From Fedora Project Wiki
< Extras | SteeringCommittee
2006 December 28 FESCo Meeting
Note: due to the holiday season is was expected that probably only small number of people could participate. FESCo nevertheless choose to run a small and meeting without actually voting on anything.
Members
Present
- thl
- bpepple
- tibbs
- dgilmore
Absent
- jwb
- jeremy
- scop
- warren
- ch4chris
- abadger1999
- spot
- awjb
- rdieter
Summary
Minimal approve messages
Background: Some packages ( krename: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=220210 and pyfribidi: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=219071 ) were not branched by our human cvs guards because it was unclear if a full review had beed done for them -- at branch request time there was only a small note "APPROVED" in a comment of thereview bugs without details what things had been checked
- dgilmore: the reason i refused to branch those two packages was that the review just said i approve this; [...] i had no way to know if any checking or anything was done
- tibbs: I agree that minimal reviews are bad (although they used to be more common) but I'm not sure the two in question were minimal.
- thl: I'm wondering if we should add a rule like "the review has to at least mention 8 points he checked when approving a package"
The decision went towards a proposed new rule: "the reviewer has to at least mention that he checked the license, if the sources match upstream and 5 other points he checked when approving a package". Dgilmore will post to f-e-l about this whole thing in more detail and start a public discussion before FESCo discusses this further.
Site note: the packages will probably get branched now, but the general problem remains.
How to get something realized in Extras
thl wrote http://www.fedoraproject.org/wiki/Extras/Schedule/HowToGetSomethingRealized , tibbs improved it slightly -- FESCO members that were around liked it, Will get posted to f-e-l for further comments.
free discussion around Extras
- there are still so many broken deps -> seems the ones that are left need real work.
- XulChris: "is a dist tag of fc#.foo okay for 3rd party repos?" Some disussions around this, the general consensus was "yes". XulChris will probably make a wiki page for guidelines (no rules -- people don't have to follow it, but they can if they want) on setting up a 3rd party repo and what dist-tag to use.
Log
00:01 < thl> | FESCo meeting ping -- abadger1999, awjb, bpepple, c4chris, dgilmore, jeremy, jwb, rdieter, spot, scop, thl, tibbs, warren
00:01 --- | thl has changed the topic to: FESCo meeting -- Meeting rules at [WWW] http://www.fedoraproject.org/wiki/Extras/Schedule MeetingGuidelines -- Init process
00:01 < thl> | Hi everybody; who's around?
00:01 < dgilmore> | pong
00:01 <@ tibbs> | I'm here.
00:01 * | bpepple is here.
00:01 * | thl seems to be one minute late
00:01 --> | |DrJef| (spacious... he's so) [n=jefrey@fedora/Jef] has joined #fedora-extras
00:01 * | nirik is in the rabble stands.
00:02 < thl> | hmmmm, that makes four FESCo members...
00:02 < dgilmore> | thl: there is probably a few on holiday
00:02 < thl> | do we want to run a normal meeting with only four? I don't think that makes to much sense
00:02 < thl> | dgilmore, yeah, probably
00:02 * | thl was not at work today, too
00:02 < bpepple> | thl: probably not.
00:03 < dgilmore> | thl: that doesnt make much sense we cant decide on anything
00:03 < mmcgrath> | pong
00:03 < thl> | dgilmore, shall we discuss the branching issue roughly?
00:03 * | thl tries to find it on the schedule
00:04 < dgilmore> | thl: sure
00:04 <@ tibbs> | Yes, please.
00:04 * | lmacken will be around
00:04 < bpepple> | dgilmore: Is that about the minimal reviews?
00:04 < dgilmore> | the reason i refused to branch those two packages was that the review just said i approve this
00:04 < dgilmore> | bpepple: yeah
00:04 * | bpepple agrees with dgilmore on this.
00:04 < thl> | we talked about that ages ago
00:05 < dgilmore> | i had no way to know if any checking or oanything was done
00:05 <@ tibbs> | I agree that minimal reviews are bad (although they used to be more common) but I'm not sure the two in question were minimal.
