FESCo log for 2009-01-23

---Topic for #fedora-meeting is FESCo meeting -- init
---Topic for #fedora-meeting set by jds2001 at Fri Jan 23 11:58:14 2009
jds2001FESCo meeting ping -- bpepple, dgilmore, dwmw2, jwb, notting, nirik, sharkcz, jds2001, j-rod
j-rodpresent
*notting is here
*sharkcz present
*jwb is here
*nirik is here.
*bpepple is here.
jds2001dgilmore: ping
jds2001oh well, let's get started I guess
dgilmorejds2001: sorry i heard this is for cardinals fans only.  ill have to take my exit :)
jds2001lol :)
jds2001ok, let's get started
---jds2001 has changed the topic to: FESCo Meeting - agenda at https://fedorahosted.org/fesco/report/9 -- tickets
jds2001.fesco 26
zodbotjds2001: #26 (Archer) - FESCo - Trac - https://fedorahosted.org/fesco/ticket/26
jds2001anyone here to represent this?
nottingi don't see tromey
dgilmorereading the feature page i dont have any questions
jds2001OK, does anyone have questions that we should postpone this?
jds2001me either
notting+1 from me
sharkczno questions and +1
jds2001+1 here
bpepple+1 looks straight-forward to me.
dgilmore+1
j-rodlooks good to me, +1
jwb+1
jwbi need to step away for just a minute.  be right back
nirikso this is a gdb fork... it replaces current gdb? or exists alongside it?
jds2001I see six +1's, so we've approved the Archer feature
jds2001i think it replaces it.
jds2001at least from what I see on the page, esp the contingency plan
j-roddoesn't seem to be a fork, but rather a merger of upstream development branches
nirikok... +1 from me as well.
jds2001moving right along...
jds2001.fesco 27
zodbotjds2001: #27 (IBus) - FESCo - Trac - https://fedorahosted.org/fesco/ticket/27
markmchttps://fedoraproject.org/wiki/Features/IBus
jds2001anyone around to represent this one?
nottingit's very much the wrong time of day for them
jds2001yeah
bpepple+1, but i'd like to see them answer Rathann's questions.
dgilmorethe test plan is pretty vague
dgilmorebut otherwise +1
j-rodjoy. yet another input method.
nirikI think the answer to the first question is yes, not sure about the second.
notting+1
nirikj-rod: yeah, we are leaving behind a trail of them. ;)
jds2001+1 here too
nirik+1 here...
sharkcz+1
nottingj-rod: scim upstream is dead :/
jwb0
j-rodsimilar to the font reorg... can this be the last time we have yet another input method framework?
jds2001i see six +1's, so we've approved the IBus feature
bpepplej-rod: yeah, that would be nice.
j-rodToo late anyway, but I'm a 0 as well.
jds2001it will be in the IRC log :)
jds2001.fesco 30
zodbotjds2001: #30 (KVM PCI Device Assignment - http://tinyurl.com/cq6w93) - FESCo - Trac - https://fedorahosted.org/fesco/ticket/30
nirikthis looks nice and cool. +1 here.
dgilmore+1 i want to use it
jds2001+1 here too
sharkcz+1
*nirik has a modem card for * that he would love to be able to passthru to a guest.
j-rod+1, yes please
nottingah, was going to ask about graphics cards. :)
bpepple+1
jds2001i saw markmc join :)
markmcnirik, yeah, one of the most common requests :)
nottingmarkmc: is this 'all' PCI, including -E, mini, mini-E, cardbus, etc?
markmcnotting, outta luck for now
markmcnotting, bleh, PCI-X is what works well
markmcnotting, everything is else *shrug* try and see
dgilmorenirik: i want to move asterisk to its own virtual machine.  but need to pass through the tdm400
notting+1
sharkczmarkmc: what are the reasons?
jwbmarkmc, then can you rename the feature to PCI-X?
markmcdgilmore, see "phone line termination cards" :)
nottingback in 2 min... sorry
markmcsharkcz, jwb, VT-d is all tied up in PCI-X details
markmcsharkcz, jwb, but you can use e.g. a PCI device behind a conventional bridge
markmcsharkcz, jwb, just not assign another device behind that bridge to another guest
jwbbut unless you're testing that and know it works...
