From Fedora Project Wiki

Fedora Packaging Committee Meeting of {2007-06-12}

Present

  • AxelThimm (thimm)
  • DavidLutterkort (lutter)
  • JasonTibbitts (tibbs)
  • RalfCorsepius (racor)
  • RexDieter (rdieter)
  • TomCallaway (spot)
  • ToshioKuratomi (abadger1999)
  • VilleSkyttä (scop)

Writeups

No writeups this week.

Votes

The following proposals were considered:

IRC Logs

[12:01]  * spot shakes his broom at the kde folks
[12:02]  <spot> get offa my lawn!
[12:02]  <rdieter> oh, forgot, ok, I guess we need to finish up.  Thanks Than, Kevin, Sebastian (+ everyone I forgot). :)
[12:02]  <spot> yeah, we need to do "real work". ;)
[12:03]  <lutter> howdy
[12:03]  * tibbs here
[12:03]  <lutter> spot: did the terms for the fpc change ?
[12:03]  <rdieter> here (still)
[12:03]  * lutter is startled by real work
[12:03]  <spot> terms?
[12:03]  --> scop has joined this channel (n=scop@cs181043142.pp.htv.fi).
[12:03]  <spot> oh, no. i'm just poking fun at the kde sig. ;)
[12:03]  <Kevin_Kofler> rdieter: Before I leave: is there still the KDE meeting at 2000 UTC? If not, let's remove it from the wiki. :-)
[12:04]  <spot> we dont actually do any real work in this meeting.
[12:04]  <rdieter> Kevin_Kofler: I'll be here for this week anyway, otherwise yeah, remove it (for now).
[12:05]  <spot> abadger1999: alive?
[12:05]  <spot> ok. well, lets get started
[12:06]  <spot> rdieter: your cmake writeup is the only pending item
[12:06]  <thimm> Is there anything that needs full voting body?
[12:06]  <rwmjones> is there an agenda?
[12:06]  * spot cracks the whip
[12:06]  <thimm> I will need to leave early
[12:06]  <rdieter> spot: writup is done.
[12:06]  <spot> yes. there is. :)
[12:06]  <spot> rdieter: please take it off GuidelinesTodo
[12:06]  * rdieter thought he did...
[12:06]  <spot> i might have a cached copy
[12:07]  <rdieter> oops, didn't hit save.
[12:07]  <spot> ok, so the first issue is: http://fedoraproject.org/wiki/PackagingDrafts/ScriptletSnippetsFixes
[12:07]  <spot> The only improvement I can think of for that would be an example.
[12:08]  <spot> the language is correct, but very technically thick.
[12:08]  <tibbs> That and the first sentence is a little confusing.
[12:08]  <spot> I'm a big fan of clarifying examples.
[12:08]  <scop> the main scriptletsnippets page is full of examples...
[12:09]  <spot> scop: i know, but i think an example of where to use : and exit 0 would be useful
[12:09]  <thimm> Maybe shorten to what the guideline is and then explain in further paragraphs?
[12:09]  <spot> since most people are conditioned/used to using || :
[12:09]  <thimm> Like "avoid non-zero exit codes of scriplets as these can mess up a whole yum transaction"?
[12:09]  <lutter> it's hard to understand what exactly you are supposed to do
[12:10]  <spot> It could be as simple as "Processes executed in scriptlets should never exit with a non-zero exit code. Here are the reasons why. Here are some ways that you can prevent it."
[12:10]  <thimm> lutter: "Make your scripts failproof, don't reply on rpm to cater for you"
[12:11]  <spot> (obviously, I'm oversimplifying, but the idea is the same)
[12:12]  <spot> scop: thoughts?
[12:13]  * abadger1999 growls at at pidgin
[12:13]  <scop> no particular thoughts, as long as the info is correct, I don't really care
[12:13]  <spot> so... do we want to approve this as is, or do we want to rework it somewhat?
[12:14]  <scop> if someone wants to rework, feel free
[12:14]  <tibbs> I don't have any real complaints about what's in the draft.
[12:14]  <spot> neither do i. i think its ok to vote on this draft, we can always improve it if people get confused.
[12:14]  <rdieter> approve as-is (for now): +1
[12:14]  <spot> +1
[12:14]  <tibbs> But some of the examples earlier in the page will need to be altered as well, won't they?
[12:15]  <scop> +1
[12:15]  <scop> tibbs, I'll have a look if this passes
[12:15]  <spot> tibbs: yeah, the examples need to be altered to reflect this.
[12:15]  <tibbs> +1
[12:15]  <rdieter> thimm?
[12:16]  <scop> lutter?
[12:16]  <spot> abadger1999?
[12:16]  <rdieter> bueller?
[12:16]  <lutter> +1
[12:16]  <abadger1999> +1