00:05 < thl> | I'm wondering if we should add a rule like "the review has to at least mention 8 points he checked when approving a pacakge"
00:05 <@ tibbs> | Does anyone have links handy?
00:05 < nirik> | he since added ot those reviews with a bunch of checklist items.
00:05 < bpepple> | tibbs: I disagree. He didn't mention anything that he checked in it.
00:05 * | dgilmore would like that at least there is written in the approval post, that 1) i built this in mock, 2) the spec meets the guidelines and 3) the md5sum matches upstream
00:05 <@ tibbs> | Ah:
00:06 <@ tibbs> | krename: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=220210
00:06 <@ tibbs> | pyfribidi: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=219071
00:07 --- | thl has changed the topic to: FESCo meeting -- minimal approve messages
00:07 <@ tibbs> | Hmm, no krename isn't one of them.
00:07 <@ tibbs> | No, neither of those is correct. Crap, which are the problem reviews:?
00:08 < thl> | http://fedoraproject.org/wiki/Extras/CVSSyncNeeded?action=recall&rev=2462
00:08 < thl> | tibbs, those were it afaics
00:08 < bpepple> | tibbs: He only put in the details after dgilmore informed him he wouldn't branch. His initial review consisted of this: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=219071#c2
00:08 < thl> | welöl, those reviews are not the only ones where it happend
00:08 <@ tibbs> | There has never been a requirement that reviewers actually post the checklist.
00:09 < thl> | so I still suggest something like "the review has to at least mention that it was build in mock, the sources match upstream and 5 other points he checked when approving a pacakge"
00:10 < dgilmore> | thl: +1
00:10 < bpepple> | thl: That sounds fine with me.
00:10 < thl> | tibbs, no, not yet, but I feel uncompfortable if people just write "approved" -- especially if they are new in the game (which is not the case here)
00:10 <@ tibbs> | It was also never a requirement that reviewed packages were built in mock.
00:10 < bpepple> | thl: Maybe they should also have to verify the package license in the review also.
00:10 < dgilmore> | long time reviews need to set an example for new reviewers
00:10 < bpepple> | dgilmore: +1
00:11 <@ tibbs> | mock takes either significant infrastructure or significant time (or both) to run.
00:11 --> | schlobinux1 (Xavier Bachelot) [n=xavierb@...] has joined #fedora-extras
00:11 <@ tibbs> | We want to encourage reviews, not put up a pile of beauracracy making it more difficult to find reviewers.
00:11 < bpepple> | tibbs: How else can a new reviewer easily check that all the BR are there?
00:11 < dgilmore> | tibbs: yes but one of the review guidelines is that it must build in mock on at least one arch
00:11 < thl> | next round: "the reviewer has to at least mention that he checked the license, the sources match upstream and 5 other points he checked when approving a pacakge"
00:12 < bpepple> | thl: +1
00:12 < dgilmore> | thl: that is fine also
00:12 < dgilmore> | +1
00:12 <@ tibbs> | dgilmore: The only mention of mock in the review guidelines is a "should".
00:12 < bpepple> | tibbs: maybe the guideline need to be modified.
00:12 <@ tibbs> | The requirement is that it compile and build into binary rpms on at least one supported architecture.
00:12 <@ tibbs> | Well, cart before the horse, then.
00:12 <-- | schlobinux_ has quit (Read error: 110 (Connection timed out))
00:13 < thl> | tibbs, +1, I don't consider the mock stuff to important
00:13 < thl> | it was more important in the fedora.us days when there was a semi-automatical buildsys
00:13 <@ tibbs> | Plain reviews were the norm before I started doing the long checklists.
00:13 * | mmcgrath notes it'll build in mock before it gets to our end users anyway
00:13 <@ tibbs> | And there were complaints at the time that I was posting too much.
00:13 < thl> | mmcgrath, +1
00:14 < bpepple> | Does Toshio python qa tool still work?