markmc(and more horrible details ad nauseum)
sharkczmarkmc: ok
jwbi just want to make sure the feature is promoting what is actually being tested
markmcjwb, not in anger yet, but we certainly want to get there
jwbso we don't get "ZOMG THIS DOESN'T WORK WITH MY PCI-E SOUND CARD" or something ridiculous
markmcjwb, if it hasn't had enough testing or is plane broken, we'll be taking it out of the UI and release notes etc.
sharkczyea, put there some pointer for required/recommended HW
jwbmarkmc, i think you're misunderstanding what i'm asking...
markmcjwb, one idea is just to expose it as "KVM PCI Network Interface Assignment" to limit it to a specific set of hardware
markmcjwb, except that would rule out the TDM cards, so ....
jwbmarkmc, all i'm asking is that you do this: KVM_PCI-X_Device_Assignment
jwbfor now
wwoodsIIRC it requires some kernel bits which, last time we turned them on, they kind of blew up a lot of machines with faulty BIOSes
markmcjwb, I understand, but I think not - e.g. the TDM cards aren't PCI-X AFAIK
markmcwwoods, CONFIG_DMAR is enabled again now AFAIR
markmcwwoods, said blowing up needs to be just fixed if it still exists
markmcwwoods, e.g. more black listing
markmcconfig-x86_64-generic:CONFIG_DMAR=y
markmcconfig-x86_64-generic:CONFIG_DMAR_GFX_WA=y
markmcconfig-x86_64-generic:CONFIG_DMAR_FLOPPY_WA=y
dgilmoremarkmc: so will it work with my pci tdm400?  and can i test that today?
j-rodthe PCI-X part... is that "any PCI or PCI-X card behind a PCI-X bridge" ?
j-rods/any/most/
j-rodor some
markmcdgilmore, there's no UI yet or code to unbind the card in the host except, but you could try with qemu-kvm -pcidevice
markmcj-rod, any PCI-X card, I think
markmcj-rod, the details are vague, but the plan is to not present any un-assignable cards in the UI
j-rodmarkmc: I think I asked that wrong...
wwoodsmarkmc: right, that was it - is there an obvious way to determine when "my machine is broken" is due to DMAR
wwoodsso we can make sure bug reports get assigned to the right people to update the blacklsits
markmcwwoods, I think they typically just don't boot, not 100% sure
j-rodmarkmc: what I'm meaning to ask is whether most any device sitting behind a pci-x bridge should be able to be passed through, or can one also pass a tdm card through if its sitting in a plain pci slot?
markmcwwoods, but I haven't noticed any new reports in quite a while
wwoodsmarkmc: so is there a magic "nodmar" flag that would disable that support (and thus prove that dmar is the cause of the problem)?
markmcj-rod, yes, you could assign the tdm card to a guest
j-rod(trying to understand if the 'VT-d is all tied up in PCI-X details" comment means a PCI-X bridge is required)
markmcj-rod, but e.g. if you had another tdm card behind the same bridge, you couldn't assign that to another guest
markmcj-rod, any machine with VT-d will have PCI-X as the root AFAIK
markmcwwoods, yes, I think there is a flag like that
*markmc looks
j-rodmarkmc: ah, ok, that more or less answers what I'm asking then :)
jwbperhaps stating that on the page would help
jwbbecause not everybody will know wtf VT-d means
markmcjwb, intel_iommu=off
sharkczand where to expect it
j-roddoes the same hold true for AMD?
markmcjwb, it's similar to "you need VT support for KVM", that's all
markmcj-rod, don't know the details of AMD IOMMU at all
wwoodsmarkmc: "intel_iommu=off" is the magic "dmar-b-gon" flag?
markmcwwoods, looks like it, yeah
wwoodsmarkmc: any idea when that change went into the rawhide kernel?
wwoodswhere should I tell testers to report DMAR-related failures, and what info do they need to include in the report?
