Folks,
We are hosting another one of our regular Fedora 15 hardfp Virtual Fedora Activity Day today Friday August 26th 2011, at 14:00UTC (10:00 Eastern Daylight Time). The purpose of this session is to co-ordinate the continued bootstrap of F15 hardfp (hardware floating point).
Previously, we performed initial bootstrap (stage1), then proceeded to get RPM working natively within an ARMv7 hardfp environment (stage2), then got yum and mock working (stage3). We are now in what we are calling "stage4" in which we are rebuilding the world of packages using mock, but are not yet using Koji to do so - that will be stage5, the final stage before we are able to say that we have completed bringup.
We need helping building packages within a mock environment for F15. To do this, we have made your life as a (potential) volunteer somewhat more straightforward through the creation of a standard image that you can install onto an ARMv7 compatible board (tested on Panda, we believe this should also work on Beagle, and can be made to run on Tegra-2 boards like Trimslice with a replacement kernel - ask us for assistance). Once you have done this, the resulting system will contain a simple script that you can use to randomly retrieve a package that needs building and have it build without you having to even issue a single mock command :)
To participate, visit the following link:
http://scotland.proximity.on.ca/fedora-arm/f15hardfp/bootstrap/rootfs_stage4...
Follow the instructions contained within the readme file to extract the "uboot", "boot", and "root" archives onto an SD Card for your device. These instructions will destroy whatever is already installed. If you are familiar with using chroots, you can also retrieve the "root" image and chroot into it if you would prefer (after bind mounting /proc, /dev, etc.). We generally prefer not using chroots at this stage however.
Once you have followed the setup instructions:
0). Configure /etc/sysconfig/network and /etc/sysconfig/network-scripts/ifcfg-eth* according to your setup. The ones in that image statically assign "tenaya.bos.jonmasters.org".
1). Login to your board as root (password is "fedora"). Change the password if you would like to do so.
2). Become the "builder" user. You can either "su - builder" or you can "passwd builder", and then login remotely/on the console.
3). Optional, but recommended, use "screen -S builder" or similar to start a screen session that you can leave running.
4). Run the simple building script: ./arm-rebuild.sh
5). Monitor the arm mailing list. It is likely that a newer version of the builder script will be posted later today or before Monday. When that happens, we may change the SSH keys used in the build image, which would have the effect of forcing everyone to update. The purpose of that being to prevent any existing builders running the older script version. Instructions will be posted for updating a couple files will be posted.
Please join us on IRC: #fedora-arm on irc.freenode.org.
Further documentation will be provided on the Fedora ARM wiki today, however this email should provide sufficient setup information now.
Jon.
On 26 Aug 2011 10:51, "Jon Masters" jcm@redhat.com wrote:
Folks,
We are hosting another one of our regular Fedora 15 hardfp Virtual Fedora Activity Day today Friday August 26th 2011, at 14:00UTC (10:00 Eastern Daylight Time). The purpose of this session is to co-ordinate the continued bootstrap of F15 hardfp (hardware floating point).
Previously, we performed initial bootstrap (stage1), then proceeded to get RPM working natively within an ARMv7 hardfp environment (stage2), then got yum and mock working (stage3). We are now in what we are calling "stage4" in which we are rebuilding the world of packages using mock, but are not yet using Koji to do so - that will be stage5, the final stage before we are able to say that we have completed bringup.
We need helping building packages within a mock environment for F15. To do this, we have made your life as a (potential) volunteer somewhat more straightforward through the creation of a standard image that you can install onto an ARMv7 compatible board (tested on Panda, we believe this should also work on Beagle, and can be made to run on Tegra-2 boards
The builder I've enabled runs on the new OLPC within a chroot and that seems to work very well. I'm not sure if the kernel is a hfp one, but I don't think that was a requirement.
like Trimslice with a replacement kernel - ask us for assistance). Once you have done this, the resulting system will contain a simple script that you can use to randomly retrieve a package that needs building and have it build without you having to even issue a single mock command :)
To participate, visit the following link:
http://scotland.proximity.on.ca/fedora-arm/f15hardfp/bootstrap/rootfs_stage4...
Follow the instructions contained within the readme file to extract the "uboot", "boot", and "root" archives onto an SD Card for your device. These instructions will destroy whatever is already installed. If you are familiar with using chroots, you can also retrieve the "root" image and chroot into it if you would prefer (after bind mounting /proc, /dev, etc.). We generally prefer not using chroots at this stage however.
Once you have followed the setup instructions:
0). Configure /etc/sysconfig/network and /etc/sysconfig/network-scripts/ifcfg-eth* according to your setup. The ones in that image statically assign "tenaya.bos.jonmasters.org".
1). Login to your board as root (password is "fedora"). Change the password if you would like to do so.
2). Become the "builder" user. You can either "su - builder" or you can "passwd builder", and then login remotely/on the console.
