Java SIG has agreed on submitting new Java Packaging Guidelines draft for FPC review/vote: https://fedoraproject.org/wiki/User:Akurtakov/JavaPackagingDraftUpdate
Log from the meeting: http://meetbot.fedoraproject.org/fedora-meeting-1/2012-10-02/java_sig.2012-10-02-15.00.log.html
Diff from current guidelines: https://fedoraproject.org/w/index.php?title=User:Akurtakov/JavaPackagingDraftUpdate&diff=305374&oldid=275277
Since the diff is relatively big a summary: 1. No longer require 2+ jars to be in subdirectory (there was no technical need) 2. Add standardization for compatibility packages 3. Remove no longer needed parts about Maven 2 4. Improve add_maven_depmap documentation 5. Add suggestions for pom_ macros instead of patching 6. installation/use of J2EE APIs standardization (initial version) 7. JNI guidelines simplification & examples
FPC will wait on Java draft until an answer for FESCo ticket 961 is received
https://fedorahosted.org/fesco/ticket/961 => FESCo voted to allow java to not be multilib.
Following up on the discussion in today's FPC meeting, I've sent a message to the packaging list to discuss the overarching question of whether arch-specific files in noarch packages should be allowed to use %{_prefix}/lib rather than %{_libdir}
http://lists.fedoraproject.org/pipermail/packaging/2012-November/008757.html
Hopefully we'll hash this out this week and can make progress on both that generally and the Java guidelines specifically at next week's meeting.
As a general policy, the following was approved:
"If a package is explicitly not multilib enabled, then it may use %{_prefix}/lib instead of %{_libdir}."
(+1:6, 0:0, -1:1)
Additionally, the Java draft was approved:
A slight amendment to the wording of the general policy:
"If a package has been granted an explicit exception from FESCo to permit it to not be multilib enabled, then it may use %{_prefix}/lib instead of %{_libdir}."
Due to wording of Java guidelines, can we consider all Java JNI packages to be allowed non-multilib? I would guess so mainly due to https://fedorahosted.org/fesco/ticket/961
Yes. We considered that FESCo exception to apply to all Java JNI packages.
During FPC's yesterday's meeting the question of "how does multi-arched Ubuntu handle this issue" has popped up.
From what I can gather from a short glimpse into their quantal packages, they seem to be installing arch'ed java parts using[[BR]] /usr/lib/jvm/<java-variant>-<arch>/[[BR]] as '''installation prefix'''[[BR]]
e.g.[[BR]] /usr/lib/jvm/java-7-openjdk-amd64[[BR]] /usr/lib/jvm/java-7-openjdk-common[[BR]] /usr/lib/jvm/java-6-openjdk-i386[[BR]]
I.e. they seem to have implemented a clean multiarch-separation.
Great so you looked at the way Ubuntu packages JVM. Too bad you didn't find the time to look at our JVM packaging.
I am this: <-> close to making this response ad-hominem. Instead I'll be constructive: {{{ $ rpm -ql java-1.7.0-openjdk | grep '/usr/lib/jvm/' | head -n1 /usr/lib/jvm/java-1.7.0-openjdk-1.7.0.9.x86_64 }}}
Making JVM multilib is not the same as making Java stack support multilib. Suggested reading: New Java guidelines which actually describe why it's the way it is (and yes, I believe you should have actually read them before voting on it)
Updated and about to announce.
Metadata Update from @corsepiu: - Issue assigned to spot
Login to comment on this ticket.