wwoodsI assume bz.r.c, rawhide/kernel
markmcwwoods, before may 2008 AFAICS
markmcwwoods, bugzilla or fedora-kernel@
wwoodsmarkmc: really? I thought we disabled it until just recently
markmcwwoods, okay, re-enabled on 2009-01-13
markmc* Thu May 22 2008 Dave Jones <davej@redhat.com>
markmc- Disable CONFIG_DMAR. This is terminally broken in the presence of a broken BIOS
markmc* Tue Jan 13 2009 Kyle McMartin <kyle@redhat.com>
markmc- Enable CONFIG_DMAR (and GFX_WA and FLOPPY_WA) requested by Gerd.
wwoodsexcellent. IIRC We've had some reports of rawhide kernels not booting recently, so I'll direct testers to try intel_iommu=off
markmcwwoods, great
wwoodsif that fixes their problem, what info is needed to update the blacklist? output of dmidecode?
markmcwwoods, not sure, TBH
markmcwwoods, iommu@lists.linux-foundation.org is probably a sensible place to send the details
wwoodsfair 'nuff. I'll just direct everyone to a single bug and if we see problems we'll comment there
wwoodsthanks for the info
jds2001alrihgt, I saw seven +1
jds2001so we've approved the KVM PCI Device Assignment feature
jds2001.fesco 31
zodbotjds2001: #31 (SSSD - https://fedoraproject.org/wiki/Features/SSSD) - FESCo - Trac - https://fedorahosted.org/fesco/ticket/31
markmcthanks guys
nottingalas. not system security data daemon?
jds2001thanks alot markmc for taking the time to answer all the questions :)
jds2001nah
jds2001the system security services daemon :)
jds2001cool, i was about ready to say sgallagh wasn't around :)
jwbso this all seems like really cool stuff i'll never sue
jwber, use
nottingsgallagh: so, this is intended to replace the current combination of nss_ldap, pam_krb5, pam_ccreds?
nirikyeah, same here. ;)
jwband someone explain how the term "technology preview" got into a Fedora feature page...
jwbfedora itself is a technology preview...
jds2001hehe i was wondering the same thing.
sgallaghnotting: short-term it will be complementary. medium-/long-term, yes.
jds2001so all this does is provide caching for offline auth?
jds2001or am I missing something totally?
sgallaghjds2001: It's also an integration point for central HBAC management
nottingsgallagh: is the nss stuff client/server so that we're not sucking $random_lib_of_the_day into process-space?
jwbHBAC ==
jwb?
jwboh
sgallaghnotting: Yes, it's client-server.
sharkczhost based access control?
sgallaghjwb: Host-based access control
dgilmorei like the idea of SSSD  i wish it was more complete
jwbthanks.
nottingsgallagh: well, that takes all the entertainment out of nss_ldap bugs
sgallaghSSSD is being developed concurrently with FreeIPA v2
dgilmorei think its going to need lots of testing
jwbsgallagh, what are your chances of getting this testable by Beta?
sgallaghjwb: I don't have the schedule handy, what's the beta date?
jds2001mid-march iirc
jwb3-10
jwbMarch 10
jwbthat's when Beta freeze is
jwbFeature freeze is a week before
f13actually
f13Feature freeze is March 3rd
jwbi said that
f13jwb: we were typing at the same time (:
jwbi win.  w00t
sharkczthe idea is great and should close some gaps
jwbi think it sounds really cool.  i'm just wondering if we should defer until we get a better feel for it being complete in time
f13well.
f13a feature can always be dropped if it's not ready in time
sgallaghSorry, please hold
nottingi'm +1 for the feature - we can drop it/not publicize it if it's not ready
sgallaghConversing with my colleagues on that date
f13it just needs a good contengency plan
nirikyeah, +1 here, and if it's not ready we can try again for f12.
sgallaghf13: Contingency is: don't use it
sharkcz+1 as notting
jds2001+1 if it's that simple.
sgallaghIt's not replacing anything in this release
bpepple+1
jds2001drop it later if need be.
f13sgallagh: sure, I haven't looked at the scope yet to see if there would be any changes that would have to be rolled back.
