 12:57:52 indeed, managed separately Mar 07 12:58:02 it's not that nice but it makes the separation cleaner Mar 07 12:58:05 don't know if it's worth it Mar 07 12:58:07 not vastly cleaner than a search and replace then Mar 07 12:58:19 true. A bit more declarative Mar 07 12:58:54 anyway this issue blocks the large transition so the sooner it is resolved the better imho Mar 07 12:59:37 Laney, the sooner is calling sed then. Or I can try coming up with a patch - which if not what upstream would make it is still a patch. Mar 07 12:59:42 sed would unblock the transition Mar 07 12:59:51 and we could work on an upstreamable fix in the meantime Mar 07 12:59:57 OK, go for that then Mar 07 13:00:13 Should I try this an dupload in Ubuntu? Mar 07 13:00:40 yeah, if you're confident it will work Mar 07 13:00:54 i.e. if you don't feel the need to test build on the porterbox Mar 07 13:01:10 Laney, I test on my panda first just to be sure Mar 07 13:01:18 ok Mar 07 13:08:12 Laney, I see sed is no stranger to debian/rules in ghc, called a few times already Mar 07 13:09:03 janimo`: Just make sure whatever you do is also undone in the clean target, on the off chance that someone builds the source on an armhf box. Mar 07 14:00:35 infinity, good point wrt undoing. Not sure I understand what happens though if it is built on an armhf box? Mar 07 14:07:20 janimo`: If you build the source (ie: debuild -S) on an armhf box, the patched bits would stay patched and never get unpatched on, say, armel. Mar 07 14:07:35 janimo`: Hence why clean needs to undo whatever you do. Mar 07 14:08:36 infinity, if I debuild -S on x86 before uploading is this still an issue in this particular case? Mar 07 14:08:57 janimo`: Not if you never touched it on armhf in that source tree. Mar 07 14:09:19 janimo`: (ie: if the sed/patch never ran, then building the source on x86 would work around debian/rules being incorrect) Mar 07 14:09:34 janimo`: Still, clean should always return you to a pristine source, that's kinda the point. ;) Mar 07 14:09:50 (Not that there aren't plenty of broken source packages for which this isn't true) Mar 07 14:10:13 infinity, ok. I know clean should clean up, just forgot when I explciitly need to do something about it Mar 07 14:10:36 janimo`: Always. :P Mar 07 14:11:16 if I upload this change - which is not yet certain, depending on result on panda - it hopefully is a very shortlived hack, just to get other package sunblocked Mar 07 14:11:18 janimo`: But I haven't looked at ghc's debian/rules, it could be a lost cause anyway. If it's seding the source elsewhere, odds are the clean target's already wrong. Mar 07 14:11:38 (Unless clean very carefully reverses every single one of those) Mar 07 14:11:50 infinity, it is sedding ony in debian/tmp - that is less evil I guess. various install time changes in .desktop files and such from what I saw Mar 07 14:12:01 Ahh, yeah, that's fine. Mar 07 14:12:07 Since clean will wipe out debian/tmp anyway. Mar 07 14:12:29 Really, the best way to go, if it's a quiltish package, is just to have a patch, but don't have it in the series. Mar 07 14:12:51 And just patch manually iff armhf, and unconditionally (ignoring errors) unpatch in clean. Mar 07 14:12:56 nah, luckily sedding instead of patching is not a widespread habit. I am not advocating it either, but the cost of coming up in short time with a patch spanning multiple haskell fiels and horrid M4 code seems a bit higher now Mar 07 14:13:11 Heh. Mar 07 14:13:11 it is 3.0 so I think quiltish Mar 07 14:13:13 Fair enough. ;) Mar 07 14:13:35 only patching via vi is nastier (chort tied to testicles, etc) Mar 07 14:13:43 chord Mar 07 14:14:28 No one's stopping you from rgrep | xargs sed && dpkg-source --commit && remove from series. Mar 07 14:14:31 Ish. Mar 07 14:14:52 But anyhow, as you note, it's a short-term hack, doesn't matter deeply. Mar 07 14:15:35 And I guess the real fix needs ghc to take configure flags for float-abi, so it's not just always doing the one (wrong or right, depending) thing for armv7? Mar 07 14:16:09 fpc needs similar pain applied. Mar 07 14:16:48 right, the real fix would be - AFAICT - adding code to aclocal.m4 to pass a new flag to the build depending on _VFP_ABI (whatever is's called) Mar 07 14:17:07 and that flag being added in the haskell sources and one of the ARM Arch parameters Mar 07 14:17:40 or hmm, maybe a configure flag would be indeed easier than changing m4 - did not look at what exaclty that build system does Mar 07 14:18:31 but still unless the patch to haskell is a hack and piggybacks floatabi flag of AR ISA extensions (such as NEON, VFP) where it does not logically belong Mar 07 14:18:48 it means quite more than an oneliner in .hs files Mar 07 14:19:25 every pattern match on ArchARM changed to carry one more param, even if unused in all cases Mar 07 14:20:10 so I am not very