--- bpepple has changed the topic to: FESCo meeting -- Meeting rules at http://www.fedoraproject.org/wiki/Development/Schedule/MeetingGuidelines -- Init process
dwmw2_gonebrb. reboot
bpeppleFESCo meeting ping -- bpepple, caillon, c4chris, dgilmore, dwmw2, f13, jeremy, jwb, notting, spot, nirik, tibbs, warren
Hi everybody; who's around?
* warren back
nirik is here.
jeremy is here
notting is here
* bpepple wonders if poelcat is about, if so we could start off with Features.
* spot is here
jwb is here
c4chris is here
poelcat here
tibbs-phone here
bpeppleok, looks like most folks are here, so we can get started.
--- bpepple has changed the topic to: FESCo-Meeting -- New Features - http://fedoraproject.org/wiki/Features/Dashboard - poelcat
bpepplepoelcat: you want to lead?
--- poelcat has changed the topic to: vote http://fedoraproject.org/wiki/Releases/FeatureXULRunner
poelcatsure :)
* tibbs-phone is without power.
notting+1, it's already in
bpepple+1 here also, but I do have a question.
dwmw2and even seems to be working
caillonbpepple, yes?
* dgilmore is here
bpepplewhat do we do if some of the packages that depend on xulrunner aren't ported over by F9?
caillonbpepple, drop them?  we've got pretty much all of them done.  the list needs to be updated.
jeremybpepple: I suspect they have to be dropped
jeremyunless caillon wants to do compat-firefox ;-P
* jeremy hides
jeremyoh, and +1
bpeppleok, just was wondering.  thanks.
tibbs-phoneCouls someone echo the tpoic changes to me? My client isn't showing them to me.
caillon0 from me, btw.
* poelcat notes the +1s have it
--- poelcat has changed the topic to: vote http://fedoraproject.org/wiki/Features/SecurityAudit
poelcattibbs-phone:  vote http://fedoraproject.org/wiki/Features/SecurityAudit
f13tibbs-phone: http://fedoraproject.org/wiki/Features/SecurityAudit
* nirik waits for the wiki.
nottingnirik: could be here a while then
nirik+1, this needs more publicity. ;)
f13+1, but what makes this different from the hundreds of other new packages we have?
and is there a plan to install this by default ?
dwmw2f13: I could ask that of many 'features'. I've always been a little confused about the process
notting+1. would like to see some sample reports to make sure they're not espousing crack
* caillon echoes dwmw2
poelcatnotting: can you add that to the feature page?
dgilmoreid like to get a better idea of how this works
caillonbut security gets people's attention
poelcatf13: it is "new" ?
f13poelcat: I'm assuming so.
f13first build 2008-01-23 09:58:49
tibbs-phoneThe wiki looks bad on a Treo, BTW.
f13maybe if this was featured in say a security spin, that might make it a feature?
dwmw2hm. ff3 refuses to show me the sectool page because of an invalid cert
poelcatis this teh 'first' package of its kind in Fedora?
f13poelcat: doubtful.
nirikthere are not many tests.
tibbs-phoneDidn't we have rkhunter?
nirikyes, rkhunter is back... I revived it.
* poelcat notes the +1s have it--please add any questions or comments to the feature page
dgilmoretibbs-phone: we have rkhunter aide etc
niriksample output: http://www.scrye.com/pastebin/28
poelcatanything else on this one?
dwmw2what makes this a 'feature'?
* nirik is reconsidering his +1, but perhaps it's too late?
jwbit's never too late
* notting backs down to 0
f13ugh, that output is not so good
looks like it's bitching about thigns we do by default.
nirikyeah. ;(
nottingand just wrong in some places
* f13 backs down to a 0 as well
dgilmoreme is 0
tibbs-phoneI like the idea of a well integrated tool.
caillonnirik, which mode is that in?  paranoid?
nirikrkhunter has a bunch more tests... but it's specific for looking for rootkits
caillonthere are 5 modes of increasing level of bitchiness
nirikcaillon: -L 5 (so all tests I guess), unclear if thats paranoid.
tibbs-phoneBut it does need to be correct, and shouldn't complain about the default install.
* nirik would like to go to 0
dwmw2we should make this kind of tool complain about over-paranoia too.