jwb+1
f13(I'm not actually in FESCo, just lurking)
jds2001f13: you're always welcome here :)
sgallaghf13: Rollback should be limited to removing from nssswitch.conf and pam config
f13sgallagh: see, there /is/ a contingency plan!
sgallaghI put that in the feature page, I thought
f13you may have (:
sgallaghAh, not enough detail. Sorry
jds2001i see six +1's, so we've approved the SSSD feature, with the understanding that if it's not testable by beta, we'll drop it.
j-rod+1, just concerned about testability by beta freeze
*j-rod tardy again
jds2001(same goes for any feature, really)
jds2001j-rod: get on the ball! :D
jds2001.fesco 32
zodbotjds2001: #32 (Instant on Fedora - https://fedoraproject.org/wiki/Features/instant-on-Fedora) - FESCo - Trac - https://fedorahosted.org/fesco/ticket/32
j-rodsorry, saw a shiny object...
bpeppledoes this actually have someone working on it?
jds2001this seemed like more of a wishlist item to me, there's no scope.
f13j-rod: looking out the window at your car again?
j-rodf13: mmmmaaaaaaaayybe
nirikyeah, without someone doing the work, I don't see that we can approve this...
jwbpoelcat, ping?
jds2001he "Don't know the work involved to get a working strip down kernel and X"
*nirik sees no answer to nottings question on this on the talk page
bpepple-1, unless it's clear that someone is actually working on this.
poelcatjwb: pong
notting-1 from me
jwbpoelcat, quick favor to ask.  could you not send features to FESCo that have 0% completion?
nirik-1 here.
sharkcz-1
j-rod-1
jds2001-1 here, no one's actually working on it.
jwb-1
poelcatjwb: ummm, that is not a requirement of the feature process... I can't that support that
jwbok, fair enough
poelcatit was never the intention of the feature process that something had to be done to propose it
jds2001i see 7 -1's, so we've declined the Instant-On Fedora feature, since nobody appears to be working on it.
jwbyeah, ok
poelcatwe can change the process, but it is not fair to legislate as we go :)
jwbyep
dgilmore-1
jds2001onto the FPC report
jds2001.fesco 29
zodbotjds2001: #29 (Review FPC Report - http://tinyurl.com/bfuuld) - FESCo - Trac - https://fedorahosted.org/fesco/ticket/29
---sdziallas_ is now known as sdziallas
poelcatif there is no one to actually work on, that makes a little more sense, but just saying 0% done, therefore not accepted doesn't make sense to me
*bpepple doesn't have any objections to the FPC guideline updates.
nirikI'm a bit worried about how vuage the 'Non obvious items' thing is... but hopefully people will use common sense.
dgilmore+1 to explictRequires
jds2001yeah, same here.
jds2001but +1 since I hope folks will use common sense.
dgilmoreHaskell id like to know what the actual changes were
nirikI guess it can be revised if it causes problems... no objections here to the other items.
sharkcz+1, but the explicit requires can have some support directly in rpmbuild
notting-1 to ExplicitRequires, both because non-obvious seems to be overly specified, and the ExplicitRequires isn't confined to library dependencies
jds2001http://fedoraproject.org/w/index.php?title=PackagingDrafts%2FHaskell&diff=72514&oldid=62373
dgilmoresymlinks +1
jwbmaybe we should take these one at a time
jds2001yeah
jds2001so for explicitrequires, I'd like to see a little less guidance on what "non-obvious" means.
jds2001rwmjones hit the nail on the head I think there.
dgilmoredont existing guidelines make "Since <code>%{_docdir}/ghc/libraries</code> is owned by <code>ghc-doc</code> it is recommended to subpackage haddock documentation in a <code>doc</code> subpackage, which can require <code>ghc-doc</code"  a must
---stickster is now known as stickster_food
nottingfor example, i don't think every package that doesn't use autoconf needs comments on why %configure isn't used
*rwmjones did?
rwmjonesah ok
*nirik is confused. We are talking about which item?
jds2001explicitrequires
jds2001which includes a clause about commenting "non-obvious" items
sharkczit is a SHOULD but not MUST
nirikok, and the non-obvious part of it?
jds2001and enumerates those items.
