Greetings,
I was just testing the latest 'master' branch, and it seems that the 'rpms' Makefile target is broken. Along with the change to support posting bodhi comments (5d37e59128c1913946c02e6a1d899a5576f37801), we set the group of fas.conf to the user 'autotest'. On my primary system, I don't have autotest installed, so the 'make rpms' target fails as follows:
make[1]: Leaving directory `/tmp/jlaska-rpm/BUILD/autoqa-0.4.3' + install -m 644 autoqa.conf repoinfo.conf /tmp/jlaska-rpm/BUILDROOT/autoqa-0.4.3-1.fc14.x86_64/etc/autoqa/ + install -m 640 -g autotest fas.conf /tmp/jlaska-rpm/BUILDROOT/autoqa-0.4.3-1.fc14.x86_64/etc/autoqa/ install: invalid group `autotest' error: Bad exit status from /tmp/jlaska-rpm/TMP/rpm-tmp.Hj4RAu (%install)
An easy fix would be to add "BuildRequires: autotest" to the spec file, but for some reason that doesn't seem right. I can't think of any, but should we consider alternatives to adding a BuildRequires for autotest?
Thanks, James
----- Original Message -----
Greetings,
I was just testing the latest 'master' branch, and it seems that the 'rpms' Makefile target is broken. Along with the change to support posting bodhi comments (5d37e59128c1913946c02e6a1d899a5576f37801), we set the group of fas.conf to the user 'autotest'. On my primary system, I don't have autotest installed, so the 'make rpms' target fails as follows:
make[1]: Leaving directory `/tmp/jlaska-rpm/BUILD/autoqa-0.4.3'
- install -m 644 autoqa.conf repoinfo.conf
/tmp/jlaska-rpm/BUILDROOT/autoqa-0.4.3-1.fc14.x86_64/etc/autoqa/
- install -m 640 -g autotest fas.conf
/tmp/jlaska-rpm/BUILDROOT/autoqa-0.4.3-1.fc14.x86_64/etc/autoqa/ install: invalid group `autotest' error: Bad exit status from /tmp/jlaska-rpm/TMP/rpm-tmp.Hj4RAu (%install)
An easy fix would be to add "BuildRequires: autotest" to the spec file, but for some reason that doesn't seem right. I can't think of any, but should we consider alternatives to adding a BuildRequires for autotest?
Thanks, James
I don't have experience with building RPMs, maybe we can ask on #devel what the correct approach is? From my laic point of view we require autotest group to be present when building package and therefore autotest is our build requirement (is there any harm in that?).
We could also probably change the fas.conf's group in postinstall script, but that would break the 'make clean install' functionality.
On Thu, 2011-01-06 at 02:59 -0500, Kamil Paral wrote:
----- Original Message -----
Greetings,
I was just testing the latest 'master' branch, and it seems that the 'rpms' Makefile target is broken. Along with the change to support posting bodhi comments (5d37e59128c1913946c02e6a1d899a5576f37801), we set the group of fas.conf to the user 'autotest'. On my primary system, I don't have autotest installed, so the 'make rpms' target fails as follows:
make[1]: Leaving directory `/tmp/jlaska-rpm/BUILD/autoqa-0.4.3'
- install -m 644 autoqa.conf repoinfo.conf
/tmp/jlaska-rpm/BUILDROOT/autoqa-0.4.3-1.fc14.x86_64/etc/autoqa/
- install -m 640 -g autotest fas.conf
/tmp/jlaska-rpm/BUILDROOT/autoqa-0.4.3-1.fc14.x86_64/etc/autoqa/ install: invalid group `autotest' error: Bad exit status from /tmp/jlaska-rpm/TMP/rpm-tmp.Hj4RAu (%install)
An easy fix would be to add "BuildRequires: autotest" to the spec file, but for some reason that doesn't seem right. I can't think of any, but should we consider alternatives to adding a BuildRequires for autotest?