encouraged to come up with a patch knowing almost for certain it should be rethought/rewritten to match upstream's taste and future plans in this area Mar 07 14:23:55 janimo`: Assuming upstream cares at all about the distinction, yeah. Mar 07 14:24:01 janimo`: Many upstream don't seem to. Mar 07 14:24:37 they have flags for NEON, VFP, VFPD16, preV7 ARM, preV6 ARM. I assume they care :) Mar 07 14:25:29 which reminds me, I should ping llvm debian maintainer. All this hassle would be avoided if llc just defaulted to hardfloat if called on a hardfloat system. Mar 07 14:25:44 at the cost of tripping up if crossbuilt from armel sometimes Mar 07 14:26:01 I wonder if llvm should have different defaults as gcc does in our default installs Mar 07 14:40:36 Probably, but I'm not terribly familiar with how llvm internals work. Mar 07 14:41:10 Individual packages clearly shouldn't have to know or care about what flags they're building with, though. The toolchain should default to something sane. Mar 07 15:56:58 micahg, at least chromium did not fail with V8/HARD flag missing this time. gles/egl missing is at least known territory Mar 07 15:59:47 did it fail again ? Mar 07 15:59:48 bah Mar 07 16:00:36 would it make sense to assume libgles2-mesa-dev [armel] is why it fails? Mar 07 16:00:42 ie. use armhf Mar 07 16:00:58 I'm not so versed on multi-arch, though Mar 07 16:01:13 kalikiana, that is exactly why it probably fails indeed Mar 07 16:01:24 I know I saw that before, not sure why I did not correct it then Mar 07 17:59:44 Hello! Mar 07 18:04:32 Can anyone suggest a good board/general resources to start on ARM? Mar 07 18:06:50 Several, depending on budget and what you want to accomplish. beagleboard.org has decent boards that are supported by Ubuntu, as does pandaboard.org. Also, Freescale has a similar platform called the Quickstart. Mar 07 18:08:45 All three are fully supported by Ubuntu with images. Any armv7 platform will run the Ubuntu software load as well, provided you can find a kernel. Mar 07 18:13:23 I went through the freescale, what's the difference between all the above you said and reaspberry pi? Mar 07 18:14:48 Main difference: We don't support the Raspberry PI due to architecture differences (it is not armv7 and would require a top-down rebuild of our arm pool - ~15k packages). Mar 07 18:15:19 We just don't currently have the resources. Mar 07 18:17:13 Hello. I'm new to Linux and trying to change the mux values in mux.c and mux44xx.c so the expansion ports J3 and J6 on the pandaboard will read GPIO rather than the standard setup. I haven't been able to figure out how to do this. I was in the pandaboard forum and they suggested I come over here for instructions. They said to start by getting a working build environment, but I don't actually know how to do that, as my Linux experien Mar 07 18:17:49 GrueMaster: Well, thanks for the input. I'll find about the other two manufacturer you said Mar 07 18:19:27 beagleboard.org has three different platforms (and three different price points). Panda is overall the best for most desktop usages (dual core, video acceleration (closed source), BT, wifi & ethernet). Mar 07 18:19:40 Freescale has native SATA. Mar 07 18:20:02 So it really depends on what you want to use it for. Mar 07 18:25:00 can anyone help me, or point me to where I can get help? Mar 07 18:25:30 It depends on the type of help. irc is a really big place. Mar 07 18:28:50 help with changing the the mux setup on linux on the pandaboard. Mar 07 18:29:01 the pandaboard irc sent me over here Mar 07 18:29:40 * GrueMaster goes to look at the pandaboard backscroll. Mar 07 18:33:14 MRSL: Best info I can give you is this: https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel Mar 07 18:33:44 As to gpio manipulation, that is #pandaboard. Mar 07 18:34:22 okay, thanks Mar 07 18:34:30 You can use your panda to recompile the kernel, but it can take a few hours. Cross-compiling on my Core2Quad takes ~20 minutes. Mar 07 18:35:56 how do you cross-compile? Mar 07 18:36:08 Sorry I can't help much more than that. For Ubuntu kernel specific questions, you might get more help on #ubuntu-kernel. We can help with the arm environment here. I know it is a PITA to bounce around IRC like this. Mar 07 18:36:34 The wiki pointer I gave you has specific instructions for the omap4 kernel cross-compile near the bottom. Mar 07 18:37:27 alright thanks, I'll head there. And, its better than before, when I was absolutely stuck Mar 07 18:37:55 What are you running on your desktop? Mar 07 18:40:11 Currently the board is running on its own. Ubuntu 10.4 runs on the desktop that I used to originally boot the board Mar 07 18:42:59 One way to get a working cross-compile environment on your desktop is to download http://cdimage.ubuntu.com/ubuntu-core/daily/current/precise-core-i386.tar.gz and use it as a