fedora-meeting
LOGS
19:59:59 <jonmasters_> #startmeeting
19:59:59 <zodbot> Meeting started Wed May  2 19:59:59 2012 UTC.  The chair is jonmasters_. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:59:59 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
20:00:13 * pbrobinson is excited.... I think
20:00:27 <djdelorie> and after this, the FEDORA LEG MEETING!  ;-)
20:00:32 <jonmasters_> #chair jonmasters ctyler pbrobinson dgilmore bconoboy orc_fedo
20:00:32 <zodbot> Current chairs: bconoboy ctyler dgilmore jonmasters jonmasters_ orc_fedo pbrobinson
20:00:45 <orc_fedo> 'ARM' becuase the BBC Acron will live forever
20:00:49 <jonmasters_> Alrighty...let's round up the troops. Just a moment. STAY TUNED!
20:00:49 <orc_fedo> Acorn
20:01:36 <orc_fedo> second prize is TWO weeks in Akron
20:01:53 <jonmasters_> lol
20:02:27 * maxam is in
20:02:55 * jwb lurks
20:03:32 <jonmasters_> Ok, let's do it.
20:03:41 <jonmasters_> #topic 0). Current build status
20:03:53 <jonmasters_> #topic 0). Current build status - Gnarly bugs and build failures
20:04:16 <jonmasters_> so...? Current bugs that have come up since last week. Anything exciting, pbrobinson ?
20:04:45 * jonmasters_ is looking over the latest emails
20:05:02 <pbrobinson> jonmasters: we've had another ICE come up in gcc, new build is building at the moment
20:05:11 * maxam fossjon agreene Frojoe (the CDOT ARM folks) are in
20:05:22 <pbrobinson> and I'm literally going through diffs to fix random packages now
20:05:34 <jonmasters_> ok, let's divide out F17 and rawhide. For F17 failures...
20:06:11 <pbrobinson> llvm isn't closed off and is blocking most of our outstanding builds where we're missing single NVR revisions
20:06:29 <jonmasters_> pbrobinson: I'm going to do LLVM. Nobody in the tools team has cycles
20:06:57 <jonmasters_> unless someone more suited volunteers right now, I'm on it as top priority from tonight on
20:07:02 <pbrobinson> it would be good to have a fix for llvm this week so it can't be tagged into the f17 final stable
20:07:11 * djdelorie has ordered five more days from eBay but they haven't shown up yet...
20:07:30 <jonmasters_> #action jonmasters to fix LLVM
20:07:35 <jonmasters_> ok, what else?
20:07:53 <ctyler> jonmasters_: I've been way short of cycles for llvm to this point but have some this week, would be pleased to hack on that with you
20:08:08 <jonmasters_> On the plus side, I fixed audit in the kernel, and we discovered the userspace package never worked on little ARM - but that's being fixed, not a blocker or an issue
20:08:17 <jonmasters_> ctyler: ok, I'll keep you informed
20:08:30 <jwb> jonmasters_, iirc, your fix isn't what upstream is going to go with, right?
20:08:38 <jwb> jonmasters_, or did rmk take your patch over will's?
20:08:49 <pbrobinson> the remainder of f17 is a number of atomics related builds which bconoboy says the openmpi patch is a good example of a fix so if people could look at patches for those if people have a cycle here and there
20:08:56 * jwb is asking because i've been trying to follow which patch to grab
20:09:25 <jonmasters_> jwb: right, I only fixed userspace register corruption. Then I spent all night reading the audit code to understand what the rest of it was doing. I talked with Will this morning, we're going with his because the original code also was looking in the wrong place for syscall entry/exit
20:09:45 <jwb> k
20:09:56 <jonmasters_> jwb: the original code in addition to breaking userspace also didn't have the right logic. I have now tested the additional patch and spoken with Eric and Russell privately about fixing audit userspace
20:10:58 <jonmasters_> for the record, audit userspace assumes big endian because there was an (not unsurprising) confusion about what the different ABIs meant on ARM. That will be fixed. For now, audit actually doesn't do anything and we happen not to notice because Fedora has only one rule "delete all rules" :)
20:11:24 <jonmasters_> it could be worse, certain - cough - other distros ship audit and it does *nothing* and in facts corrupts userspace if you turn it on...so that was well tested by them ;)
20:11:46 <jonmasters_> anyway...other than LLVM, pbrobinson what about F17 bugs beyond that?
