Re: unicode error with mergerepos against CentOS 5.
2009/3/27 Seth Vidal : > > > On Fri, 27 Mar 2009, Steve Traylen wrote: > >> On Wed, Mar 18, 2009 at 2:23 PM, Steve Traylen wrote: >>> >>> On Wed, Mar 18, 2009 at 2:18 PM, Seth Vidal >>> wrote: On Wed, 18 Mar 2009, Steve Traylen wrote: >> >> the metadata you're pulling in was written with a much older >> createrepo >> which was less picky about some of the data it wrote back out. So you >> end >> up >> with some pkgs with dodgy data in the changelogs. > > Mirroring in CentOS5 and then recreating from scratch the repodirs with > a > Fedora > 10 createrepo the the repository is created spitting out some warnings. What ver of yum and createrepo do you have when you're running both createrepo and mergerepo? >>> >>> $ rpm -qf /usr/bin/yum \ >>> /usr/bin/mergerepo \ >>> /usr/bin/createrepo \ >>> /usr/libexec/kojid/mergerepos >>> yum-3.2.21-2.fc10.noarch >>> createrepo-0.9.6-3.fc10.noarch >>> createrepo-0.9.6-3.fc10.noarch >>> koji-builder-1.3.1-1.fc10.noarch >>> >> >> Hi Seth, >> >> Is there any workaround or anything for this? >> >> Steve > > I doubt it'll help much - but try 0.9.7 and 3.2.22 that are in rawhide. We > fixed a few more unicode explosions but I'm sure we're still missing some > > -sv Same with rawhide. Workaround , removing by hand the copyright symbol present in primary.xml. recorded here: http://createrepo.baseurl.org/ticket/3 > > -- > Fedora-buildsys-list mailing list > Fedora-buildsys-list@redhat.com > https://www.redhat.com/mailman/listinfo/fedora-buildsys-list > -- Steve Traylen -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: Error from repoview when composing with pungi.
On Wed, 18 Feb 2009, seth vidal wrote: On Wed, 2009-02-18 at 16:43 -0500, seth vidal wrote: On Wed, 2009-02-18 at 14:35 -0700, William F. Acker WB2FLW +1 303 722 7209 wrote: I back leveled sqlite on the build machine which didn't help. I know that's not conclusive since some packages such as Anaconda are downloaded from the repo and never cached. So, before putting the old sqlite into the repo, I checked to see the when I was last successful VS when I installed sqlite. It turns out that both yum and sqlite were installed on 2/5, and my last successful spin was on 2/8. I don't see any packages installed since them that are obviously connected to this problem. I'm not sure if this is sqlite anymore - I've been mucking with it today. Oddly on rawhide here I can't make it happen. ah ha! My sample size was too small. I found out what was causing the problem somehow we had two packages VLGothic-fonts and vlgothic-fonts - repoview case normalizes for the .html files it makes - hence this problem. Fedora was not supposed to have those two pkgs - it was an oversight and that has been fixed. I'll also change repoview to not case normalize the filenames. The new repoview did the trick. On F9 and F10, the install media built quite nicely. I haven't installed from them yet, but I don't see any reason why there'd be a problem with that. Thanks a gig! -- Bill in Denver -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: bind-mounting files in mock
On Friday, March 27 2009, Jesse Keating said: > On Fri, 2009-03-27 at 10:37 -0500, Clark Williams wrote: > > It's not described in the man page, so I'm a bit leery about depending > > on that behavior. That being said, it's not a normal use-case for mock > > builds, so it'll probably only bite the livecd creation stuff it the > > bind-mount behavior changes. > > > > Ok, I'll pull it in for the next release. > > I need that functionality too when building install images. Rather than > see more and different code paths when chroot generating, I think there > would be some value in using the same code path regardless of how the > chroot is used. I think the bind mounting of dev/ is only there for the > loop entries, I don't know of any other mock consumers that require a > real /dev/ tree. We might consider just always file bind mounting a few > loop entries in and only having one code path. You can't really just do a few as you may need however many there are on the real system. If the idea is to make them by default it's probably better to just make n of them based on the host. But then the chroot does have a way of affecting outside of it :/ Jeremy -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: bind-mounting files in mock
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Fri, 27 Mar 2009 08:57:17 -0700 Jesse Keating wrote: > On Fri, 2009-03-27 at 10:37 -0500, Clark Williams wrote: > > It's not described in the man page, so I'm a bit leery about depending > > on that behavior. That being said, it's not a normal use-case for mock > > builds, so it'll probably only bite the livecd creation stuff it the > > bind-mount behavior changes. > > > > Ok, I'll pull it in for the next release. > > I need that functionality too when building install images. Rather than > see more and different code paths when chroot generating, I think there > would be some value in using the same code path regardless of how the > chroot is used. I think the bind mounting of dev/ is only there for the > loop entries, I don't know of any other mock consumers that require a > real /dev/ tree. We might consider just always file bind mounting a few > loop entries in and only having one code path. That works for me. I'm currently trying to figure out the regression with 3rd party repos, so if you have the time and can gin up a patch, that'd be great; if not I'll work on it after I figure the BZ out. Clark -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.9 (GNU/Linux) iEYEARECAAYFAknNF3QACgkQHyuj/+TTEp2kZgCfdrRSiGQLWukV3OT03EXFosZv 8VEAoLmdboXfdgVafNi7SYgfz4qnc+KZ =qdbd -END PGP SIGNATURE- -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: unicode error with mergerepos against CentOS 5.
