**** BEGIN LOGGING AT Sun Feb 14 02:59:57 2010 Feb 14 12:19:29 lool Feb 14 12:19:59 ogra Feb 14 12:32:11 saeed Feb 14 12:35:48 hi armin Feb 14 13:22:53 saeed: Hey Feb 14 14:18:52 as you know, the suspend button worked for me Feb 14 14:19:30 but when I convert it to KEY_POWER , and configure the g-p-m to "Ask Me" when power pressed Feb 14 14:19:39 nothing happens Feb 14 14:20:09 when I configure the g p m to suspend when power pressed, it works Feb 14 14:45:35 saeed: So, g-p-m is capturing your button press, but not launching the GUI when configured to "Ask Me"? Feb 14 15:06:13 saeed: In a terminal, killall gnome-power-manager and then run gnome-power-manager --verbose Feb 14 15:06:32 saeed: That should give you some debugging info on what GPM is seeing Feb 14 15:52:25 persia yes exactly Feb 14 16:08:54 What is everybody using to build for ARM? Are there particular compilers that are optimized for ARM? Feb 14 16:09:18 codesourcery's Feb 14 16:12:16 armin76: fastest? Feb 14 17:07:46 therealgalen: We use native compilation in Ubuntu: compiling ARM on ARM. Feb 14 17:29:45 persia: I don't mean the hardware so much as the software - the compiler Feb 14 17:29:45 a lot Feb 14 17:29:55 of vendors claim to have "faster" or "better" compilers Feb 14 17:30:31 Ah. I don't think Ubuntu claims faster or better, but I know a lot of work has gone into making it appropriate for ARMv7+Thumb2 Feb 14 17:31:13 But no special secret sauce: most of the changes are upstream, so I suspect anyone could grab them if they felt it was a competition. Feb 14 17:31:55 * persia is more concerned about making the software work than about the compiler being especially fast Feb 14 17:32:17 persia: architecture-specific issues, or just general development? Feb 14 17:32:27 Both :) Feb 14 17:33:30 Although a lot of what I personally do is actually neither, but rather just things like improving developer tools to make it easier to do those, or reviewing cases where something didn't build for reasons involving how the distribution is developed, etc. Feb 14 17:36:11 well, i am looking at a variety of companies offering gcc alternatives claimed to be faster; i was hoping to hear from people who have tried them Feb 14 17:40:50 therealgalen: its known to be the most optimized, yes Feb 14 17:41:35 armin76: are there any benchmarks *not* from the vendor online? Feb 14 17:42:05 armin76: optimised for which? execution time, code size, compilation time, etc.? Feb 14 17:42:33 * persia is fairly certain it's not possible to optimise for those three things simultaneously, and suspects there are other similar variables. Feb 14 17:43:45 execution performance primarily; compilation time does not matter, code size does not matter Feb 14 17:47:36 persia: no clue, its just what they tend to say Feb 14 17:47:47 persia: in fact they tend to say that vanilla gcc sucks Feb 14 17:47:54 therealgalen: no clue, sorry Feb 14 17:48:50 therealgalen: From what I understand, it really depends on which code you need so optimised. Feb 14 17:49:07 Try a few different toolchains, and see which one compiles your application in the fastest way. Feb 14 17:53:21 hmm, ok Feb 14 17:53:34 persia: most of the functions are kernel stuff, ath9k, ec. Feb 14 17:54:33 The kernel gets kinda fussy about opimisation. Generally, I think it's better to walk the code path and optimise in code than try to adjust toolchains for that class of issues. Feb 14 17:54:49 Just because the kernel code makes some assumptions about how it will be compiled. Feb 14 19:35:14 oh heya, where do I get the