3). Optional, but recommended, use "screen -S builder" or similar to start a screen session that you can leave running.
4). Run the simple building script: ./arm-rebuild.sh
5). Monitor the arm mailing list. It is likely that a newer version of the builder script will be posted later today or before Monday. When that happens, we may change the SSH keys used in the build image, which would have the effect of forcing everyone to update. The purpose of that being to prevent any existing builders running the older script version. Instructions will be posted for updating a couple files will be posted.
Please join us on IRC: #fedora-arm on irc.freenode.org.
Further documentation will be provided on the Fedora ARM wiki today, however this email should provide sufficient setup information now.
Jon.
-- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
On Fri, Aug 26, 2011 at 11:41 AM, Niels de Vos devos@fedoraproject.org wrote:
The builder I've enabled runs on the new OLPC within a chroot and that seems to work very well. I'm not sure if the kernel is a hfp one, but I don't think that was a requirement.
Excellent! The kernels we provide atm are not hfp afaik -- but it shouldn't make any difference.
I've found my XO-1.75s, attached to a USB HDD, to be an excellent build machine :-)
cheers,
m
On Fri, 2011-08-26 at 09:51 -0400, Jon Masters wrote:
To participate, visit the following link:
http://scotland.proximity.on.ca/fedora-arm/f15hardfp/bootstrap/rootfs_stage4...
Follow the instructions contained within the readme file to extract the "uboot", "boot", and "root" archives onto an SD Card for your device. These instructions will destroy whatever is already installed. If you are familiar with using chroots, you can also retrieve the "root" image and chroot into it if you would prefer (after bind mounting /proc, /dev, etc.). We generally prefer not using chroots at this stage however.
NOTE: The readme has been updated as of 1pm EDT today to correct a few issues. Please reload it. The current version should work on all OMAPs using x-loader, no matter how finnicky they are about partitioning.
Jon.
On Fri, 2011-08-26 at 09:51 -0400, Jon Masters wrote:
To participate, visit the following link:
http://scotland.proximity.on.ca/fedora-arm/f15hardfp/bootstrap/rootfs_stage4...
Just a quick update that we've had mock builders running around the clock and that, at last count we had built almost 9,000 native ARMv7 RPMs for the Fedora 15 bootstrap effort. I'll keep everyone updated. If you would like to contribute build cycles, please see my earlier mail, and ping us on #fedora-arm (irc.freenode.org) if you need assistance.
Hopefully, within a week or so we can consider switching to a full Koji environment, at which point we'll just have to do one more mass rebuild (stage5) before we /should/ have everything built correctly in Koji. We can do a preliminary alpha release of F15 soon thereafter. Folks are working on Anaconda, but I've no status update on that front yet.
Jon.
On Mon, Aug 29, 2011 at 4:06 AM, Jon Masters jcm@redhat.com wrote:
On Fri, 2011-08-26 at 09:51 -0400, Jon Masters wrote:
To participate, visit the following link:
http://scotland.proximity.on.ca/fedora-arm/f15hardfp/bootstrap/rootfs_stage4...
Just a quick update that we've had mock builders running around the clock and that, at last count we had built almost 9,000 native ARMv7 RPMs for the Fedora 15 bootstrap effort. I'll keep everyone updated. If you would like to contribute build cycles, please see my earlier mail, and ping us on #fedora-arm (irc.freenode.org) if you need assistance.
Why don't we just get it running in koji? Once we can get a armv5tel/armv7hl running in koji we don't need to have people off doing their own thing and we can start moving forward as a group.
Peter
On Mon, 2011-08-29 at 19:13 +0100, Peter Robinson wrote:
On Mon, Aug 29, 2011 at 4:06 AM, Jon Masters jcm@redhat.com wrote:
On Fri, 2011-08-26 at 09:51 -0400, Jon Masters wrote:
To participate, visit the following link:
http://scotland.proximity.on.ca/fedora-arm/f15hardfp/bootstrap/rootfs_stage4...
Just a quick update that we've had mock builders running around the clock and that, at last count we had built almost 9,000 native ARMv7 RPMs for the Fedora 15 bootstrap effort. I'll keep everyone updated. If you would like to contribute build cycles, please see my earlier mail, and ping us on #fedora-arm (irc.freenode.org) if you need assistance.
Why don't we just get it running in koji? Once we can get a armv5tel/armv7hl running in koji we don't need to have people off doing their own thing and we can start moving forward as a group.
We need a provisional set of repos to populate Koji. Arguably we might have enough now to get away with it, but we'd still wind up with lots of buildroot and false FTBFS type failures as Koji couldn't find deps. For better or worse, Dennis and I felt it was better to do a mock run first.
