Packaging:Minutes20070911

Fedora Packaging Committee Meeting of {2007-09-11}

Present

• DavidLutterkort (lutter)
• JasonTibbitts (tibbs)
• JesseKeating (f13)
• RalfCorsepius (racor)
• TomCallaway (spot)
• ToshioKuratomi (abadger1999)
• VilleSkyttä (scop)

Writeups

No new guidelines this week.

Votes

No votes this week.

Other Discussions

The Python eggs proposal has undergone further revision: https://fedoraproject.org/wiki/PackagingDrafts/PythonEggs It is mostly finished and a vote next week is likely. If you have something to add, please submit your comments soon.

IRC Logs

[11:59]  * spot is here
[11:59]  * f13 is here.
[11:59]  <f13> head deap in gpg crud.
[12:01]  * abadger1999 here
[12:01]  * scop is here
[12:02]  * spot counts 4 heads
[12:03]  <spot> i know rdieter isn't going to be here, thimm hasn't showed up in a long time
[12:03]  <racor> racor is here
[12:03]  <spot> ok, thats 5
[12:04]  <abadger1999> lutter, tibbs: You guys here?
[12:04]  <spot> the only thing on my agenda is the http://fedoraproject.org/wiki/PackagingDrafts/PythonEggs
[12:05]  <abadger1999> Are the new initscripts ready as well?
[12:05]  <lutter> abadger1999: here now
[12:05]  * f13 reviews eggs
[12:05]  <spot> abadger1999: i don't know, the existing draft still has a lot of open questions
[12:05]  <scop> initscripts stuff still has the "open questions" chapter
[12:06]  <abadger1999> Okay.  So eggs it is.
[12:06]  <spot> abadger1999: is jeremy willing to make the python changes needed to support the egg draft?
[12:06]  <abadger1999> I've talked to jeremy and we won't be restoring eggs-in-distutils for F8.  But he can't think of a reason off hand that it wouldn't go into F9.
[12:07]  <abadger1999> He will want to examine it more after F8 is out, though.
[12:08]  <jeremy> I'll probably try to examine it before f8 is out even so that we can go into f9 with a "we're flipping the switch the day after"  plan
[12:08]  <abadger1999> Excellent
[12:09]  <spot> so, just because i'm dumb
[12:09]  <spot> eggs aren't "rebuilt" from "source", right?
[12:10]  <abadger1999> The eggs in the draft are.
[12:10]  <abadger1999> There's two uses of the word egg.
[12:10]  <lutter> what does rebuilding involve ?
[12:11]  <abadger1999> egg can mean file format like "jar" for java
[12:11]  <abadger1999> or it can mean collection of python module and metada
[12:11]  <spot> OK, so it would be very nice to see a section that says "What is an Egg?" which describes this
[12:12]  <spot> i'm not opposed to eggs inside rpms, but i am opposed to packaging prebuilt eggs into Fedora rpms
[12:12]  <abadger1999> This is because eggs can be installed several different ways.  As a zip file with the filname *.egg, as a directory in site-packages, or as a module in site-packages with either a file or directory of metadata
[12:12]  <lutter> abadger1999: can there be binaries in an egg (e.g. pytjhon bindings) ?
[12:12]  <abadger1999> spot: Sounds good.
[12:12]  <abadger1999> lutter: Yes there can.
[12:13]  <lutter> is that handled cleanly by the egg system ?
[12:13]  <abadger1999> When that happens, the filename includes the arch it was built for.
[12:13]  <spot> abadger1999: i think we want to make it crystal clear that eggs from upstream with prebuilt binary content is NOT ok to just slide into a package
[12:13]  <lutter> what I meant was: does the egg system build this stuff in a way that's multilib compatible ?
[12:13]  <abadger1999> That sounds good.  I'll add a section about using upstream eggs as source only, not as something that can be dropped into the filesystem.
[12:15]  <abadger1999> lutter: It should.  setuptools uses the definitions of the site-packages directories from distutils.  Since we're defining it on x86_64 as lib for noarch and lib64 for arch specific it should pick those up automatically.