ARM toolchain for cross-compiling stuff? Feb 14 19:35:25 apt-cache search gcc doesn't show anything related to ARM. Feb 14 19:36:28 We don't cross-compile :) Feb 14 19:36:42 we just rebuilt everything natively. Feb 14 19:37:10 How'd you get the first one up, then? =þ Feb 14 19:37:23 Used Debian's compilation. Feb 14 19:37:34 they stole it :D Feb 14 19:37:49 * persia prefers to think of it as "sharing" Feb 14 19:53:38 * DanaG wishes somebody would make something like that "touchbook" but with the new Marvell 1.2GHz thingy. Feb 14 19:56:40 The worst embedded thing I've ever (tried to) use: Microblaze. It was so bad, that building the cross-compiler made the HOST compiler segfault! Feb 14 20:04:57 hmm, what systems do you run the armel compilation on? Feb 14 20:06:43 I'm not sure precisely. I suspect there's at least one package that shows it in the build log, but :) Feb 14 20:06:51 I do know that they are ARMv7 machines. Feb 14 20:07:18 Personally, I do test-compilation on a Sharp Netwalker, but we only upload sources. Feb 14 20:13:32 Oh yeah, worst distro fail I've ever seen was Angstrom on a Zaurus SL-5500. I can understand drivers not working... but to have your kernel config for that board not enable the ALREADY-IN-KERNEL drivers for that board? That's just epic fail. Feb 14 20:13:54 I wish I could get Ubuntu running on the thing. Feb 14 20:16:38 The 5500? That's too old. Feb 14 20:16:53 Jaunty should run on the 1000, 1100, 3000, 3100, 3200, and 3300 Feb 14 20:16:58 But karmic or lucid won't. Feb 14 20:18:33 My gripe was that they discontinued OpenZaurus in favor of the supposedly superior "Angstrom" -- but then seemingly didn't even TRY to make the old hardware work. Feb 14 20:19:20 Having it work slowly and poorly... fine.... but disabling in-kernel touchscreen driver? fail. Feb 14 20:19:32 Anyway, now I'm using a beagleboard, and that's cool. Feb 14 20:19:39 Well, Sharp discontinued the device, which may have had something to do with that. Feb 14 20:19:56 You're running lucid on your beagleboard? Feb 14 20:20:00 Yeah. Feb 14 20:20:21 Still, they could've at least set CONFIG_UCB1100_TS true, or whatever that was. Feb 14 20:20:27 anyway, enough bashing Angstrom. Feb 14 20:20:44 I'm using Ubuntu on it because it'll be just like my main development environment. Feb 14 20:20:57 Too big for the onboard flash, though -- I'm using a 2-gig SD card. Feb 14 20:21:07 hmm, is there anything useful I COULD use the onboard flash for? Feb 14 20:21:30 I think we still need a slightly better way to dispatch test-builds to other machines though. I have to manually scp sources to armel or powerpc devices from my laptop to build them. Feb 14 20:21:38 But other than that, I think it works just fine. Feb 14 20:21:56 How big is the onboard flash? Feb 14 20:22:45 I think it's just 256 megs. Feb 14 20:22:49 i didn't know the bb had internal flash Feb 14 20:22:54 And u-boot takes up some of that. Feb 14 20:23:14 That's tight. You might be able to stuff your kernel and initramfs in there. Feb 14 20:23:29 But I don't think you can do anything advanced with it. Feb 14 20:23:30 Might as well just leave it on initramfs. Feb 14 20:24:27 now, this would be cool, if not for the tiny battery: http://eeepc.net/superthin-netbook-runs-on-marvell-armada-510/ Feb 14 20:26:45 Oh yeah, I did have a lot of trouble getting the usb host-to-device networking working.... it was encoding the packets badly (off-by-one error). Feb 14 20:26:56 Nobody seems to want to put a full-size battery in with a low-power device. I really wouldn't mind an