The Koji build will happen soon. We need to do one more build to make sure we have every build dep in place during builds. Although we suspect we're good even with stage4, we'd like to make sure we don't have packages failing to enable features during build by doing it again in what we call stage5. Then, we're done. Lots of nice things could be done to improve this in the future, like bootstrap deps, etc.
We'll keep an eye on builds this week. Some packages need huge resources to build, so they might be removed from the general list and built on a box we have that has a lot more RAM than many of the other builders.
Jon.
On Mon, Aug 29, 2011 at 7:22 PM, Jon Masters jcm@redhat.com wrote:
On Mon, 2011-08-29 at 19:13 +0100, Peter Robinson wrote:
On Mon, Aug 29, 2011 at 4:06 AM, Jon Masters jcm@redhat.com wrote:
On Fri, 2011-08-26 at 09:51 -0400, Jon Masters wrote:
To participate, visit the following link:
http://scotland.proximity.on.ca/fedora-arm/f15hardfp/bootstrap/rootfs_stage4...
Just a quick update that we've had mock builders running around the clock and that, at last count we had built almost 9,000 native ARMv7 RPMs for the Fedora 15 bootstrap effort. I'll keep everyone updated. If you would like to contribute build cycles, please see my earlier mail, and ping us on #fedora-arm (irc.freenode.org) if you need assistance.
Why don't we just get it running in koji? Once we can get a armv5tel/armv7hl running in koji we don't need to have people off doing their own thing and we can start moving forward as a group.
We need a provisional set of repos to populate Koji. Arguably we might have enough now to get away with it, but we'd still wind up with lots of buildroot and false FTBFS type failures as Koji couldn't find deps. For better or worse, Dennis and I felt it was better to do a mock run first.
The Koji build will happen soon. We need to do one more build to make sure we have every build dep in place during builds. Although we suspect we're good even with stage4, we'd like to make sure we don't have packages failing to enable features during build by doing it again in what we call stage5. Then, we're done. Lots of nice things could be done to improve this in the future, like bootstrap deps, etc.
We'll keep an eye on builds this week. Some packages need huge resources to build, so they might be removed from the general list and built on a box we have that has a lot more RAM than many of the other builders.
2 points here.
1) If the dep is missing it needs to be built..... that's the same whether its mock or koji.
2) need to do the same on armv5tel even if its for the core buildroot too. I don't see any movement on this.
Peter
On Monday, August 29, 2011 01:22:18 PM Jon Masters wrote:
On Mon, 2011-08-29 at 19:13 +0100, Peter Robinson wrote:
On Mon, Aug 29, 2011 at 4:06 AM, Jon Masters jcm@redhat.com wrote:
On Fri, 2011-08-26 at 09:51 -0400, Jon Masters wrote:
To participate, visit the following link:
http://scotland.proximity.on.ca/fedora-arm/f15hardfp/bootstrap/rootfs_ stage4_20110825/
Just a quick update that we've had mock builders running around the clock and that, at last count we had built almost 9,000 native ARMv7 RPMs for the Fedora 15 bootstrap effort. I'll keep everyone updated. If you would like to contribute build cycles, please see my earlier mail, and ping us on #fedora-arm (irc.freenode.org) if you need assistance.
Why don't we just get it running in koji? Once we can get a armv5tel/armv7hl running in koji we don't need to have people off doing their own thing and we can start moving forward as a group.
We need a provisional set of repos to populate Koji. Arguably we might have enough now to get away with it, but we'd still wind up with lots of buildroot and false FTBFS type failures as Koji couldn't find deps. For better or worse, Dennis and I felt it was better to do a mock run first.
The Koji build will happen soon. We need to do one more build to make sure we have every build dep in place during builds. Although we suspect we're good even with stage4, we'd like to make sure we don't have packages failing to enable features during build by doing it again in what we call stage5. Then, we're done. Lots of nice things could be done to improve this in the future, like bootstrap deps, etc.
We'll keep an eye on builds this week. Some packages need huge resources to build, so they might be removed from the general list and built on a box we have that has a lot more RAM than many of the other builders.
before we can consider building in koji we need a few things
1) armv5tel and armv7hl match in builods 2) rpm support for armv7hl is finalised 3) yum support for armv7hl is finalised 4) koji supports hardfp by seperating out the rpms into their own repo
koji support needs yum support and yum support needs rpm support
and we need to get started building things for armv5tel.
we can not really consider the stage3 rpms. they were build by hand, a few of with --nodeps to get around missing things.
find /srv/pub/fedora-arm/stage4/ -name *armv7hl.rpm|wc -l 6546
still quite a number of rpms but there is many missing. we should do a full run of things outside of koji, the main reason being that koji only allows us to build a nvr once. so to do it all in koji we would need to edit a crap load of spec files. to get rebuilding of things done.
Dennis
-
Dennis Gilmore -
Jon Masters -
Martin Langhoff -
Niels de Vos -
Peter Robinson