WARNING: Your SSH _PUBLIC_ keys are not publicly readable.
nirikthe gui is kinda pretty.
poelcatdwmw2: re: your question about "is it a feature": http://fedoraproject.org/wiki/Features/Policy#definition
dwmw2does it meet any of those 5 criteria?
poelcatdwmw2: that's up to you all to decide :)
caillon(because marketing can't decide this)
dwmw2thanks for the link. IS that new? I don't remember seeing it last time I was confused :)
caillonit seems that it meets 3a.
and/or 5
poelcatdwmw2: no FESCo voted on it before F9 feature process started :)
bpepplepoelcat: correct.
jeremydwmw2: I think we voted on it when you were in mongolia :-)
f13the feature doesn't have to be perfect for us to accept it
nirikIf it was more integrated or had some super fedora related feature it would be 5, but I don't think this is.
* bpepple sticks with his
f13it's in a testible state, we could potentially get it fixed before the final release.
nottingno, but it worries me if it's enough un-perfect
* dwmw2 remains ambivalent
jeremy thinks sectool could be a cool feature, but it should be well-integrated and on by default if we're going to say it's a feature
f13 +1's jeremy
nirik nods at jeremy. agree
jeremyas it stands right now, I don't see anything that makes me think that's the case
f13so lets let it get it's bugs worked out in F9, to make it a "FEATURE" in F10
tibbs-phoneOn by default carries its own issues.
dwmw2yeah. I don't think we're ready for that. Carry the package by all means, but I don't think it's a feature
dwmw2not for f9, at least
* dwmw2 revises to -1
nirik+1 for 'come back in f10 with cool integration and more features'
dwmw2that too
* nirik also goes for -1 for the feature now.
c4chrisok, what jeremy said seems fair enough
jeremycan we also add the feedback of "it'd be nice if you'd blog about what you're working on some too"? :)
notting-1, come back in f10.
warrenOK, -1
* warren behaves like sheep.
* dwmw2 doesn't envy bpepple counting the votes for this one
bpeppledwmw2: yeah, let's get another count of hands.
c4chrismeh, that's why we pay him... oh wait...
caillon+1 because it meets our definition of a feature, IMO
caillonour definition may be dumb, but it's our definition.
dwmw2caillon: not _all_ work that fedora people happen to do upstream meets 3a.
f13caillon: that's our definition of 'can it be proposed' IMHO
not 'should it be approved'
bpeppleok, I see five '-1', one '0', and two '+1'.  anyone else?
* poelcat notes that after a 'recount' this feature is NOT accepted
jeremypoelcat: did you miss any hanging chads? :)
poelcatplease add your concerns to the feature page to help guide the owner
to greatness for F10
--- poelcat has changed the topic to: vote http://fedoraproject.org/wiki/Features/NetworkManager-MobileBroadband
dwmw2tibbs-phone: topic changed to mobile broadband in NM
dwmw2does this include Bluetooth support?
who'd want to connect their phone to their computer with a wire in this day and age?
dgilmore+1  i use it frequently
jwbdwmw2, we already have "improved bluetooth support" as a separate feature
bpeppledwmw2: I don't see any mention of bluetooth support on the feature page.
* dwmw2 uses modemlights_applet instead
nottingdwmw2: it's supposed to
dwmw2jwb: and one of the major missing parts of that was the NM bit :)
dgilmoredwmw2: i have a mobile card built into my laptop
caillondwmw2, bluetooth has it's own feature page and bluetooth is a component of "MoreNM"
dwmw2+1 either way, I suppose
dgilmoredwmw2: it also covers pcmcia and pcexpress cards
bpeppleok, that's eleven '+1'.
--- poelcat has changed the topic to: vote http://fedoraproject.org/wiki/Releases/FeatureFirefox3
jwbtibbs-phone, FF3
jwbthough it renders tracks all funky
dgilmoredwmw2: and with a usb cable and a .fdi file will work for phones..  and im sure bluetooth also
tibbs-phone+1 gots to have it.
bpeppleyeah, there's eleven '+1', and one '0'.