chroot. That's what I am doing (and I also have Lucid on my desktop). Mar 07 18:45:07 Okay. On a side note, using the command from the wiki page to get the source gives an error saying that 'You muust put some 'source' URIs in your sources.list'. Do you know what that means? Mar 07 18:46:56 Yes. You will need to edit /etc/apt/sources.list. The easiest thing is to copy the lines that start with "deb " and change the copies to "deb-src". Mar 07 18:47:12 Then run "sudo apt-get update" Mar 07 18:48:55 where is /ect? I tried to access the file and it says the directory doesn't exist. Mar 07 18:54:38 /etc not /ect. Mar 07 18:56:17 Bugger. Now I feel thick. Yay for sleep depravation!!! Mar 07 18:56:50 Heh. Mar 07 19:01:37 hi Mar 07 19:02:36 I am trying to create a custom root fs .. for FriendlyARM Tiny 6410 ... and facing kernel panic ... any direction on where to look will be very helpful Mar 07 19:03:05 Is it based on ubuntu? Mar 07 19:05:17 Alright, it started to download the source. so thanks again. Mar 07 19:05:33 had run the ubuntu image, but given the limited capabilities of the board ubuntu was very slow ... so I was trying to get the simplest setup Mar 07 19:05:46 GrueMaster: had run the ubuntu image, but given the limited capabilities of the board ubuntu was very slow ... so I was trying to get the simplest setup Mar 07 19:06:53 GrueMaster: is there a way to strip the gui from ubuntu arm ... I mean strip it to tis bare skeleton ... so that my low end board would be able to use it Mar 07 19:07:03 if it is at least armv7, you can use the ubuntu-core image as a start for your rootfs. Mar 07 19:07:33 Or you could use an ubuntu-server image. They are headless and will even run on a Beagle C4. Mar 07 19:09:46 GrueMaster: okay, thanks Mar 07 19:12:17 GrueMaster: its arm 11 Mar 07 19:12:29 http://www.friendlyarm.net/products/tiny6410 Mar 07 19:13:00 Ouch. We don't support anything lower than armv7 in our images. All binaries are built for this. Mar 07 19:13:21 GrueMaster: so I start by downloading the ubuntu-core image Mar 07 19:13:35 It won't run on arm 11. Mar 07 19:14:06 so v7 is higher than arm 11 Mar 07 19:14:37 confusing arithmetic :P ... okay ... anyways .. any workarounds? Mar 07 19:14:47 Arm 11 is armv6 (iirc) architecture. Arm numbers were very confusing in the past. Mar 07 19:15:11 And no, we can't work around this without a complete rebuild. Mar 07 19:16:14 Won't happen any time soon. Not enough resources to rebuild & retest. We are having a hard enough time fixing all the broken packages for our current builds. Mar 07 19:17:10 What I have been trying to do is get a system up with just busybox ... and my aim is to run python on the board and be able to use usb dongles for wifi and bluetooth on it Mar 07 19:17:33 GrueMaster, any directions on how I should proceed Mar 07 19:18:37 You will probably want to use debian for binary packages. We simply can't help here. Beyond that, I have no suggestions. Mar 07 19:21:14 Would it be completely crazy to try and build everything from source ... or would such an action drive me crazy via kernel panics ... what do you suggest? Mar 07 19:23:35 GrueMaster: Mar 07 19:24:59 If you were to rebuild from source, again you would have to use something like Debian as our source is set to build armel & armhf, both of which are defined as armv7 in our build tools. Mar 07 19:25:25 And at this point you are no longer using ubuntu. Which is why I suggested debian. Mar 07 19:32:22 GrueMaster: where on debian's site exactly do I go looking for its arm port ? I apologize if the intellectual standard of the questions is low ... Mar 07 19:32:45 I don't know. Try asking in #debian. Mar 07 19:36:18 anuvrat: if you want a minimal embedded system you should consider using openembedded Mar 07 19:37:27 anuvrat: if you want to use ubuntu try ubuntu-core Mar 07 19:38:01 ... friendly arm thats armv5 right? Mar 07 19:40:08 xranby_ac100: ubuntu-core is armv7 only. Mar 07 19:43:37 anuvrat: i installed ubuntu from http://elinux.org/BeagleBoardUbuntu . I then install x11 and use ~/.xinitrc to launch my app when i startx Mar 07 19:44:28 anuvrat: which friendlyarm board do you got? Mar 07 19:44:40 i notice that some of their boards are based on a cortex-a8 Mar 07 19:45:14 like the http://www.friendlyarm.net/products/tiny210 this board can run ubuntu-core Mar 07 19:45:38 i know on my xm the precise images boot to initramfs, havent had time to debug yet Mar 07 19:46:26 xranby_ac100: looks like similar hardware specs to the beagle xm Mar 07 20:38:38 xranby_ac100: smplman, GrueMaster I have FriendlyARM Tiny 6410 http://www.friendlyarm.net/products/tiny6410 Mar 07 20:42:45 janimo`: seems that one of the build-deps of armel only, that should be [armel armhf] as kalikiana pointed out to me, I'll fix for the next upload **** ENDING LOGGING AT Thu Mar 08 02:59:58 2012