[12:16]  <lutter> also, your writeup is really nice to give some background and details on eggs, but it would be good to have a section somewhere that has a short summary of the musts/must nots/shoulds that reviewers have to look out for
[12:17]  <lutter> also: the fact that a pkg provides egg metadata should be indicated with something like a Provides: python-egg(foo)
[12:18]  * spot nods, seems logical
[12:18]  <abadger1999> lutter: ack on the musts/shoulds.
[12:18]  <abadger1999> Regarding the egg metadata... Can we roll this into a comprehensive virtual provides for python similar to other languages?
[12:19]  <spot> abadger1999: maybe... example?
[12:19]  <abadger1999> This assumes that we'll be reverting to upstream WRT eggs & distutils for F9.
[12:20]  <abadger1999> Provides: python(SQLAlchemy) = %{version}
[12:20]  <lutter> that assumes that everything comes as an egg, too, doesn't it ?
[12:22]  <abadger1999> lutter: yes, which is why I specified the assumption about reverting to eggs for distutils in F9.
[12:22]  <spot> well, if everything is eggs, then that seems ok.
[12:23]  <abadger1999> I guess we'd want to defer that until after we know for sure that we'll be doing that for F9, though.
[12:23]  <lutter> if that can be done, sounds fine by me
[12:24]  <spot> abadger1999: we can look at this next week, with some of the suggested changes?
[12:24]  <spot> i don't think there's any rush on this one, given that it won't be applicable until f9
[12:25]  <abadger1999> spot: Err... Actually there are some things that are broken for F8 unless we get at least part of this in.
[12:25]  <abadger1999> The TurboGears stack is not functional on Rawhide at the moment.
[12:26]  <spot> hrm.
[12:26]  <abadger1999> Well, partially functional depending on what options you want to use.
[12:26]  <abadger1999> This is the "Multiple versions" portion of theguidelines.
[12:27]  <spot> well, i'm ok with the draft with the extra section describing eggs and explicitly forbidding prebuilt binary content
[12:27]  <tibbs> Sorry, folks, I was called away.
[12:27]  * spot would like to read that draft before voting on it though
[12:27]  * abadger1999 looks at the F8 schedule
[12:29]  <abadger1999> K.  I'll write the draft with a definition of egg formats and the explicit prebuilt prohibition.
[12:29]  <tibbs> Forgive my confusion, but what does "revert to eggs" mean?  What's being reverted?
[12:29]  <abadger1999> ... and get packages into the review and approval queue pending approval of the guidelines.
[12:29]  <abadger1999> There should be enough time.
[12:30]  <abadger1999> tibbs: Python 2.5 creates eggs (as in egg-info) for any module built with distutils (pretty much everything.)
[12:30]  <abadger1999> But our python2.5 has a patch to disable that.
[12:30]  <abadger1999> So it's removing the patch and reverting to upstream.
[12:31]  <tibbs> OK, thanks.  To satisfy my curiosity, why are we patching it now?
[12:32]  <abadger1999> tibbs: jeremy doesn't know.  Possibly it's because eggs didn't seem to have any benefits on an rpm managed system at the time.  Now that we know that they can support plugins at runtime they have more benefit.
[12:33]  <spot> ok, with that settled until next week, are there any other items for discussion today? (if you do not have something, please say "no")
[12:33]  <abadger1999> no
[12:33]  <spot> no
[12:34]  <tibbs> no
[12:34]  * spot eyes lutter, racor, f13
[12:34]  <lutter> nothing from me
[12:34]  <racor> no
[12:35]  <racor> as usual, I've got to go ... dinner is ready, bye.
[12:35]  <spot> ok, i'll assume that f13 has nothing and close out the meeting
[12:35]  <spot> thanks all
[12:36]  <abadger1999> Thanks spot