For my work I'm trying to package objectweb-asm3 for EPEL7 (and asked the current maintainer to grant me co-maintainer privileges).
However, in just doing a scratch build on koji to see if it builds, I get the following error in root.log:
DEBUG util.py:283: Getting requirements for objectweb-asm3-3.3.1-10.el7.src DEBUG util.py:283: --> ant-1.9.2-9.el7.noarch DEBUG util.py:283: --> maven-local-3.4.1-6.el7_0.noarch DEBUG util.py:283: Error: No Package found for objectweb-asm3
How do you build a package in koji when the package requires itself to be available?
On Wed, 27 Aug 2014 14:17:30 -0400 "Darryl L. Pierce" mcpierce@gmail.com wrote:
For my work I'm trying to package objectweb-asm3 for EPEL7 (and asked the current maintainer to grant me co-maintainer privileges).
However, in just doing a scratch build on koji to see if it builds, I get the following error in root.log:
DEBUG util.py:283: Getting requirements for objectweb-asm3-3.3.1-10.el7.src DEBUG util.py:283: --> ant-1.9.2-9.el7.noarch DEBUG util.py:283: --> maven-local-3.4.1-6.el7_0.noarch DEBUG util.py:283: Error: No Package found for objectweb-asm3
How do you build a package in koji when the package requires itself to be available?
hm, there is usually a bootstrap flag in such situations, but can't see it here ...
Dan
On 08/27/2014 08:17 PM, Darryl L. Pierce wrote:
How do you build a package in koji when the package requires itself to be available?
I seen a lot of packages where it was needed for %check section. If this is the situation (you have to investigate that). Just comment out that BR and %check section. Build it. Then uncomment and build again. Of course if you can put there %boostrap macro, you will easy work for others.
On Thu, 28 Aug 2014 13:55:14 +0200 Miroslav Suchý msuchy@redhat.com wrote:
On 08/27/2014 08:17 PM, Darryl L. Pierce wrote:
How do you build a package in koji when the package requires itself to be available?
I seen a lot of packages where it was needed for %check section. If this is the situation (you have to investigate that). Just comment out that BR and %check section. Build it. Then uncomment and build again. Of course if you can put there %boostrap macro, you will easy work for others.
And you might try building an older version that didn't have yet the dependency first, then build the latest. You can also talk to the maintainers, because they should know how to resolve the cyclic dependency properly
Dan
2014-08-28 14:05 GMT+02:00 Dan Horák dan@danny.cz:
On Thu, 28 Aug 2014 13:55:14 +0200 Miroslav Suchý msuchy@redhat.com wrote:
On 08/27/2014 08:17 PM, Darryl L. Pierce wrote:
How do you build a package in koji when the package requires itself to be available?
I seen a lot of packages where it was needed for %check section. If this is the situation (you have to investigate that). Just comment out that BR and %check section. Build it. Then uncomment and build again. Of course if you can put there %boostrap macro, you will easy work for others.
And you might try building an older version that didn't have yet the dependency first, then build the latest. You can also talk to the maintainers, because they should know how to resolve the cyclic dependency properly
So, a lot of options.
Wouldn't it be better if we had a packaging guideline forcing the spec to either have something like a %bootstrap macro, or at least contain a comment _documenting_ how to bootstrap the package?
- Thomas
On 08/28/2014 05:17 PM, Thomas Moschny wrote:
Wouldn't it be better if we had a packaging guideline forcing the spec to either have something like a %bootstrap macro, or at least contain a comment_documenting_ how to bootstrap the package?
I totally agree.
On Thu, Aug 28, 2014 at 05:17:53PM +0200, Thomas Moschny wrote:
2014-08-28 14:05 GMT+02:00 Dan Horák dan@danny.cz:
On Thu, 28 Aug 2014 13:55:14 +0200 Miroslav Suchý msuchy@redhat.com wrote:
On 08/27/2014 08:17 PM, Darryl L. Pierce wrote:
How do you build a package in koji when the package requires itself to be available?
I seen a lot of packages where it was needed for %check section. If this is the situation (you have to investigate that). Just comment out that BR and %check section. Build it. Then uncomment and build again. Of course if you can put there %boostrap macro, you will easy work for others.
And you might try building an older version that didn't have yet the dependency first, then build the latest. You can also talk to the maintainers, because they should know how to resolve the cyclic dependency properly
So, a lot of options.
Wouldn't it be better if we had a packaging guideline forcing the spec to either have something like a %bootstrap macro, or at least contain a comment _documenting_ how to bootstrap the package?
+1 on the guideline.
On Thu, Aug 28, 2014 at 02:05:59PM +0200, Dan Horák wrote:
On Thu, 28 Aug 2014 13:55:14 +0200 Miroslav Suchý msuchy@redhat.com wrote:
On 08/27/2014 08:17 PM, Darryl L. Pierce wrote:
How do you build a package in koji when the package requires itself to be available?
I seen a lot of packages where it was needed for %check section. If this is the situation (you have to investigate that). Just comment out that BR and %check section. Build it. Then uncomment and build again. Of course if you can put there %boostrap macro, you will easy work for others.
And you might try building an older version that didn't have yet the dependency first, then build the latest. You can also talk to the maintainers, because they should know how to resolve the cyclic dependency properly
I reached out to him a few weeks ago and haven't heard anything back yet.
On 08/28/2014 01:55 PM, Miroslav Suchý wrote:
On 08/27/2014 08:17 PM, Darryl L. Pierce wrote:
How do you build a package in koji when the package requires itself to be available?
I seen a lot of packages where it was needed for %check section. If this is the situation (you have to investigate that). Just comment out that BR and %check section.
Examples? At least in some cases these likely are packaging bugs.
Ralf
On 08/28/2014 03:49 PM, Ralf Corsepius wrote:
Examples? At least in some cases these likely are packaging bugs.
I was looking for those examples and I realized I lied a little. So corrections:
I saw a lot of examples where BR had *circular* dependencies because of %check section.
For example whole Ruby On Rails stack. Huge amount of packages needs some rubygems for %check, but those rubygems need that original gem as BR of itself. E.g: rubygem-tilt, rubygem-thor, rubygem-fomatador, rubygem-rspec-expectations, rubygem-mocks, rubygem-rack-protection, rubygem-diff-lcs, rubygem-sinatra.
On 08/28/2014 05:27 PM, Miroslav Suchý wrote:
On 08/28/2014 03:49 PM, Ralf Corsepius wrote:
Examples? At least in some cases these likely are packaging bugs.
I was looking for those examples and I realized I lied a little. So corrections:
I saw a lot of examples where BR had *circular* dependencies because of %check section.
OK. this is not unusual. The "bootsrap" rpm flag trick might be a proper work-around for these
Note you'll have to build all packages twice. Once with a "bootstrap flag" set during a "bootstrap" from scratch and a second time without "bootstrap flag" set, when incrementally building.
(Initially I was suspecting some testsuites being run against older, installed packages and not against the just having been built versions. We've had some of these cases, in the past.).
For example whole Ruby On Rails stack. Huge amount of packages needs some rubygems for %check, but those rubygems need that original gem as BR of itself. E.g: rubygem-tilt, rubygem-thor, rubygem-fomatador, rubygem-rspec-expectations, rubygem-mocks, rubygem-rack-protection, rubygem-diff-lcs, rubygem-sinatra.
My wild guess would be ruby likely is in a similar situation as perl (which applies the bootstrap trick).
Ralf
packaging@lists.fedoraproject.org
-
Dan Horák -
Darryl L. Pierce -
Miroslav Suchý -
Ralf Corsepius -
Thomas Moschny