--- poelcat has changed the topic to: vote http://fedoraproject.org/wiki/Features/GlobalMenu
dwmw2tibbs-phone: global menu
nottingbpepple: and the '0' is more of 'my feature, abstain', anyway
bpepple-1, for the reasons mclasen noted.
warrenThis needed to happen upstream a long time ago
* tibbs-phone has power now, but in fsck hell.
warrenconsider for F10 if it becomes more mature and accepted upstream
notting-1. no not-a-chance-of-upstream features, plz
jwbalso, benefit's shouldn't start with "It's just nice."
tibbs-phone0 just not familiar enough with the issue.
warrenHave we ever had such a negative vote before? =)
jeremy-1 given upstream resistance
bpeppleok, I see nine '-1', and one '0'.
warrenjwb, you are so negative.
dgilmore, you are so neutral.
* nirik chants the upstream mantra.
dwmw2not that upstream resistance is always sane, in gnome. But still...
nottingwarren: i'm not sure we've ever had a feature proposed where the upstream maintainer of the package that is being patched explicitly said 'no' before
dgilmorei personally dont much care what gnome does
jwbwarren, i am
nirikshall we move on?
poelcatthat's all i've got for voting
* poelcat notes that http://fedoraproject.org/wiki/Features/Dashboard#novote are not part of the "accepted" feature list. Should any of them be tuned up for Fedora 9 acceptance and voted on by Tuesday (Feature Freeze)?
dgilmorewarren: i was for FF3 and NM-mobile
nottingpoelcat: afaik, voluemcontrol isn't happening
jeremyI've actually got something that I haven't had a chance to write up yet
dwmw2poelcat: want to have a vote on the ppc64 build page size? I'm buggered if I know which is best.
jeremybut which we should probably do as a feature
jwbdwmw2, that's not a feature
dwmw2, it's a separate topic
bpeppledwmw2: that's next on the agenda after features.
jeremylivecd persistence
nottingprobably need to harass clumens & pjones about firstboot and efi
jeremy(well, liveusb persistence)
dgilmorejeremy: its not a feature if you dont propose it
nottingjeremy: surely that's liveusb persistence
jeremynotting: yeah
poelcatalso a sanity check of http://fedoraproject.org/wiki/Releases/9/FeatureList to make sure everything is there that should be
f13if propose you do not, feature it is not.
jeremynotting: I know peter's working on feverishly getting the efi stuff _working_ right now
nottingjeremy: -EBIOS?
* poelcat has tried to stay on top of everything, but could ahve missed one
nirikso, vote by email if submitted before tuesday?
f13that said, I think we could have a special meeting on Monday to approve any last minute features
dgilmoreBuetooth 0% done?
jeremyf13: probably wise
nirikor that.
f13since the freeze will be late monday night
so that tuesday's rawhide == frozen content
tibbs-phoneI have no problem with an extra meeting.
bpepplef13: I'm fine with having a meeting for any last minute features on Monday.
nottingas long as it's not 1PM
* nirik is fine with that.
bpepplewhat time would work best for people?
dwmw2it may shock you to learn that I won't be around next week.
dgilmoredwmw2: sking doesnt count
nottingso, is
warrenpoelcat, K12Linux is at 75% or so
tibbs-phonesomething +-2 hrs of this meeting is fine with me.
nottingso, is  is the 'where in the world is dwmw2' game a f10 feature?
dgilmorethe amount of low completion right now scares me some
bpepplehow about after the rel-eng meeting on monday?
wwoodshere's a question: if I want to get system-summary (the "About This Computer" dialog) added to the default System menu.. does that require a Feature?
dgilmorebpepple: sounds sane
dwmw2notting: heh. For that we have dopplr.com
c4chrisbpepple: what time is releng meeting?
bpepplewhich would be 19:00 utc.
tibbs-phonedgilmore: scares you in a good or bad way?
bpepplec4chris: 18:00
nottingdwmw2: but that doesn't have rockapella
dgilmoretibbs-phone: in a bad way
dwmw2notting: now that _is_ a feature
* c4chris might be able to make it at 19:00 utc
bpepplewwoods: not necessarily.
nottingi'd give a half-hour/hour cushion to the rel-eng meeting, as meeting-on-freeze day could run long
bpepplenotting: sounds reasonable.  let's plan on having it at 19:30 on Monday.
anything else we need to discuss in regard to features?
f13wwoods: I don't think so.
wwoods: just make it happen.
poelcatbpepple: that's all from my side
bpepplepoelcat: great.  thanks!