On Fri, 27 Mar 2009, Steve Traylen wrote: On Wed, Mar 18, 2009 at 2:23 PM, Steve Traylen wrote: On Wed, Mar 18, 2009 at 2:18 PM, Seth Vidal wrote: On Wed, 18 Mar 2009, Steve Traylen wrote: the metadata you're pulling in was written with a much older createrepo which was less picky about some of the data it wrote back out. So you end up with some pkgs with dodgy data in the changelogs. Mirroring in CentOS5 and then recreating from scratch the repodirs with a Fedora 10 createrepo the the repository is created spitting out some warnings. What ver of yum and createrepo do you have when you're running both createrepo and mergerepo? $ rpm -qf /usr/bin/yum \ /usr/bin/mergerepo \ /usr/bin/createrepo \ /usr/libexec/kojid/mergerepos yum-3.2.21-2.fc10.noarch createrepo-0.9.6-3.fc10.noarch createrepo-0.9.6-3.fc10.noarch koji-builder-1.3.1-1.fc10.noarch Hi Seth, Is there any workaround or anything for this? Steve I doubt it'll help much - but try 0.9.7 and 3.2.22 that are in rawhide. We fixed a few more unicode explosions but I'm sure we're still missing some -sv -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: bind-mounting files in mock
On Fri, 2009-03-27 at 10:37 -0500, Clark Williams wrote: > It's not described in the man page, so I'm a bit leery about depending > on that behavior. That being said, it's not a normal use-case for mock > builds, so it'll probably only bite the livecd creation stuff it the > bind-mount behavior changes. > > Ok, I'll pull it in for the next release. I need that functionality too when building install images. Rather than see more and different code paths when chroot generating, I think there would be some value in using the same code path regardless of how the chroot is used. I think the bind mounting of dev/ is only there for the loop entries, I don't know of any other mock consumers that require a real /dev/ tree. We might consider just always file bind mounting a few loop entries in and only having one code path. -- Jesse Keating Fedora -- Freedom² is a feature! identi.ca: http://identi.ca/jkeating signature.asc Description: This is a digitally signed message part -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: bind-mounting files in mock
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Wow, that's bizarre (but works). :) It's not described in the man page, so I'm a bit leery about depending on that behavior. That being said, it's not a normal use-case for mock builds, so it'll probably only bite the livecd creation stuff it the bind-mount behavior changes. Ok, I'll pull it in for the next release. Clark On Fri, 27 Mar 2009 10:39:54 -0400 Jay Greguske wrote: > Hi Clark, > > If you're bind mounting a file, you have to bind it to an existing file, > not a directory. Try touching /dev/myloop first, and then mounting the > real loop device to it. > > Thanks, > -jay > > > Clark Williams wrote: > > -BEGIN PGP SIGNED MESSAGE- > > Hash: SHA1 > > > > On Wed, 18 Mar 2009 14:52:35 -0400 > > Jay Greguske wrote: > > > > > >> Hello, > >> > >> While trying to get livecd-creator working in a mock-built chroot, I > >> discovered that only directories could be bind-mounted using the > >> bind_mount plugin. I made a few code changes and attached a patch for > >> your consideration that enables the bind-mounting of files. Like > >> directories, the bind-mounting of files will require > >> 'internal_dev_setup' be set to False, otherwise the mount command will > >> fail. > >> > >> In the chroot configuration file the following syntax would be allowed > >> (which looks just like that which was used for bind-mounting directories): > >> > >> config_opts['plugin_conf']['bind_mount_opts']['files'].append(('/dev/loop0', > >> > >> '/dev/loop0')) > >> > >> Please let me know what you think, and thanks in advance for your time! > >> > >> - Jay > >> > > > > Jay, > > > > Finally got some time to pull in mock patches and work on bugs. Sorry > > it's taken this long... > > > > I'm not seeing how your stuff would work. A bind-mount only works on > > directories, not files. If I try to bind-mount /dev/loop0 somewhere by > > hand I get the following: > > > > $ mkdir -p /tmp/foo && cd /tmp/foo > > $ sudo mount -n --bind /dev/loop0 . > > mount: Not a directory > > $ mkdir dev > > $ sudo mount -n --bind /dev/loop0 dev > > mount: Not a directory > > > > What am I missing here? > > > > Clark > > -BEGIN PGP SIGNATURE- > > Version: GnuPG v2.0.9 (GNU/Linux) > > > > iEYEARECAAYFAknM5GYACgkQHyuj/+TTEp1CWQCgsPClxT4qV2XCQx3oyHihhBlE > > t9UAoIekmO6wJCgp9T+OwukEguLJ21Qz > > =63u9 > > -END PGP SIGNATURE- > > > -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.9 (GNU/Linux) iEYEARECAAYFAknM8swACgkQHyuj/+TTEp2i/QCfZa1y4KBC2z4A3ZP9+rO8muZF Q3gAoItg2iK5NMf9vnfVLgaT4BSTWhLC =j3Sw -END PGP SIGNATURE- -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: bind-mounting files in mock