20:12:01 <jonmasters_> sorry to look to you pbrobinson I've been bogged down a bit, know you have too
20:12:03 <orc_fedo> jonmasters_: I am sure you wrote and upstreamed a unit test to demonstrate that
20:12:42 <jonmasters_> orc_fedo: oh, you just need to turn audit on by running auditd in userspace and the system will fall over
20:13:02 <pbrobinson> jonmasters: most of the packages aren't looking bad, there's a few harder ones like fpc that could people can look at, if people have cycles they can ping me and I can give a current list of packages
20:13:19 <pbrobinson> I've been fixing a lot of broken mainline packages so we can build them on ARM
20:13:19 <jonmasters_> pbrobinson: can you send another problem children email? (packages)
20:13:41 <pbrobinson> jonmasters_: sure, I'll follow them up with the one from last week
20:13:53 <FunkyPenguin> pbrobinson: sorry for the noob question, but is there an easy webview of package status for fedora on arm?
20:14:14 <jonmasters_> ctyler: (jonmasters looks to stage left and gives a pondering gaze toward G+) you say you're getting into full gear for the summer...can your guys start by automating the problem children package list emails?
20:14:16 <dgilmore> hey all
20:14:23 <jonmasters_> FunkyPenguin: ^^^ :)
20:14:24 <pbrobinson> a lot of fixes aren't glamorous and need to be fixed on mainline. I've been fixing packages broken from F-11 and 12 :-/
20:15:01 <jonmasters_> ctyler: this is the delta comparisons between PA and SA, never built, built, etc.
20:15:06 <ctyler> ah
20:15:09 <ctyler> yes
20:15:13 * ctyler looks over at fossjon
20:15:14 <pbrobinson> FunkyPenguin: no but koji queries can tell you. If you query mainline f17 tag and get a package != fc17 it's broken and needs to be looked at
20:15:25 <jonmasters_> would be good, then pbrobinson can co-ordinate with you and we can move it from a manual to mostly auto process
20:15:46 <jonmasters_> ctyler, fossjon: do you think we might get something in a week or two?
20:15:49 <ctyler> #action fossjon to automate problem child e-mails
20:15:56 <jonmasters_> (does not need to be sexy)
20:16:00 <FunkyPenguin> ta
20:16:19 <pbrobinson> ctyler: I can send an overview if the relatively manual process I use if it'd be helpful?
20:16:26 <ctyler> jonmasters_: fossjon says by EOW
20:16:28 <jonmasters_> ok, for rawhide I see lots of Haskell stuff. I know somebody loves Haskell so I'll shut up...but seriously, what other F18+ stuff?
20:16:33 <ctyler> pbrobinson: sure, would be great
20:16:50 <jonmasters_> #action pbrobinson to send an overview of the manual process for automation
20:17:07 <pbrobinson> haskell is broken on mainline too because they're upgrading, juhp is going to look at it this week to unblock rawhide stuff
20:17:07 <djdelorie> aside from failed builds, how are we doing on "keeping up" with PA build rates?
20:17:10 * ctyler thinks of Haskell, Ocaml as hipster langauges
20:17:20 <dgilmore> djdelorie: we are closer than ppc and s390
20:17:30 <dgilmore> djdelorie: not too far behind
20:17:49 <djdelorie> I mean, I think a critical requirement for ARM is to have enough build resources that, over time, we don't fall further and further behind
20:18:15 <pbrobinson> we're keeping up quite well, on rawhide we'll unblock a lot once the latest gcc is built as it should allow a couple of blockers to unclog a lot
20:18:16 <dgilmore> djdelorie: right. as we get enterpise hardware we will know better how we do
20:18:45 <jonmasters_> ok, oops
20:18:51 <djdelorie> and enough hw to not fall over when we pick up a new release?  Is F18 == rawhide at the moment?
20:18:59 <ctyler> djdelorie: yes
20:19:03 <jonmasters_> #topic 0). Current build status - Sending out regular problem package mails? (already covered)
20:19:17 <jonmasters_> shall we go onto atomics?
20:19:30 <jonmasters_> #topic 0). Current build status - Atomics on older processors (v5)/LLVM/etc.
20:19:52 <jonmasters_> current recommendation is to just use the openmpi example bconoboy pointed to rather than add it to my todo - all in favor?
20:20:02 <djdelorie> jomasters_: three topic #0's ?