--- bpepple has changed the topic to: FESCO-Meeting -- Page size on PPC/PPC64 builders - https://www.redhat.com/archives/fedora-devel-list/2008-March/msg00072.html - dwmw2
poelcatyou're welcome
dwmw2not sure what to say that wasn't already in the above-linked mail
* nirik is happy to defer to the ppc SIG/team on this. ;)
nirik: I'm in two minds.
jwbdoes anyone need a paraphrase, or did we all read it?
nottingare we discussing or voting?
dwmw2there are reasonable arguments on both sides
f13my arguments are to leave the builders on the RHEL5 kernel.  Builds that fail due to /kernel/ need to be fixed, and I"m working on having a ppc64 box with the RHEL5 kernel available here in the next week or so.
jwbnotting, both
dwmw2I'm probably with f13 but could be swayed with a feather
f13going to a differnet kernel for the builders will be an ongoing maint headache, and provide a 'gotcha' to maintainers when their packages get built for RHEL
tibbs-phoneI don't think that the extra testing is a bad thing.
nottingmy opinion is that it is failing builds in a way that users can't test with fedora is bad
warrenI'm for 64KiB
jwbnotting, that's my primary concern
nirikis there any idea if rhel6 or ongoing rhel will keep that pagesize?
f13notting: I agree, and that's why I want to provide the test box
warrenHow many packages were found to be broken by this?
* spot abstains
dwmw2nirik: rhel6 will almost certainly keep it 64KIB
tibbs-phonebut it means that one platform that almost nobody has shows the difficult bugs.
nirikwarren: 2 I think
jwbnotting, as f13 said, we're working on getting a RHEL5 ppc64 box FASized
f13notting: it's sitting here, waiting for me to figure out wtf is going on with selinux, before I drag it to olpc.
warren2 out of how many?
build fails or runtime fails?
dwmw2warren: not many, and it was _generic_ bugs which were found
c4chriscould someone clarify: 1 - the build fails ? or 2 - teh built package does not work ?
f13the build failed.
jwbalso, i'm questioning glibc's behavior to being with in the mysql fail
nirikbuild failure
nottingthere are plenty of ways to check for bugs (MALLOC_CHECK_, MALLOC_PERTURB_, umask, etc.) that we don't do in the buildsystem; i don't see why page size should be special-cased
jwbf13, mysql was actually a test fail
f13jwb: as part of the build.
warrenand a fix for mysql was found?  suitable for upstream?
f13the build /task/ failed.
jwbwell, yes.  but it wasn't a _compile_ fail is my point
dwmw2notting: also there are plenty of things we do which _are_ only relevant at build time, not run time.
tibbs-phonefinally out of fsck.
dgilmoref13: it cant live at the OLPC office  most likely
f13dgilmore: bwah?!
nottingneed the extra space for negroponte's replacement? :)
jwbok, let's get back to that
c4chriscan a package built on 64KiB pages then mysteriously fail when deployed on a 4KiB page machine ?
dwmw2I agree with the argument about consistency with what Fedora ships.
dgilmoref13: they seem to be cracking down on non olpc owned hardware permanently hosted in the office
jwbc4chris, typically not
nirikas long as a super fast link isn't needed, you're welcome to ship it here. ;)
nottingc4chris: if it's dumb, yes. (a dumb package can also fail on 64kb when built on 4kb)
dwmw2although I'm also inclined to consider consistency with what RHEL will ship, and that's more relevant because building on 64KiB apges and running on 4KiB pages works much better than vice versa
warrendwmw2, Fedora's kernel isn't 64KiB page sizes? why not?
jwbwarren, because it makes no sense
dwmw2warren: because that would be stupid. IBM want it for their workload on their machines.
nottingwarren: Really Bad Idea on lowmem machines (ps3, etc)
jwbwarren, i sent an email about that in the thread
dwmw2bad idea on most machines, really.
warrenPS3 is barely usable anyway, but I digress...
dwmw2it makes me shudder to think that we switch back to 4KiB _just_ because it's finding generic bugs and making us have to fix them :)
c4chrisI tend to like having bugs shaken out of packages
dwmw2yeah, I like that too. I just wish we could do it on x86_64 so people stop bitching about ppc :)
warrenHow difficult will it be to provide a host for people to do test builds on and fix their software?  If we can provide that to FAS cvsextras auth, then I think it is feasible to stick with 64KiB.
