Re: [Sugar-devel] SoaS on the XO progress
On Sat, Jan 24, 2009 at 2:12 AM, Marco Pesenti Gritti marc...@sugarlabs.org wrote: On Fri, Jan 23, 2009 at 10:41 PM, Chris Ball c...@laptop.org wrote: D'oh, I think OLPC asked Warren to push that X driver update, and it sounds like we didn't test it properly. Is there a good way to select which version of an RPM to use in kickstart, maybe? Which package are you referring to here exactly? I'm confused because F10 and rawhide seem to have the same versions of X packages, and joyride works but F10 doesn't. Hm, reverting from 2.11 to 2.10 actually fixes it. So I don't get why 2.11 works in joyride. Marco ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
On Sat, Jan 24, 2009 at 2:12 AM, Marco Pesenti Gritti marc...@sugarlabs.org wrote: On Fri, Jan 23, 2009 at 10:41 PM, Chris Ball c...@laptop.org wrote: D'oh, I think OLPC asked Warren to push that X driver update, and it sounds like we didn't test it properly. Is there a good way to select which version of an RPM to use in kickstart, maybe? Which package are you referring to here exactly? I'm confused because F10 and rawhide seem to have the same versions of X packages, and joyride works but F10 doesn't. Using joyride xorg.conf doesn't help either. Maybe some kind of interaction with the kernel? /me has no clue... Marco ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
Marco Pesenti Gritti wrote: On Sat, Jan 24, 2009 at 2:12 AM, Marco Pesenti Gritti marc...@sugarlabs.org wrote: On Fri, Jan 23, 2009 at 10:41 PM, Chris Ball c...@laptop.org wrote: D'oh, I think OLPC asked Warren to push that X driver update, and it sounds like we didn't test it properly. Is there a good way to select which version of an RPM to use in kickstart, maybe? Which package are you referring to here exactly? I'm confused because F10 and rawhide seem to have the same versions of X packages, and joyride works but F10 doesn't. Hm, reverting from 2.11 to 2.10 actually fixes it. So I don't get why 2.11 works in joyride. Marco I just wanted to add that it's possible to select a specific version in a kickstart file. Though, as far as I know, it's a revisor-only feature. You can run revisor and add the --kickstart-exact-nevra option then: --kickstart-exact-nevra Force Revisor to interpret the package manifest as complete package nevra (name, epoch, version, release and architecture). Implies --kickstart-exact 'yum install revisor' should do the trick [1]. ;) --Sebastian [1] https://fedorahosted.org/revisor/ ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
On Wed, Jan 21, 2009 at 04:51:04AM +0100, Marco Pesenti Gritti wrote: 2 X fades a couple of times and then hangs the system. Describe this further? -- James Cameronmailto:qu...@us.netrek.org http://quozl.netrek.org/ ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
On Thu, 2009-01-22 at 18:01 -0500, Chris Ball wrote: Hi Marco, Some extra steps: * Clone git://git.fedoraproject.org/spin-kickstarts * If you want F10 apply spin.patch * Apply live.patch in /usr/lib/python2.5/site-packages/imgcreate * sudo yum -y install livecd-tools mtd-utils crcimg * If you're building on F10, s/ext4/ext3/ in fedora-live-base.ks * sudo livecd-creator --cache=cache -c fedora-livecd-desktop.ks I'm still not able to get the images I create to boot on the XO -- could you confirm that you're using the Fedora kernel and initrd? I wonder what else could be different between our setups.. Thanks! - Chris. Chris what are you using for the /boot/olpc.fth file? I have luck using the below, in that file, on a usb-drive: \ Boot script for USB boot hex rom-pa fffc7 + 4 $number drop h# 2e19 [if] patch 2drop erase claim-params : high-ramdisk ( -- ) cv-load-ramdisk h# 22c +lp l@ 1+ memory-limit umin /ramdisk - .f000 and ( new-ramdisk-adr ) ramdisk-adr over /ramdisk move ( new-ramdisk-adr ) to ramdisk-adr ; ' high-ramdisk to load-ramdisk [then] : set-bootpath-dev ( -- ) /chosen find-package if ( phandle ) bootpath rot get-package-property 0= if ( propval$ ) get-encoded-string ( bootpath$ ) [char] \ left-parse-string 2nip( dn$ ) dn-buf place( ) then then /sd dn-buf count sindex 0= if sd: else u: then BOOTPATHDEV $set-macro ; set-bootpath-dev text ks=hd:LABEL=XSRepo:/ks.cfg to boot-file ${BOOTPATHDEV}\boot\initrd.img expand$ to ramdisk ${BOOTPATHDEV}\boot\vmlinuz expand$ to boot-device unfreeze boot The above code is found in the latest livecd-tools and is added if you pass --XO on the cmdline. I'm booting the xs-05 installer with the above code, you would have to edit:to boot-file for that boot arguments you want and the vmlinuz and initrd.img lines or make links pointing to the real files. Jerry ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