20:20:04 <jonmasters_> (we point people to that, if needed, we write up something around it)
20:20:09 <jonmasters_> djdelorie: they're subpoints
20:20:28 <jonmasters_> djdelorie: ok, 0.1,0.2,0.3 really I guess
20:20:29 <pbrobinson> jonmasters: +1
20:20:49 <djdelorie> +1
20:20:50 <jonmasters_> ok, we point people to that and if they need help, we explain it with a wiki page I or someone with time can write
20:21:06 <jonmasters_> #topic 1). Fedora 17 Beta - What are the constraints on getting this out next week?
20:21:13 <jonmasters_> Yup! It's Beta Time! Wooo!
20:21:20 <djdelorie> define "beta"
20:21:23 <pbrobinson> does anyone have some cycles to use the openmpi patch to produce patches for other packages with atomics issues? I can provide a list
20:21:50 <dgilmore> jonmasters_: we cant push something not defined
20:21:51 <jonmasters_> so dmarlin_ is doing good work, but we won't have a generic image solution this week. Therefore, my proposal is we bless specific versions of bconoboy's images as F17 beta
20:22:19 <jonmasters_> pbrobinson: mostly needs someone who knows the code, except in cases of obvious macros/etc.
20:22:19 <djdelorie> do we need to "bless" the repo too?  I've been yum updateing my F17 for a while now with no problems...
20:22:55 <jonmasters_> dgilmore: true. But worst is doing nothing. We did an alpha that was just whatever (no offense) pbrobinson put together, so I think at a minimum we should push something from bconoboy's images soon
20:23:04 <dgilmore> djdelorie: the repo is ever changing.  we could make a Fedora repo that matches primary install tree
20:23:07 <jonmasters_> shall we talk about the release requirements then?
20:23:21 <pbrobinson> I'm not sure that blocking the repos makes much sense at the moment, I'm working as much as possible so we'll be able to have a repo as close to the f17 gold as possible for final
20:23:21 <ctyler> pwhalen: ping!
20:23:22 <pwhalen> heres a cleaned up version of the release criteria ctyler and myself worked on - http://etherpad.proximity.on.ca:9001/p/k8c7SAPEhA
20:24:41 <ctyler> NB. this is different from the one from earlier today in that the strikeout text is removed ^
20:27:04 <djdelorie> do we have a known-good qemu kernel now?  Frame buffer et al ?
20:27:21 <jonmasters_> ok, so what's the action here? Are we going to review this later and take the F17 beta discussion to email/possibly have another sync here before next week?
20:27:21 <pbrobinson> djdelorie: almost
20:27:34 <jonmasters_> pbrobinson: you switched to vexpress, right?
20:27:43 <pbrobinson> jonmasters_: yes
20:27:50 <jonmasters_> ok, I can test
20:28:08 <djdelorie> we'll need a write-up wiki page for each target we support
20:28:20 <jonmasters_> ctyler: btw, where do we stand with the nice generic virt-manager libvirt provisioned qemu stuff?
20:28:42 <jonmasters_> djdelorie: yea, should not block beta, but it won't make it useful without at least a couple of them in place
20:28:46 <pbrobinson> jonmasters_: I know rich jones has been working on that
20:28:54 <pbrobinson> or at least investigating it
20:29:02 <jonmasters_> so we're only suggesting Panda and vexpress for the beta?
20:29:09 <ctyler> jonmasters_: if we have the kernel pieces together, the rest is in place
20:29:20 <pbrobinson> what about trimslice?
20:29:22 <jonmasters_> I think we should try to hit vexpress, Panda, and perhaps TS
20:29:35 <ctyler> yeah, wasn't sure about including TS
20:29:36 <jonmasters_> (I'll get you a working TS kernel)
20:29:40 <ctyler> ok
20:29:43 <bconoboy> highbank?
20:29:50 <pbrobinson> and it would be nice to get a remix of the Raspberry Pi too with X as I'm being asked about it regularly
20:29:58 <ctyler> do we have an open video driver for TS or PB?
20:30:11 <ctyler> pbrobinson: remix is in the works. Fighting kernel build issues atm.
20:30:23 <pbrobinson> is there a means of people other than those directly involved to run highbank?
20:30:42 <jonmasters_> I think we should ask Mark L. about that
20:30:48 <dmarlin_> jonmasters_: do we have a kernel fix for 1) TS USB resets, and 2) device-tree on Panda ?
20:30:50 <jonmasters_> if there is, then we should consider it
20:30:52 <ctyler> I think we should start with a really tiny set of targets and grow it rather than the other way around.