[12:16]  <lutter> (sorry, on the phone)
[12:16]  <racor> +1
[12:16]  <thimm> +1
[12:16]  <spot> ok, thats a pass.
[12:16]  <spot> next item: http://fedoraproject.org/wiki/PackagingDrafts/OCaml
[12:16]  <rwmjones> yes ...
[12:17]  <rwmjones> I summarised everything that happened since last tuesday here:
[12:17]  <rwmjones> https://www.redhat.com/archives/fedora-packaging/2007-June/msg00070.html
[12:17]  <rwmjones> & I hope that covers all the things people raised last week
[12:17]  <thimm> Looks like it passed
[12:18]  <tibbs> I think these guidelines are to the point where we can start to make use of them.
[12:18]  <spot> The draft looks good to me.
[12:18]  <thimm> Why do we need signature hashes for ocaml and not all the rest of the languages?
[12:18]  <rwmjones> because of the very strict dependencies that compiled code has
[12:18]  <tibbs> "rest of the languages" being Perl, Python, etc?
[12:18]  <rwmjones> it's very different from dynamic languages
[12:18]  <rwmjones> a good analogy are the kernel module symbols
[12:19]  <rwmjones> which have hashes because of super-strict type (and other) dependencies
[12:19]  <tibbs> It's not really up to us to judge the ABI constraints of a language anyway.
[12:19]  <spot> agreed.
[12:19]  <rdieter> we don't have to like it though... :)
[12:20]  <tibbs> That's what's necessary; our goal is to figure out how to make reasonable guidelines.
[12:20]  <spot> I think these guidelines are extremely well polished and written.
[12:20]  <rdieter> vote on draft?
[12:20]  <spot> +1
[12:20]  <tibbs> +1
[12:20]  <rdieter> +1
[12:20]  <rwmjones> https://www.redhat.com/archives/fedora-packaging/2007-May/msg00105.html <-- explanation of the hashes
[12:20]  <tibbs> I guess the RPM folks were amenable to getting the dependency generators into rpm-build?
[12:20]  <scop> +1
[12:21]  <rdieter> tibbs: yes
[12:21]  <lutter> +1
[12:21]  <rwmjones> https://www.redhat.com/archives/fedora-devel-list/2007-June/msg01071.html <-- into RPM
[12:21]  <spot> racor, abadger1999, thimm?
[12:21]  <abadger1999> +1
[12:21]  <tibbs> Should we wait for that or just go ahead with the _use_internal_depencency_generator bit?
[12:22]  <rwmjones> I'm happy to remove those from the specfiles when/if they go in to rpmbuild
[12:22]  <racor> +1, sorry I am distracted and probably will have to leave suddenly
[12:22]  <spot> racor: thats understandable, thank you for your time.
[12:22]  <tibbs> Because I'm concerned that turning off the dependency generation will cause other things to break.
[12:22]  <rdieter> rwmjones:+1
[12:22]  <spot> rwmjones: now is the time to be pushing on the rpm maintainers for these changes in rawhide, very early in the f8 cycle
[12:22]  <rwmjones> the ocaml-find-provides/requires scripts call /usr/lib/rpm/find-{provides,requires}
[12:22]  <rwmjones> I don't know if that is sufficient
[12:23]  <tibbs> That should be OK.
[12:23]  <spot> the draft passes.
[12:23]  <abadger1999> rwmjones: Just a note -- they probably won't hit older rpm versions so we'll need to keep it around for a while and note at which Fedora Release it is no longer required.
[12:23]  <tibbs> Obviously that will have to go before it gets into rpm-build, though.
[12:23]  <rwmjones> thanks ... quick question, do I need to rename that page?
[12:23]  <spot> rwmjones: no, abadger1999 will take care of it
[12:23]  <rwmjones> ok, thanks
[12:23]  <abadger1999> rwmjones: I'll do it
[12:23]  <spot> Next item: EmacsenAddOns
[12:24]  <spot> this one looks like its waiting on some bugzillas to be resolved
[12:24]  <spot> does anyone here have experience/care about emacs? :)
[12:25]  <scop> I do, both :)
[12:25]  <spot> ok, i'll wait for you to tell me that one is ready.
[12:25]  <scop> ok
[12:25]  <spot> is there any thing else anyone would like to discuss?
[12:26]  <tibbs> I use emacs but I know zip about packaging.
[12:26]  <tibbs> abadger1999: Did you want to propose something about static libraries?
[12:26]  <scop> I added the pkgconfig file to the latest xemacs package builds, dunno what's the status of emacs
[12:26]  <abadger1999> I haven't written anything up yet but there are several issues.