Sayamindu Dasgupta wrote: FWIW, I have noticed mmap errors while trying to deal with large files (~70MB) on the standard OLPC builds. localedef does not work in the XO for this (strace shows that it chokes when trying to mmap /usr/lib/locale/locale-archive) I don't think it's related to the size of the file. Writeable file mappings are just not supported by jffs2, and cause mmap() to return an error. glibc likes to do it when building the locale-archive, and I vaguely remember I had a workaround for this in our fork of the glibc rpm. -- // Bernie Innocenti - http://www.codewiz.org/ \X/ Sugar Labs - http://www.sugarlabs.org/ ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
C. Scott Ananian wrote: It's probably worth reading through the pilgrim 'streams.d/olpc-development.stream' file to see if there are other fixes you are missing. Indeed. Starting with a white-room F10 build is going to cause many such regressions, and re-discovering all the associated workarounds is going to cost a lot of time. -- // Bernie Innocenti - http://www.codewiz.org/ \X/ Sugar Labs - http://www.sugarlabs.org/ ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
On Fri, Jan 23, 2009 at 12:01 AM, Chris Ball c...@laptop.org wrote: I'm still not able to get the images I create to boot on the XO -- could you confirm that you're using the Fedora kernel and initrd? Yup, sure. The only difference is that the initrd contains more modules, because of the livecd-tools change I posted as a patch. Here is my image: http://www.sugarlabs.org/~marco/f10.img http://www.sugarlabs.org/~marco/f10.crc Could you give it a try? That would tell us if it's something different in the build environment or something else... Marco ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
Yay, stock F10 is now running GNOME fine. Some more things I found: * I was not using -a when copying the files, so owners and all kind of other stuff was not preserved. That's the cause of the hal failure, new script attached. * With selinux enabled you cannot login from the console. Perhaps jffs2/selinux issues? I'm just disabling it for now. -selinux --enforcing +selinux --disabled * rpm does not work. It complains about mmap failure when reading the db, jffs2 related? * X doesn't actually crash the system, it's just very broken :) If you switch to vt and back you can also see some parts of the gdm screen. It looks definitely like a regression in the updates. If I build an image with updates disabled X works fine. Marco livecd-iso-to-xo.sh Description: Bourne shell script ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
On Fri, Jan 23, 2009 at 10:47 PM, Marco Pesenti Gritti marc...@sugarlabs.org wrote: Yay, stock F10 is now running GNOME fine. Some more things I found: * I was not using -a when copying the files, so owners and all kind of other stuff was not preserved. That's the cause of the hal failure, new script attached. * With selinux enabled you cannot login from the console. Perhaps jffs2/selinux issues? I'm just disabling it for now. -selinux --enforcing +selinux --disabled * rpm does not work. It complains about mmap failure when reading the db, jffs2 related? FWIW, I have noticed mmap errors while trying to deal with large files (~70MB) on the standard OLPC builds. localedef does not work in the XO for this (strace shows that it chokes when trying to mmap /usr/lib/locale/locale-archive) -sdg- -- Sayamindu Dasgupta [http://sayamindu.randomink.org/ramblings] ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
On Fri, Jan 23, 2009 at 12:17 PM, Marco Pesenti Gritti marc...@sugarlabs.org wrote: * rpm does not work. It complains about mmap failure when reading the db, jffs2 related? As far as I know, jffs2 doesn't support writable mmaps. In the debian ports, we add some special magic to tell apt not to use them ( http://wiki.laptop.org/go/Installing_Debian_as_an_upgrade ); and the following code from pilgrim appears to do something similar: echo - Configuring RPM to not use writable mmap (as jffs2 don't support it) mkdir -p $INSTALL_ROOT/etc/rpm cat EOF $INSTALL_ROOT/etc/rpm/macros.rpmdb %__dbi_cdb create private nommap EOF It's probably worth reading through the pilgrim 'streams.d/olpc-development.stream' file to see if there are other fixes you are missing. --scott -- ( http://cscott.net/ ) ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
Hi Marco, Got it working now. livecd-creator should be run on x86_64 as: sudo setarch i386 livecd-creator ... * X doesn't actually crash the system, it's just very broken :) If you switch to vt and back you can also see some parts of the gdm screen. It looks definitely like a regression in the updates. If I build an image with updates disabled X works fine. D'oh, I think OLPC asked Warren to push that X driver update, and it sounds like we didn't test it properly. Is there a good way to select which version of an RPM to use in kickstart, maybe? I tried building Rawhide, and got the error you mentioned, linked below. Running bash-3.2# mount -t jffs2 mtd0 /mnt works, though, so I'm not sure what's up / why it's not creating /dev/root properly. Jeremy, any ideas? http://dev.laptop.org/~cjb/f11-boot Thanks, - Chris. -- Chris Ball c...@laptop.org ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