Thanks, James
I don't have experience with building RPMs, maybe we can ask on #devel what the correct approach is? From my laic point of view we require autotest group to be present when building package and therefore autotest is our build requirement (is there any harm in that?).
Until now it wasn't a build requirement, but with the fas.conf addition, you are right, this might be the most sensible approach. Not sure why I'm hesitant to do it though. Unless something better comes along, we'll go with that for the next release.
We could also probably change the fas.conf's group in postinstall script, but that would break the 'make clean install' functionality.
I don't think that would cut the package review mustard.
Thanks for the input! James
Not that I'd know anything about packaging, but would not be something like (pseudocode):
dirty_user = False if not exists_user(autotest): create_user(autotest) dirty_user = True
# do magical build stuff
if dirty_user: remove_user(autotest)
So you have 'the thing' in makefile, and you don't need to add additional build dep?
joza
----- Original Message ----- From: "James Laska" jlaska@redhat.com To: "AutoQA development" autoqa-devel@lists.fedorahosted.org Sent: Thursday, January 6, 2011 2:31:32 PM Subject: Re: 'rpm' make target failing
On Thu, 2011-01-06 at 02:59 -0500, Kamil Paral wrote:
----- Original Message -----
Greetings,
I was just testing the latest 'master' branch, and it seems that the 'rpms' Makefile target is broken. Along with the change to support posting bodhi comments (5d37e59128c1913946c02e6a1d899a5576f37801), we set the group of fas.conf to the user 'autotest'. On my primary system, I don't have autotest installed, so the 'make rpms' target fails as follows:
make[1]: Leaving directory `/tmp/jlaska-rpm/BUILD/autoqa-0.4.3'
- install -m 644 autoqa.conf repoinfo.conf
/tmp/jlaska-rpm/BUILDROOT/autoqa-0.4.3-1.fc14.x86_64/etc/autoqa/
- install -m 640 -g autotest fas.conf
/tmp/jlaska-rpm/BUILDROOT/autoqa-0.4.3-1.fc14.x86_64/etc/autoqa/ install: invalid group `autotest' error: Bad exit status from /tmp/jlaska-rpm/TMP/rpm-tmp.Hj4RAu (%install)
An easy fix would be to add "BuildRequires: autotest" to the spec file, but for some reason that doesn't seem right. I can't think of any, but should we consider alternatives to adding a BuildRequires for autotest?
Thanks, James
I don't have experience with building RPMs, maybe we can ask on #devel what the correct approach is? From my laic point of view we require autotest group to be present when building package and therefore autotest is our build requirement (is there any harm in that?).
Until now it wasn't a build requirement, but with the fas.conf addition, you are right, this might be the most sensible approach. Not sure why I'm hesitant to do it though. Unless something better comes along, we'll go with that for the next release.
We could also probably change the fas.conf's group in postinstall script, but that would break the 'make clean install' functionality.
I don't think that would cut the package review mustard.
Thanks for the input! James
_______________________________________________ autoqa-devel mailing list autoqa-devel@lists.fedorahosted.org https://fedorahosted.org/mailman/listinfo/autoqa-devel
On Fri, 2011-01-07 at 08:19 -0500, Josef Skladanka wrote:
Not that I'd know anything about packaging, but would not be something like (pseudocode):
dirty_user = False if not exists_user(autotest): create_user(autotest) dirty_user = True # do magical build stuff if dirty_user: remove_user(autotest)So you have 'the thing' in makefile, and you don't need to add additional build dep?
Thanks for the suggestion. That would definitely do the job. I think the wrinkle comes because multiple packages are not supposed (or allowed, I'm not sure) to create/remote the same user/groups. We'd end up with the logic you reference above in both autoqa and autotest spec files (I think). I suspect package reviewers would frown on that and recommend we just rely on autotest for it's user/group definition.
http://fedoraproject.org/wiki/Packaging:UsersAndGroups
Thanks, James
autoqa-devel@lists.fedorahosted.org