[12:27]  <spot> ok, i have one minor item that I came up with
[12:27]  <spot> http://fedoraproject.org/wiki/PackagingDrafts/DirectoryOwnershipImprovement
[12:27]  <scop> +1
[12:27]  <rdieter> +1
[12:28]  <spot> +1 (i like my own drafts, usually)
[12:28]  <tibbs> +1 my perl example is overly complicated anyway.
[12:28]  <scop> actually s/acceptable/mandatory/
[12:28]  <scop> or something to that effect
[12:28]  <scop> (last sentence of the draft addition)
[12:28]  <spot> ah, thats a good point. I'll reword that before committing
[12:29]  <abadger1999> +1
[12:29]  <spot> In this case, each package must own the /usr/share/Foo/Animal/ directory.
[12:29]  <lutter> +1
[12:30]  <spot> ok, it passes
[12:30]  <abadger1999> BTW, caillon asked if directory ownership problems could be addressed in rpm (No bug filed that I know of)
[12:30]  <thimm> +1 (just for counting votes ;)
[12:30]  <tibbs> I love counting votes.
[12:30]  <spot> abadger1999: so, the answer is almost certainly yes, ajax has also pointed this out
[12:31]  <spot> if the guidelines become obsoleted by rpm, then we have less guidelines, but our job is not to point to rpm and demand fixes. :)
[12:31]  <thimm> How would rpm manage directory ownership?
[12:31]  <thimm> This is a cross-package problem
[12:31]  <rdieter> abadger1999,thamm: fwiw, that *other* rpm(5) already does enforce that (ie, no unowned dirs).
[12:31]  <thimm> And rpm never examines the dependency chains (at runtime) during the build
[12:31]  <thimm> rdieter: How?
[12:32]  <thimm> How will rpm5 know that there is no package owning "Animal"
[12:32]  <rdieter> thimm: it adds deps to parent dirs (somehow).
[12:32]  <thimm> Especially if it is not part of the BRs
[12:32]  <rdieter> thimm: only checked at install-time (afaict)
[12:32]  <thimm> So rpm5 does owenrship bloating
[12:32]  <thimm> That's not better than now
[12:32]  <thimm> Point is: Only gloabl tools can ,manage cross-package issues like directory ownerships
[12:33]  <abadger1999> caillon thought checking at install time and recording when an unowned directory was created would work.
[12:33]  <scop> adding deps to parent dirs is not ownership bloating
[12:33]  <rdieter> thimm: only at install time, pkgs don't bloat
[12:33]  <thimm> OK, as a warning then?
[12:33]  <thimm> OK, I'll have to go now
[12:33]  <scop> packges' dependencies do bloat, and yum's filelist behaviour will probably be triggered a lot
[12:33]  <thimm> Hope there is nothing where votes are 50-50
[12:33]  <spot> abadger1999: encourage caillon to patch rpm. We're always willing to let others make our life easier. :)
[12:33]  <spot> are there any other drafts or items of business?
[12:33]  <rdieter> spot: +1
[12:34]  <scop> there were some disagreements about usersandgroups, is anyone working on an alternate draft?
[12:34]  <spot> i'm going to try to work something out
[12:35]  <tibbs> About "no modifications to upstream tarball except...":
[12:35]  <tibbs> That discussion ran on into a different area.
[12:35]  <rdieter> fyi, epel list includes notice that they're going to vote on repotags...
[12:35]  <rdieter> "
[12:35]  <rdieter> We hereby like to ask the Packaging Committee to bless the use of
[12:35]  <rdieter> repotags for EPEL.
[12:35]  <thimm> FPC = pope?
[12:36]  <rdieter> not sure what that means. :)
[12:36]  <spot> Did they vote on that?
[12:36]  <rdieter> spot: going to vote.
[12:36]  <tibbs> I just wanted to ask if the general sentiment was that I could say "removal of unacceptable content ONLY".
[12:36]  <rdieter> sorry, just a proposal at this point.
[12:36]  <spot> OK, well, we'll deal with that when and if it happens.
[12:36]  <thimm> tibbs++
[12:36]  <thimm> OK, now I really need to lift, bye all!
[12:36]  <spot> thimm: thanks
[12:36]  <tibbs> Thanks, thimm.
[12:37]  <rdieter> tibbs++
[12:37]  <abadger1999> tibs: +1
[12:37]  <spot> i agree with tibbs as well
[12:37]  <abadger1999> tibbs even.
[12:37]  <scop> -1
[12:37]  <tibbs> OK, I'll go ahead and draft that for next week.
[12:37]  <tibbs> I spent all of my time on reviews last week...
[12:38]  <spot> alright. i think thats it for this week. i'm going to go feed my hungry cat.
[12:38]  <spot> thanks all.