jds2001yeah
nirikwell, some of those... (but not limited to)
nottingthe ones that seem like they should remain enumerated are modified tarballs and license/legal changes. but aren't those covered elsewhere in the guidelines?
j-rodsomething adding a note about commenting non-obvious stuff that is completely unrelated to Requires: seems, uh, out of place, on a page titled 'explicitRequires'
dgilmoresharkcz: if you put files in the filesystem not provided by the filesystem package  and you dont own the path.  you need to Requires the package that does own the path
nirikI see no problem with that list in particular...
jds2001yep
nirikj-rod: yeah, thats confusing me too... they seem pretty seperate
j-rodadding a comment about modified tarballs has nothing to do with explicit Requires
sharkczdgilmore: I mean the comments are SHOULD
j-rodso while I think commenting non-obvious things is generally a good idea, this doc isn't the right place for the guidelines
dgilmoresharkcz: yeah im getting ahead  sorry
nirikj-rod: I agree
jwbhave real life meeting in 1min
j-rodnirik: although... they're under a 'addition to packaging guidelines' heading in the page
j-rodso maybe its just the page title that is suckage
nirikcould be.
jds2001and I assume this goes into the main guidelines page once approved
jds2001rather than a subpage off of it.
j-rodbut then, what notting said: aren't most of these already covered elsewhere?
sharkczthe comments section should be somewhere under "spec readability"
niriksome of them are... this expands that I think, to 'you should document weirdness in your spec'...
sharkczhttp://fedoraproject.org/wiki/Packaging:Guidelines#Spec_Legibility
sharkczis the place I think
jds2001yeah, I was just looking for that :)
nirikso do we object to it? or just want it split out and integrated in a nice way?
*nirik doesn't have a problem with the guideline, but agrees it shouldn't be in this page at the end.
jds2001I'm not entirely sure that it needs to be enumerated such as it is.
nottingi agree with jds2001
sharkczimo split, put the "comment for non-obvious items" into ^ and make Explicit Requires separate paragraph
sharkczI think the list plus the staring sentence is nice legal work :-)
nirikyeah, you can never list everything... perhaps it should have just more explaination? ie "In order for other people reading your spec to understand it, please comment anything that you do that is non obvious or changes from standard/defaults"
nottingand i'd like Explicit requires to be somewhat rephrased/softened. we don't need to be explaining why the kernel requires mkinitrd, etc.
sharkczs/staring/starting/
nottingfor example, rephrase it as "Packages should only contain explicit Requires on other packages that are necessary the packages to function. Library dependencies should be added automatically by RPM's build process." or something along those lines
sharkczmkinitrd is an obvious item
jds2001and obvious to whom?
jds2001mkinitrd is obvious to everyone here, but joe blow might not know what it does.
nottingif you don't know what the requirements do, perhaps you shouldn't be modifying the spec
jds2001agreed :)
nottingi just find it odd that a guideline that started because a package had explict "Requires: libfoo libbar" now starts off with "Packages must not contain Explicit Requires: within the spec file, except ..."
nirikok, so where are we? kick back to FPC with comments to rework?
jds2001yeah, I think so.
jds2001all in favor of kicking back ExplicitRequires to FPC with comments to rework?
jds2001+1 here
sharkcz+1
jwb+1
notting+1
nirik+1 I guess...
bpepple0
jds2001I see five +1's, so we've rejected the ExplicitRequires guideline, with comments to rework.
jds2001on to Haskell guidelines....
jwb0
notting+1, seems fine from the diff
sharkcz+1
dgilmorejds2001: Haskell i had an issue with
nirik+1 here, does this require changes to existing packages? hopefully not too much churn
dgilmoredont existing guidelines make "Since <code>%{_docdir}/ghc/libraries</code> is owned by <code>ghc-doc</code> it is recommended to subpackage haddock documentation in a <code>doc</code> subpackage, which can require <code>ghc-doc</code"  a must
jds2001yeah, I thought that too.
dgilmoreif you put files in the filesystem not provided by the filesystem package  and you dont own the path.  you need to Requires the package that does own the path
jds2001and splitting docs is alredy part of the guidelines too.