extra 100g for an extra 10 hours of power. Feb 14 20:27:11 yeah, that sucks. Feb 14 20:27:38 My network issue: I set guest device to mac address 00:11:22:33:44:55, and the parent saw "Ethernet II" frames from address ff:ff:00:11:22:33 sending packets of type 4455. Feb 14 20:27:48 Well, it probably also costs an extra 12,000円 for that extra 100g, which makes it perceived as less competitive. Feb 14 20:28:01 What currency unit is that? Feb 14 20:28:02 Wah! That's just not right :) Feb 14 20:28:16 I finally fixed it by passing g_ether.use_eem=0 Feb 14 20:28:20 Oh, right. That gets used in multiple contexts. JPY is what I was thinking. Feb 14 20:28:22 then it used full cdc-ether instead of cdc-subset. Feb 14 20:32:15 Hi. What libc implementation is being used in Ubuntu ARM? Feb 14 20:32:42 windows! Feb 14 20:33:14 jonnor: eglibc Feb 14 20:34:25 armin76: cool, thanks! Feb 14 20:35:48 heh, note to self: ext4 as root on flash means I can't use ext2load. Feb 14 21:12:22 init: ureadahead main process (41) terminated with status 5 mountall: Could not connect to Plymouth init: procps main process (369) terminated with status 255 init: ureadahead-other main process (466) terminated with status 4 Feb 14 21:12:26 that's on the beagleboard. Feb 14 21:29:47 seems your kernel misses a good bunch of options and plymouth isnt installed Feb 14 21:31:08 none of these issues should stop it from finishing the boot though Feb 14 21:32:02 * ogra goes back to weekendish stuff Feb 14 21:34:54 init: ureadahead main process (40) terminated with status 5 Feb 14 21:35:10 oh yeah, I don't plan to use the video output for much (at least for now), anyway. Feb 14 21:35:40 Also weird: the default user's named group is not 1000 like it is on the host system. Feb 14 21:39:04 beagleboard has this: Feb 14 21:39:05 mtdparts: mtdparts=nand:512k(x-loader),1920k(u-boot),128k(u-boot-env),4m(kernel),-(fs) Feb 14 22:40:29 !find pkg-config Feb 14 22:40:34 DanaG: Found: pkg-config Feb 14 22:40:39 =þ Feb 14 22:40:52 odd that command-not-found doesn't seem to work on ARM. Feb 14 22:51:06 !find dbus-glib-1 Feb 14 22:51:07 DanaG: Found: libdbus-glib-1-2, libdbus-glib-1-2-dbg, libdbus-glib-1-dev, libdbus-glib-1-doc Feb 15 00:55:19 hi Feb 15 00:59:00 i tried to build a rootfs for qemu but everytime i get a Kernel panic - not syncing: Attempted to kill init! Feb 15 01:02:15 How are you trying to build the rootfs? Feb 15 01:03:22 like in https://wiki.ubuntu.com/ARM/RootfsFromScratch Feb 15 01:04:46 sudo rootstock --fqdn ubuntu --login ubuntu --password ubuntu Feb 15 01:05:56 and then mkdir /tmp/arm Feb 15 01:05:56 sudo mount -o loop /tmp/arm.img /tmp/arm Feb 15 01:05:56 cd /tmp/arm Feb 15 01:05:56 sudo tar xzvf ../qemu-arm-rootfs.tgz Feb 15 01:05:56 cd .. Feb 15 01:05:56 sudo umount /tmp/arm Feb 15 01:05:58 rm /tmp/arm Feb 15 01:06:30 qemu with Feb 15 01:06:33 sudo qemu-system-arm -M versatilepb -kernel /tmp/vmlinuz-2.6.28-versatile -cpu arm926 -hda /tmp/arm.img -m 256M -append "root=/dev/sda mem=256M" Feb 15 01:08:00 That certainly looks right. Feb 15 01:08:01 * persia tries Feb 15 01:11:03 oh and qemu-img create arm.img 1G Feb 15 01:11:03 sudo mkfs.ext2 -F arm.img Feb 15 01:11:38 I seem to be bumping into a bug that prevents rootstock from running at all today. Sorry. Feb 15 01:11:44 Someone else may have to help. Feb 15 01:13:37 thanks anyway! Feb 15 01:22:10 the rootfs from http://people.canonical.com/~ogra/arm/qemu/qemu-arm-rootfs.tgz works.... Feb 15 01:35:23 i give up for today Feb 15 01:35:31 bye **** ENDING LOGGING AT Mon Feb 15 02:59:57 2010