00:14 < dgilmore> | tibbs: that probably should be a must
00:14 < thl> | tibbs, well, in my memory most reviews had at least some things listed that the reviewer checked; only a small number had just the "approved" message
00:15 < thl> | but tha's history now in any case
00:15 * | nirik always used a checklist based on the 'how to review a package' page.
00:15 < |DrJef|> | bah... all this back and forth in my review over a trivial helper script that I want to place in a %docs section
00:15 < thl> | tibbs, so what do you suggest? Leave everything as it is?
00:15 < XulChris> | why dont we write a shell script that goes through the review checklist and the reviewer can mark yes no or n/a on each item then the shell script automatically generates the review text
00:16 < thl> | XulChris, no, I'd leave some freedom for the reviewers how they do there work
00:16 < bpepple> | XulChris: That's what Toshio's tool did, though it was a gtk+ gui.
00:16 < thl> | we can of course write one, but we should not force it on people
00:16 < |DrJef|> | XulChris, i like that... spot had a checklist cheatsheet in his wikipages
00:16 < thl> | XulChris, and what bpepple said ;)
00:16 < thl> | |DrJef|, there are multiple ones in the wiki iirc
00:16 <@ tibbs> | Here's a nice one:
00:16 <@ tibbs> | https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=167254
00:16 < dgilmore> | All i really want is some evience that a review was done
00:17 < |DrJef|> | thl, spot's is oldest :->
00:17 < thl> | |DrJef|, a lot of people maintain there own variant
00:17 < bpepple> | Ah, there's the name: qa-assistant.
00:17 <@ tibbs> | I don't object to asking people to post a list of what they checked.
00:17 <@ tibbs> | I just think this whole issue is rather unpleasant.
00:18 < thl> | tibbs, so in other words you'd like to stick to the current behaviour?
00:18 <@ tibbs> | No slight against dgilmore intended, but I don't recall who appointed him the gatekeeper of what a good review is.
00:19 < |DrJef|> | tibbs, i think process transparency requires a checklisting... including the should items
00:19 < dgilmore> | tibbs: any of the branches can refuse to branch if they feel the review in not appropriate
00:19 < thl> | tibbs, those that branch cvs are requires to look at the review
00:19 < |DrJef|> | tibbs, an explicit checklist continues to re-affirm best practices to new contributors
00:19 < XulChris> | dgilmore: thing is everyone will be able to do their own branches soon
00:20 < |DrJef|> | tibbs, and if saged contributors aren't doing it....new people dont respect it
00:20 <@ tibbs> | One version of the checklist I use is even online.
00:20 < dgilmore> | XulChris: yes so we need to have an automated way to ensure things are ok then
00:20 < |DrJef|> | tibbs, whatsmore.. as policy shifts... the checklisting helps reviewers tell other reviewers about policy changes they nissed
00:20 <-- | k0k has quit (Read error: 104 (Connection reset by peer))
00:20 < ecik> | my review template is at wiki as well
00:20 <@ tibbs> | But if someone wants " a checklisting... including the should items" then I don't think that even my reviews qualify.
00:20 < dgilmore> | XulChris: we can write a script that parses the review for certain things
00:21 < |DrJef|> | tibbs, should include... not must
00:21 < thl> | tibbs, that why I proposed "the reviewer has to at least mention that he checked the license, the sources match upstream and 5 other points he checked when approving a pacakge"
00:21 < |DrJef|> | tibbs, knowing what shoulds have been done.. help other people manage their time..so they dont redo shoulds
00:21 < thl> | it stays up to the review what 5 things he checks
00:22 < thl> | that afaics still should be quite easy to satisfy
00:22 < bpepple> | thl: agreed.
00:22 < dgilmore> | thl: that works and we can sript the checking of that for branching
00:22 < dgilmore> | s/sript/script/
00:22 * | nirik wonders how checking that can be scripted... since bugzilla text is free form.