20:31:01 <pbrobinson> ctyler: I would poke gregkh too see how his rasp pi rebase is going
20:31:09 <ctyler> We're talking about blocking on these devices, not limiting support to them too.
20:31:22 <jonmasters_> dmarlin_: device tree on Panda no. Well, sorta. I built one with dtb support, but it blew up on our existing panda dtb and I've not had time to look - let's ask upstream about that
20:31:32 <jonmasters_> ah right, yea, thanks ctyler
20:31:41 <pbrobinson> jonmasters: because IMO we're better off spending time on something other than highbank as it's not available
20:31:53 <jonmasters_> #idea all in favor of the limited blocking list ctyler posted? (perhaps with TS added)?
20:31:57 <djdelorie> RPi isn't "available" to most of us either though
20:32:06 <dgilmore> ctyler: panda there is DRM_OMAP and the modesetting X driver
20:32:17 <jonmasters_> djdelorie: what do you mean? Mine's coming in September ;)
20:32:22 <ctyler> djdelorie: right, hence excluding from the release criteria list for now
20:32:30 <dgilmore> ctyler: and im working on building and testing the tegra drm driver
20:32:33 * ctyler got a ship date of June 18 for his order
20:32:35 <pbrobinson> djelorie: RPi is more a marketing thing than widely available
20:32:43 <orc_fedo> jonmasters_: I would like to add: have a working image in libvirt there so one can test build
20:32:48 <jonmasters_> well look at goodie "I ordered on time" ctyler ;)
20:32:58 <nb> ctyler, where do you order them at?
20:33:02 <djdelorie> probinson: right, but the limited number of them makes it much harder for "people" to work on them.  At the moment.  In a few months, it'll be very different.
20:33:02 * jonmasters_ was a little slow off the mark
20:33:09 <pbrobinson> jonmasters_: dtb is still missing subsystems so while there's some support there it doesn't appear enough to use it generally
20:33:13 * nb should go ahead and place ambassadors an order in case we don't get ahold of a way to get them any earlier
20:33:27 <ctyler> nb: element 14 between crashes
20:33:29 <nb> ctyler, i assume they don't charge the cc until they ship?
20:33:29 <jonmasters_> #idea we get Greg-KH (of Linux Foundation) to do all the Fedora ARM rPi QE and testing on his device
20:34:02 <pbrobinson> orc_fedo: that's the vexpress qemu support
20:34:04 <jonmasters_> pbrobinson: I think we can ship F17 on Panda without dtb for now and do a kernel update later
20:34:23 <djdelorie> fb for panda?
20:34:37 <pbrobinson> jonmasters_: yes it's looking much better in 3.4 and I think 3.5 will be the usable one
20:34:39 <jonmasters_> djdelorie: honestly, I don't know. I don't use the fb on mine
20:34:49 * ctyler notes that FUDcon coupons should be fulfulled Real Soon Now
20:34:50 <djdelorie> I think "working fb" should be a beta criteria
20:34:59 <jonmasters_> ctyler: we really really need to pay them somehow
20:35:05 <djdelorie> since so much of Fedora is desktop oriented
20:35:13 <jonmasters_> ctyler: can you ping them again about that?
20:35:17 <pbrobinson> djdelorie there is some support, 3.4 (and I think 3.3) has a omap KMS driver that will work with the modesetting Xorg driver
20:35:31 <djdelorie> dumb fb is acceptable :-)
20:35:47 <orc_fedo> 16:34 * ctyler notes that FUDcon coupons ... **snort **
20:35:48 <jonmasters_> ok, I'm going to suggest we take the release criteria discussion to the list, since ctyler and pwhalen only just finished it. That will give us time to digest. Does that make sense?
20:35:58 <ctyler> +1s
20:36:05 <ctyler> s/s//
20:36:06 <djdelorie> is olpc XO-1.75 popular enough to warrant a beta?
20:36:23 <ctyler> djdelorie: I'd let downstream handle that
20:36:29 <jonmasters_> #topic 2). Secondary Architecture Promotion -  Status of ARM toward promotion criteria
20:36:33 <djdelorie> popular enough among people who would beta test, that is.  It happens to have a working fb
20:36:52 <pbrobinson> djdelorie: it's currently got a custom kernel and closed X. It's been running F-17 hardfp for months though
20:37:02 <djdelorie> ah, right, the custom kernel.  Nevermind.