On Fri, Jan 23, 2009 at 6:42 PM, C. Scott Ananian csc...@laptop.org wrote: On Fri, Jan 23, 2009 at 12:17 PM, Marco Pesenti Gritti marc...@sugarlabs.org wrote: * rpm does not work. It complains about mmap failure when reading the db, jffs2 related? As far as I know, jffs2 doesn't support writable mmaps. In the debian ports, we add some special magic to tell apt not to use them ( http://wiki.laptop.org/go/Installing_Debian_as_an_upgrade ); and the following code from pilgrim appears to do something similar: echo - Configuring RPM to not use writable mmap (as jffs2 don't support it) mkdir -p $INSTALL_ROOT/etc/rpm cat EOF $INSTALL_ROOT/etc/rpm/macros.rpmdb %__dbi_cdb create private nommap EOF Good catch, thanks! Marco ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
On Fri, Jan 23, 2009 at 10:41 PM, Chris Ball c...@laptop.org wrote: D'oh, I think OLPC asked Warren to push that X driver update, and it sounds like we didn't test it properly. Is there a good way to select which version of an RPM to use in kickstart, maybe? Which package are you referring to here exactly? I'm confused because F10 and rawhide seem to have the same versions of X packages, and joyride works but F10 doesn't. Marco ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
Hi Marco, Some extra steps: * Clone git://git.fedoraproject.org/spin-kickstarts * If you want F10 apply spin.patch * Apply live.patch in /usr/lib/python2.5/site-packages/imgcreate * sudo yum -y install livecd-tools mtd-utils crcimg * If you're building on F10, s/ext4/ext3/ in fedora-live-base.ks * sudo livecd-creator --cache=cache -c fedora-livecd-desktop.ks I'm still not able to get the images I create to boot on the XO -- could you confirm that you're using the Fedora kernel and initrd? I wonder what else could be different between our setups.. Thanks! - Chris. -- Chris Ball c...@laptop.org ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
Hi Marco, Hello, I spent some time trying to get Sugar on a stick images (which are basically livecd-tools based Fedora spins) running on the XO from nand. Converting the image to jffs2 and adding cafe_nand and jffs2 to the initrd was enough to make it boot. You're awesome! Thanks! 1 haldaemon fails to start and I couldn't find any error log. Maybe running it directly, rather than via init, will expose the problem? 2 X fades a couple of times and then hangs the system. Happy to look into this. I can reproduce 2 if I write the same image to an usb stick using Fedora XO livecd-iso-to-disk. Could be either a regression in the F10 updates or something wrong in the way SoaS images are built. I have absolutely no idea about 1, but it would seem to be related to jffs2/initrd, since it works fine from the usb stick. It looks like you're making F10 spins, then? I wonder what happens if you try a Rawhide spin instead. Could you share your instructions and code? I'd like to give this technique a try for our F11 rebase. (At which point the difference between OLPC's image and SoaS will be very small. Maybe there'll be no difference at all, even. ;-) Thanks! - Chris. -- Chris Ball c...@laptop.org ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
1 haldaemon fails to start and I couldn't find any error log. Maybe running it directly, rather than via init, will expose the problem? When running it directly it was sitting for a while and then exiting without any error (even in verbose mode). I should strace it. It looks like you're making F10 spins, then? I wonder what happens if you try a Rawhide spin instead. Could you share your instructions and code? I'd like to give this technique a try for our F11 rebase. I did it in a very manual/hacky way yesterday. Working right now on doing it more cleanly, scripting a little and documenting, will post something later tonight. (At which point the difference between OLPC's image and SoaS will be very small. Maybe there'll be no difference at all, even. ;-) Right, that's where I want to get :) Marco ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel
Re: [Sugar-devel] SoaS on the XO progress
Hi Marco, Same results as yesterday with F10 while F11 fails to /init (both images built from an F10). I gave this a try, and I get a hang after Loading ramdisk image.. in OFW when booting a Rawhide image built on a Rawhide/x86-64 host. Will try with a serial port connected tomorrow. Thanks, - Chris. -- Chris Ball c...@laptop.org ___ Sugar-devel mailing list Sugar-devel@lists.sugarlabs.org http://lists.sugarlabs.org/listinfo/sugar-devel