00:23 < thl> | well, let's stop here
00:23 < thl> | and get back to it next week
00:23 < thl> | does anybody want to bring this topic to the list for discussions?
00:24 < dgilmore> | thl: i will as i started it
00:24 < thl> | not all at once please ;-)
00:24 < thl> | dgilmore, thx
00:24 --> | rafalzaq (RafaÅ) [n=rafalzaq@...] has joined #fedora-extras
00:24 --- | thl has changed the topic to: FESCo meeting -- How to get something realized in Extras
00:24 < thl> | tibbs, thx for fixing/enhancing the page
00:24 < dgilmore> | thl: i liked what you posted
00:24 < thl> | bpepple, what do you think of it?
00:24 < thl> | and tibbs ?
00:25 < bpepple> | thl: I liked it.
00:25 <@ tibbs> | Sorry, phone.
00:25 < thl> | dgilmore, bpepple, should there be more/anything else on the page?
00:25 < thl> | tibbs, np
00:25 < thl> | btw, we are talking about http://www.fedoraproject.org/wiki/Extras/Schedule/HowToGetSomethingRealized
00:26 < thl> | I'm wondering if I should just post about it to the list
00:26 < thl> | so people can comment
00:26 < dgilmore> | thl: not that i can think of right now. But it is good to make everyone aware they dont need to be in FESCo or some other committie to get things done
00:26 <@ tibbs> | It makes sense to me. I get tired of people saying "fesco should do something" and then just leaving it for us to figure out what (if anything) to do.
00:26 < thl> | dgilmore, that eactly why I wrote it
00:26 < bpepple> | off the top of my head I can't think of anything. I'm glad you mentioned that part about what FESCo isn't.
00:26 < thl> | so shall I take it to the list?
00:27 < dgilmore> | thl: do
00:27 < bpepple> | thl: yes.
00:27 <@ tibbs> | Yes, I think so.
00:27 --> | Sonar_Guy (Scott Glaser) [n=sglaser@fedora/sonarguy] has joined #fedora-extras
00:27 < thl> | k, will do
00:27 --- | thl has changed the topic to: FESCo meeting -- free discussion around Extras
00:27 < thl> | so, anything else?
00:27 < thl> | there are still so many broken deps
00:27 < XulChris> | i have a question about last weeks meeting
00:28 <@ tibbs> | I have tried to rebuild essentially every busted dep at this point.
00:28 < XulChris> | is a dist tag of fc#.foo okay for 3rd party repos?
00:28 <@ tibbs> | The ones that are left need real work.
00:28 < thl> | tibbs, okay, many thx
00:28 < thl> | XulChris, good question...
00:28 < thl> | XulChris, you maybe should ask spot and/or the packaging committee
00:29 <@ tibbs> | Should we try to say what 3rd party repos should do?
00:29 < thl> | XulChris, I tend to think "yes"
00:29 < bpepple> | XulChris: I believe the tag should match the distro it is for. Spot would probably be the one to ask, since he felt strongly about it.
00:29 < thl> | tibbs, no, of course not, but ge can guide
00:29 <@ tibbs> | I guess if they're asking our opinion it's OK.
00:29 < XulChris> | i think we should make a wiki page for guidelines on setting up a 3rd party repo
00:29 < bpepple> | XulChris: Probably not a bad idea.
00:30 <@ tibbs> | If by "we" you mean "I" then I'm all for it!
00:30 < XulChris> | if by "I" you mean "you"... ;-)
00:31 < XulChris> | im not sure i know enough about setting up a 3rd party repo to actually write a guidelines page
00:31 <@ tibbs> | But seriously, I don't suppose it could hurt. I'm just wary of the potential for flamewars.
00:31 < |DrJef|> | XulChris, you have to be very very clear that such a page is meant as a best effort to help 3rd parties synergize with fedora packaging policy.. and its not meant as a coersive set of rules
00:31 <-- | Sonar_Guy has quit (Remote closed the connection)
00:31 <@ tibbs> | Damn, I had a nice review of libsieve written up and firefox crashed.