dwmw2s/stop/reduce their/ :)
we will definitely provide such a host
dgilmoredwmw2: is it just the mismatch in page size that found the bugs  or was it that its not 4KiB
jwbwarren, have you not been paying attention?
nottingi don't see 'it finds a bug' as a valid reason for building in a non-reproducible-on-fedora way
jwbwe've already said we were working on that like three times
f13warren: we probably can't FAS it, due to security problems, but we can provide a machine.
nottingby that logic, we should push every build through coverity
dwmw2dgilmore: it's mostly the latter.
f13notting: I don't see "it broke one package" to create an on going maint headache by having a custom kernel for our RHEL5 install
dgilmoredwmw2: so we will find such things  sparc has a minimum page size of 8KiB
f13notting: because unless we create full virtual machines for each and every build, they won't match the "fedora" they're building for
nirikdid this sort of thing ever happen back in the past with internal buildsys?
jwbnirik, no those are rhel4
dwmw2it happened around the FC6 time
f13nirik: yes, around FC6 time
dwmw2and we fixed all the packages back then.
jwboh, i wasn't aware of that
f13nirik: internal builders were RHEL4, external were 64KiB
nottingnirik: it affects what we can do with glibc
dwmw2mostly, the work is _done_ now
nirikyeah, so lets just do that again... fix packages and go on.
nottingf13: maintenance load is a separate argument
nirikreally this only affected 2 packages and they are both already fixed aren't they?
nottingf13: *cough* TUX
f13notting: that's fixed in RHEL proper
notting: we're running fully stock RHEL kernels.
jwbnotting, yeah.  different issue
dwmw2nirik: 2 packages since December when we changed, yes. And they were both real bugs.
nottingf13: it's still an issue causedfrom the same source
f13notting: yes, and /fixed/ in the same source, instead of just going "LALALALA" and waiting for a customer to find it.
dwmw2in fact, they may both have been the _same_ bug, if we consider it a bug that glibc's pthread_attr_setstacksize() doesn't actually set the damn stack size to the number you ask for ;)
jwbdwmw2, i asked jakub about that in the bug.  he's ignoring me so far
dwmw2I'm not going there :)
c4chrisso, do we vote ?
dwmw2I'm inclined to go with 64KiB pages. If it turns out to be a pain for packages, the workload mostly falls on me anyway as I trawl the FE-ExcludeArch-ppc{64,} bugs
dwmw2and we can always reconsider if it _really_ turns out to be a pain (which it hasn't so far)
tibbs+1 for 64k pages.  We can always revisit this in the future.
* dwmw2 kicks dgilmore
warrenIf everyone gets 64 votes, then people voting for +4 automatically lose...
dgilmoredwmw2: not nice
jwbit's passed
bpepplewarren: ;)
ok, anything else? Or should we move on?
dwmw2dgilmore: was that a serious vote? What for?
tibbsnotting: Your argument about reproducibility is interesting, though.
tibbsHow impossible would it be to occasionally build 64k page Fedora kernels for ppc64 so that people can reproduce if they want to?
dwmw2tibbs: it's certainly a concern. But since I also removed PAGE_SIZE from <asm/page.h> (and in fact also now removed <asm/page.h>, we're _forcing_ people to use getpagesize() instead of hard-coding a constant now :)
tibbs: not hard at all.
jwbtibbs, not hard.  yet ew
dwmw2and we'll definitely be providing build/test hosts.
tibbsIt's just a config tweak and a koji scratch build.
dgilmoredwmw2: it was a vote for 64KiB
nottingdwmw2: oh, they can still hardcode a constant if they try hard enough
dwmw2notting: yes, but then we would be justified in visiting them in the wee hours and promoting an attitude of violence
jwbi would like to point out that sparc requires 8KiB
bpeppleok, I see eight '+1', one '-1', and two '0'.
dwmw2dgilmore: oh, ok -- I thought you were just being silly.
f13tibbs: the thing is, the people complaining about this, don't have a ppc machien to run said kernel on to begin with
jwbbut then again, fedora will run like that too
tibbsf13: Then the reproducability argument doesn't make much sense; you could extend it to anything but x86.
f13tibbs: thats some of the argument.
tibbsThe people who will be complaining about sparc and arm builds won't have those machines either.
dwmw2they should have access to such machines.
dgilmoref13: the people complaning about it were RH employees  who i believe you said they had access to internal RH machines to do the testing
f13dgilmore: the internal machines weren't working
dgilmore: because to use them, you had to install Fedora, which wasn't installable
dgilmoref13: oh i did not hear that
f13granted, there were ways to work around it, if the person actually /tried/
or engaged anybody beyond a single email every few days to public lists
dgilmoretibbs: there will be a sparc machine for testing builds on
dwmw2jlaska gave me a F8 host auto-installed.