dgilmoreunless im misremebering the guidelines
dgilmorejds2001: splitting doc subpackages is only for packages with large docs
dgilmorei guess i misread what i posted
jds2001yeah, with large being a judgement call.
dgilmoreif you put docs in %{_docdir}/ghc/libraries  your going to have to Requires ghc-doc
dgilmorebut they are recommending you split those docs off
dgilmorenot as clear as it could be but its ok
jds2001+1 here too.
jds2001i see five +1's, so we've approved the Haskell Guidelines.
jds2001symlinks?
dgilmoresymlinks +1
jds2001+1
jwbwhy does symlinks exist
jwbwe need to explain to people how to use symlinks?
jds2001i guess it's to make explicit that either way is acceptable?
dgilmorejwb: most kde packages use symlinks for docs  so that khelp works
jwband?
sharkczwas there any issue that this guideline tries to solve?
jwbsharkcz, yeah that's a better way to phrase my question
sharkczjwb: :-)
nirikperhaps rpmlint could stop yelling about links, but thats not a guideline thing.
dgilmorei know in the past we say symlinks must be relative
dgilmores/say/said/
nirikdgilmore: rpmlint says that... but it wasn't a guideline I don't think
dgilmorenirik: hmm it was always treated as a guideline
jds2001yeah, reading the IRC log from FPC
nirikshouldn't be. Unless i missed where it was added.
jds2001seems to be a reaction to https://bugzilla.redhat.com/show_bug.cgi?id=474057
buggbotBug 474057: medium, low, ---, than@redhat.com, ASSIGNED, meinproc4 creates absolute instead of relative symlinks for the common directory
---stickster_food is now known as stickster
nirikthat could be closed->Notabug I think... but if there is a consistency issue, it would be nice for say all the kde packages to do the same thing with links.
sharkczperhaps the connection is that the guideline is "quiet rpmlint" and rpmlint complains about absolute symlinks ...
nirikI think many people misread that... the guidelines say that rpmlint must be run and the results pasted in the review. Nothing about if they should be all dealt with. Sometimes rpmlint is wrong.
jwbso we're writing guidelines to contradict rpmlint output
jwbfix rpmlint instead?
dgilmorejwb: it will need fixing anyway
nirik"Run rpmlint on the rpms to examine them for common errors, and fix them (unless rpmlint is wrong, which can happen, too)."
jwbi have no real objections to the guideline, but it seems superfluous to me
dgilmoreim +1  doesnt hurt to make it clear
dgilmoreneed to make sure a bug is filed for rpmlint
jds2001still only have two +1's here, anyone else?
*nirik has no problem with it, +1 here
notting+1 (although fixing rpmlint never hurts)
sharkcz+1
jwb+1
jds2001alright, I see five +1's, so we've approved the symlink guideline.
j-rod+1
jds2001er, 7 +1
jds2001:)
j-rodCRAP. late again...
j-rodI suck
j-rodoh well
*jds2001 is just too fast :)
jds2001that's all I had for today.....
---jds2001 has changed the topic to: FESCo Meeting - open floor
jds2001anybody got anything they want to talk about?
sharkczQA guidelines ?
nirikdid we want to talk about the cloning bugs thing?
jds2001yeah, Kevin withdrew his complaint, but I think we should talk about it anyhow.
bpepplenirik: didn't kevin drop that.
sharkczjds2001: I agree with your mail
jds2001did my reply to the thread make sense?
nirikbpepple: yeah, he did.
nirikjds2001: sure. I agree that any big changes in workflow should come thru fesco...
nirikI guess we can leave it at that
jds2001anything else on that topic? (or any other topic)?
*jds2001 will end the meeting in 5
jds20014
jds20013
jds20012
j-rod7
bpepplejds2001: you might want to remind folks at the beginning of the meeting who is responsible for writing the summary.
jds2001bpepple: oh right....
bpepplejust so they don't forget.
jds2001jwb: you're on the hook for one more week :)
jwbyep
jds2001nirik: then it goes to you :)
nirikthanks for running things jds2001... and for writing up the summary jwb.
nirikjds2001: ok.
jwbnp
jds2001np
jds2001-- MEETING END --

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!