Hi Clark, If you're bind mounting a file, you have to bind it to an existing file, not a directory. Try touching /dev/myloop first, and then mounting the real loop device to it. Thanks, -jay Clark Williams wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Wed, 18 Mar 2009 14:52:35 -0400 Jay Greguske wrote: Hello, While trying to get livecd-creator working in a mock-built chroot, I discovered that only directories could be bind-mounted using the bind_mount plugin. I made a few code changes and attached a patch for your consideration that enables the bind-mounting of files. Like directories, the bind-mounting of files will require 'internal_dev_setup' be set to False, otherwise the mount command will fail. In the chroot configuration file the following syntax would be allowed (which looks just like that which was used for bind-mounting directories): config_opts['plugin_conf']['bind_mount_opts']['files'].append(('/dev/loop0', '/dev/loop0')) Please let me know what you think, and thanks in advance for your time! - Jay Jay, Finally got some time to pull in mock patches and work on bugs. Sorry it's taken this long... I'm not seeing how your stuff would work. A bind-mount only works on directories, not files. If I try to bind-mount /dev/loop0 somewhere by hand I get the following: $ mkdir -p /tmp/foo && cd /tmp/foo $ sudo mount -n --bind /dev/loop0 . mount: Not a directory $ mkdir dev $ sudo mount -n --bind /dev/loop0 dev mount: Not a directory What am I missing here? Clark -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.9 (GNU/Linux) iEYEARECAAYFAknM5GYACgkQHyuj/+TTEp1CWQCgsPClxT4qV2XCQx3oyHihhBlE t9UAoIekmO6wJCgp9T+OwukEguLJ21Qz =63u9 -END PGP SIGNATURE- -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: bind-mounting files in mock
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Wed, 18 Mar 2009 14:52:35 -0400 Jay Greguske wrote: > Hello, > > While trying to get livecd-creator working in a mock-built chroot, I > discovered that only directories could be bind-mounted using the > bind_mount plugin. I made a few code changes and attached a patch for > your consideration that enables the bind-mounting of files. Like > directories, the bind-mounting of files will require > 'internal_dev_setup' be set to False, otherwise the mount command will fail. > > In the chroot configuration file the following syntax would be allowed > (which looks just like that which was used for bind-mounting directories): > > config_opts['plugin_conf']['bind_mount_opts']['files'].append(('/dev/loop0', > '/dev/loop0')) > > Please let me know what you think, and thanks in advance for your time! > > - Jay Jay, Finally got some time to pull in mock patches and work on bugs. Sorry it's taken this long... I'm not seeing how your stuff would work. A bind-mount only works on directories, not files. If I try to bind-mount /dev/loop0 somewhere by hand I get the following: $ mkdir -p /tmp/foo && cd /tmp/foo $ sudo mount -n --bind /dev/loop0 . mount: Not a directory $ mkdir dev $ sudo mount -n --bind /dev/loop0 dev mount: Not a directory What am I missing here? Clark -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.9 (GNU/Linux) iEYEARECAAYFAknM5GYACgkQHyuj/+TTEp1CWQCgsPClxT4qV2XCQx3oyHihhBlE t9UAoIekmO6wJCgp9T+OwukEguLJ21Qz =63u9 -END PGP SIGNATURE- -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list
Re: unicode error with mergerepos against CentOS 5.
On Wed, Mar 18, 2009 at 2:23 PM, Steve Traylen wrote: > On Wed, Mar 18, 2009 at 2:18 PM, Seth Vidal wrote: >> >> >> On Wed, 18 Mar 2009, Steve Traylen wrote: >> the metadata you're pulling in was written with a much older createrepo which was less picky about some of the data it wrote back out. So you end up with some pkgs with dodgy data in the changelogs. >>> >>> Mirroring in CentOS5 and then recreating from scratch the repodirs with a >>> Fedora >>> 10 createrepo the the repository is created spitting out some warnings. >> >> What ver of yum and createrepo do you have when you're running both >> createrepo and mergerepo? > > $ rpm -qf /usr/bin/yum \ > /usr/bin/mergerepo \ > /usr/bin/createrepo \ > /usr/libexec/kojid/mergerepos > yum-3.2.21-2.fc10.noarch > createrepo-0.9.6-3.fc10.noarch > createrepo-0.9.6-3.fc10.noarch > koji-builder-1.3.1-1.fc10.noarch > Hi Seth, Is there any workaround or anything for this? Steve > > >> >> -sv >> >> -- >> Fedora-buildsys-list mailing list >> Fedora-buildsys-list@redhat.com >> https://www.redhat.com/mailman/listinfo/fedora-buildsys-list >> > > > > -- > Steve Traylen > -- Steve Traylen -- Fedora-buildsys-list mailing list Fedora-buildsys-list@redhat.com https://www.redhat.com/mailman/listinfo/fedora-buildsys-list