* bpepple looks at the time, and thinks we need to move on.
jwbyes, 64KiB pages passed
f13dwmw2: but that's neither A) rawhide, nor B) 64KiB
c4chrisbpepple: +1
dwmw2f13: we installed a special kernel on it.
and I was using it to fix rawhide installation :)
f13dwmw2: right, as I said, it could be worked around with some effort.  none was applied.
dwmw2which ought to work as soon as we have the 64-bit install image back again
f13bpepple: what's next?
--- bpepple has changed the topic to: FESCO-Meeting -- Proposal to make tags in CVS "immutable" - f13
f13ah right.
dwmw2I don't like this. Mostly because I'm stupid
f13this is for our long term goal of being able to take the nvr of a package and get back to the cvs tag used to create it, and be confident that it's the same content
dwmw2I often make one-line changes to patches and retag them, or add the forgotten patch and retag
f13so that we can have a web tool for people to regenerate srpms on demand.
tibbsIf only we could make them immutable after a successful build against them.
warrenThis means force-tag wont work anymore?
f13let me finish
*BEFORE* this would go into effect, we would investigate the most common uses of force-tag, and create ways of mitigating the problem leading up to a force-tag call, as to eliminate the need for force-tag.
dwmw2is there any realistic prospect of us moving to another scm?
dwmw2my concern is that it's hard to mitigate my stupidity. I've spent over three decades trying
f13dwmw2: maybe.
nottingnot in the very near future, no
spoti hear good things about bitkeeper.
tibbsjwb: Ever, or just in the time frame that this problem needs to be solved?
jwbtibbs, within this timeframe
dwmw2tags in git are dead easy to keep :)
jwbnever say never
nirikadding more checking logic to 'make tag' might fix most of the cases.
f13nirik: that's my intial suspicion
this change won't be thrown over night
nirikie, 'you didn't cvs-add a file' 'or you have unchecked in changes to this'
f13we're coming to FESCo to ask that if we /can/ mitigate the need, can we make tags immutable.  Can we spend the effort?
bpepple+1 to f13's proposal.
tibbs+1 to that.
nirikor 'your source doesn't match, you didn't update sources and upload new tar.gz'
warren+1 if we make make tag smarter
dgilmorewarren: thats a given
warrenhow does the server enforce immutable?
appendonly or something?
tibbsCVS has a hook that can check this.
it passed
tibbstaginfo, I think.
bpeppleok, I see eight '+1', so f13's proposal has passed.
bpepplemoving on.......
--- bpepple has changed the topic to: FESCo meeting -- Free discussion around Fedora
f13it'll probably take a while to impliment, and will be brought back to FESCo before we throw the switch
bpeppleanything else that needs to be discussed today?
* jeremy has another meeting to go to
bpepplef13: cool. thanks.
c4chrisnothing here
dgilmorenothing from me
bpeppleoh, one quick note.
tibbsActually making tags immutable is absolutely trivial, BTW.
nirikyeah, making 'make tag' smart is the hard part.
bpepplestickster asked us to define our role better, so that was the reason for my note to the fesco-mailing list.  Right now it's just caillon and me working on it, so if anyone else wants to help just contact me.
tibbsI saw that and would like to help, but I'm swamped with work at the moment.
jwbbpepple, i saw that.  i'd be happy to review/comment on what you have
bpepplejwb: cool, once we have something I'll forward it to you to look at.
bpeppleok, if there's nothing else, I think we can wrap up for this week.
* bpepple will end the meeting in 60
caillonoops, got distracted
* bpepple will end the meeting in 30
bpepple will end the meeting in 15
bpepple-- MARK -- Meeting End
Thanks, everyone!

Generated by irclog2html.py 2.5 by Marius Gedminas - find it at mg.pov.lt!