20:37:06 <jonmasters_> bconoboy: you're on the hook now for writing a 10 item response to each point of our current progress, as you noted
20:37:15 <djdelorie> Mine runs with the Fedora X though
20:37:19 <pbrobinson> so it's not really a beta candidate but is very usable on F-17 anyway
20:37:28 <jonmasters_> anyone else have input on the secondary promotion stuff?
20:37:48 <djdelorie> again, we should have a "current status" wiki writup page that's updated as we go
20:37:52 <pbrobinson> jonmasters_: no because I've not had time to read it but I'm sure I will next week :-D
20:37:54 <jonmasters_> we also should write an ARM status email sometime soon. Shall we aim to do that before next week?
20:37:55 <ctyler> djdelorie: +1
20:38:23 <jonmasters_> actually, maybe djdelorie has the better approach
20:38:28 * pbrobinson was hoping to do a blog post over the weekend to coincide with the freeze
20:38:47 <jonmasters_> maybe if we write a wiki page and keep it updated, but make a concerned effort that at least once a month it's very up to date
20:39:03 <ctyler> we suck at wiki gardening
20:39:16 <jonmasters_> then we can post a link in with the meeting announcement each week to remind folks on devel@
20:39:17 <djdelorie> but email and irc are too transient.  We want to be as "findable" as possible
20:39:38 <pbrobinson> I agree, we need to suck less in the lead up to PA
20:39:48 <jonmasters_> well, I think if we make it our problem to at least update that once a month...let's get a wiki page up before next week
20:40:17 <jonmasters_> #action wiki page to be written collaboratively with current status (in lieu of monthly email) and link included in meeting invites going forward
20:40:21 <jonmasters_> ?
20:40:25 <djdelorie> +1
20:40:47 <jonmasters_> ok, anything else on that topic?
20:40:49 <orc_fedo> jonmasters_: who owns it?
20:41:16 <jonmasters_> orc_fedo: perhaps bconoboy wants to "own" it for this week since he's doing the other 10 item response - he can ping us to get us to put things in it?
20:41:17 <jonmasters_> bconoboy: ?
20:41:47 * orc_fedo has pinged last editting person on F wiki items and gotten devnull replies
20:42:08 <jonmasters_> well, we can but try :)
20:42:19 <jonmasters_> any more comments on this topic?
20:42:24 <bconoboy> jonmasters_: still on a call
20:42:48 <jonmasters_> bconoboy: ok, we can figure out later who owns this
20:42:58 <bconoboy> ... and I'm free
20:43:02 <bconoboy> reading scrollback
20:43:05 <ctyler> "still on a call" attracts action points
20:43:39 <bconoboy> jonmasters_: sign me up and let's move on
20:43:42 <jonmasters_> bconoboy: ok, the main question is in addition to the 10 item response of current status, can you own the overall "current Fedora ARM project status" wiki page creation before next week? We'll then include an ongoing link to it in meeting announcements
20:43:47 <jonmasters_> bconoboy: excellent thanks
20:43:59 <djdelorie> SA->PA status, that is.
20:44:00 <jonmasters_> #action bconoboy to own creation of said wiki of current Fedora ARM project status
20:44:14 <jonmasters_> #topic 3). Your topic here
20:44:22 <jonmasters_> Any other business? Or useful stuff?
20:44:28 <jonmasters_> Or heck, excitement?
20:44:44 <djdelorie> anyone know of a real-time kernel that supports our beta targets?
20:45:02 <djdelorie> yeah, it's a post-PA issue, but you asked :-)
20:45:09 <jonmasters_> I'd ask rostedt perhaps
20:45:24 <jonmasters_> either him or Thomas
20:45:43 <djdelorie> I've seen some excitement about ARM-based CNC controllers, Fedora 18 running heavy equipment would be cool but linuxcnc uses RTAI and it's arm support is ancient
20:46:06 <jonmasters_> just promise us you won't blame Fedora if the CNC cuts off something you'll miss
20:46:13 <djdelorie> arm PA might trigger arm support throughout the ecosphere...
20:46:41 <djdelorie> I've been told that there's a lower limit on how many such accidents you should have before you're considered "done" :-)
20:46:58 <pbrobinson> djdelorie: cool but can we discuss that on the #fedora-arm channel please
20:47:03 <jonmasters_> Iff they put Fedora ARM (or any other preempt-rt) in the flight control of a plane I'm flying on, I'm not getting on it.
20:47:16 <jonmasters_> right, any other business?
20:47:30 <jonmasters_> going once...
20:47:38 <jonmasters_> going twice...
20:48:19 <jonmasters_> #endmeeting