00:31 < thl> | tibbs, you did not get asbestos underwear for christmas?
00:31 < XulChris> | |DrJef|: yes, ultimately its just a set a guidelines
00:32 < |DrJef|> | XulChris, guidelines which will be in flux
00:32 < XulChris> | I can start a draft but ill probably need some help
00:32 < |DrJef|> | XulChris, if the language on the page isn't overly squishy, people will misread it as a written-in-stone specification for compatibility
00:33 < XulChris> | well its a wiki so people can fix up any bad language
00:33 < dgilmore> | i think its perfectly fine for 3rd party repos to use .fcx if they choose to
00:33 < |DrJef|> | XulChris, the best way to approach it is to think about people who run in house distros
00:34 < |DrJef|> | XulChris, instead of thinking.. generally public repos
00:34 < |DrJef|> | XulChris, err inhouse repos
00:34 <@ tibbs> | Of course, such a page would have to say "don't replace or upgrade existing Fedora packages", which gets you into a heap of flames already.
00:34 < XulChris> | dgilmore: i was asking about fcx.foo like fc6.lvn
00:34 < dgilmore> | |DrJef|: i have inhouse repos and use .el4
00:34 < dgilmore> | XulChris: they are free to use what they want
00:34 < XulChris> | tibbs: no we would just recommend to not do so, list reasons why, and also list exceptions
00:35 < XulChris> | dgilmore: yes they are free, but they *should* be using fcx
00:35 < |DrJef|> | dgilmore, im just suggesting that as XulChris writes this... he thinks of the audience as people with inhouse repos.. instead of the more obvious repo targets out in the wild....
00:35 < dgilmore> | XulChris: we have no say on what they do ro dont do or what they should or should not be doing
00:35 < XulChris> | dgilmore: we definately have a say on what is good practice
00:36 < dgilmore> | we can say whats good practice for Fedora
00:36 < XulChris> | and we can also say what is good practice for repos that use fedora
00:36 < dgilmore> | wheather they follow that it up t othem
00:36 < XulChris> | ofcourse
00:36 < XulChris> | we arent mandating anything
00:36 < dgilmore> | XulChris: we have zero influence on that
00:36 < thl> | XulChris, please also mention that users should set the vendor and/or packager tags
00:37 < dgilmore> | XulChris: all we can do is set guideelines for us and hope that third party repos follow them
00:37 < XulChris> | dgilmore: thats all im asking we do
00:37 < dgilmore> | XulChris: livna in my opinion should use .fcx as a disttag
00:38 < dgilmore> | but thats just my opinion
00:38 < XulChris> | ya thl just mentioned livna's dist tag will be fixed
00:38 < thl> | well, the disttag is not that easy
00:38 < XulChris> | and dribbles dist tag also because dribble will do what livna does
00:38 < ecik> | but if it uses .lvnx it is easier to check what repo package is from
00:38 < thl> | as fcx is lower then lvn
00:39 < XulChris> | thl: ya unfortunately for livna it means a rebuild, might be easier to only do it for fc7
00:39 < thl> | XulChris, I'll see what livna will do in the long term
00:39 < thl> | XulChris, yeah, probably
00:39 < thl> | anyway, anyting else regarding Extras we should discuss?
00:39 < thl> | otherwise I'll close the meeting for today
00:40 < bpepple> | of course shouldn't our dist tags be f7, since Core is eventually gonna dissappear?
00:40 < thl> | bpepple, hehe
00:40 < thl> | bpepple, that probably a decision for the PC (sorry, once again I try to move work to the PC...)
00:40 < XulChris> | fu7 (fedora universe 7) ;-)
00:40 < thl> | Fedora Nexus
00:41 < bpepple> | Oh no, another naming session. ;)
00:41 * | thl will end the meeting in 60
00:41 * | thl will end the meeting in 30
00:42 * | thl will end the meeting in 13
00:42 <-- | Belegdol has quit ("Leaving")
00:42 < thl> | -- MARK -- Meeting End