/srv/irclogs.linaro.org/2011/10/10/#linaro.txt

=== michaelh1 is now known as michaelh1|away
=== agreen-away is now known as agreen
springzliuyq, 03:00
springzliuyq, hi03:00
springzliuyq, could you help to test branch fix-861115 if you're available? I have pushed a new commit. Android on mx53 still crash with new milestone, so I can't test it myself03:06
liuyqpk03:37
liuyqspringz, ok, i will do that afternoon03:38
=== jk-- is now known as jk-
kan_hu#linaro-multimedia05:04
mabacmwhudson, could you point me to the hwpack that triggered bug 842421 for you?06:36
ubot2Launchpad bug 842421 in linaro-image-tools "linaro-media-create failed to u-boot" [Undecided,Invalid] https://launchpad.net/bugs/84242106:36
=== agreen is now known as agreen-away
=== agreen-away is now known as agreen
=== michaelh1|away is now known as michaelh1
ramanamorning08:07
Doorthe link of libertas firmware for marvell wifi chip is broken08:09
Dooranyone know where i can find them?08:09
hrwDoor: linux-firmware git?08:13
Doorthank you hrw08:13
Doori'll try08:13
michaelh1ramana: hey, could you open the call?  I don't have the password on me.08:33
michaelh1Hey toolchain people.  I'm hunting for the call password.  We'll open soon08:34
zygahi08:34
ams_csmichaelh1: I just sent it to you 08:34
hrwmichaelh1: ok08:34
zygahrw, elections are over, did you get a chance to vote?08:35
davidgilukphew, saved from the bleepy music08:36
hrwzyga: sure, as usual08:36
zygahrw, we had to go to barcelona but we managed to vote too :)08:36
dpcHi. I'm looking for a package (*.deb) that contains "boot.scr" / "boot.txt" file for Versatile Express board in Linaro.08:40
davidgilukdpc: I'm not convinced that there is one in the standard Linaro images08:45
davidgilukdpc: Certainly on this image I have here it doesn't seem to have one in /boot or in the boot partition08:45
dpcThe boot.scr is copied / generated by some script.08:47
dpcdavidgiluk: I need just the content of this file for 11.09.08:47
davidgilukdpc: the flash-kernel program is the one that generates the boot files when a new kernel is produced08:48
zmawow, made a Linaro Ubuntu image for Pandaboard and tried to run it with Qemu, crashes my host linux running Qemu.08:53
czajkowskiAloha08:59
davidgilukzma: Hmm - qemu doesn't have support for pandaboard hardware; however it shouldn't take the host out09:02
davidgilukzma: If you can repeat it, try configuring with --disable-sdl - I've seen some odd interactions between SDL and newer X09:07
zmadavidgiluk: I was trying with versatilepb arch, but I guess it's not good enough. Btw: qemu-system-arm: --disable-sdl: invalid option09:11
davidgilukzma: That's a ./configure option during the build; versatilepb is unlikely to run a panda image09:11
=== doko_ is now known as doko
ryanharkindavepiggott: are you in the Linaro Cambridge office on Wed?09:21
davidgilukpm215: Got a qemu pre to test?09:23
ryanharkindavepigott: are you in the Linaro Cambridge office on Wed? (helps if I spell your name correctly...)09:24
davepigottryanharkin: Yep. 09:24
ryanharkindavepigott: good.  my plans are very vague at the moment, but I'm in ARM in Tues and hope to pick up 2 vexpress boards for LAVA.  I thought I might drop them off on Wed *IF* I have the chance.09:25
davepigottryanharkin: That would be great! Hope to see you Wed then. :)09:26
ryanharkindavepigott: i'll let you know more as I do, but wanted to check you'd be there first, before making plans ;)09:26
davepigottryanharkin: np. Am here all week, with the exception of this afternoon, and I have other ARM visitors on Fri PM.09:27
ryanharkindavepigott: great, it'll be nice to meet you and the others if possible - and to get a tour of LAVA & the office!09:28
davepigottryanharkin: Ah, put aside a couple of days. It's massive. :D09:28
ryanharkindavepigott: lol! it'll be even massiverer soon :O09:29
* pabhishek is away: I'm busy09:30
ericm|ubuntursalveti, we found that ifconfig is missing in oneiric rootfs09:49
jk-ericm|ubuntu: replaced by `ip` perhaps?09:51
ericm|ubuntujk-, let me check09:51
ogra_that would be a very heavy linaro hack, ubuntu still uses and ships ifconfig09:52
ericm|ubuntujk-, no ipconfig as well09:54
jk-ericm|ubuntu: how about just `ip` ?09:55
ericm|ubuntujk-, interestingly oneiric rootfs was shrinked very much on snapshots.linaro.org09:55
ericm|ubuntujk-, ah I see your point09:55
* jk- is just making wild guesses :)09:55
* michaelh1 waves goodnight09:56
=== michaelh1 is now known as michaelh1|away
ericm|ubuntujk-, ip is there09:57
ericm|ubuntujk-, oneiric is using 'ip' solely for network configuration???09:58
* ericm|ubuntu goes a bit mad09:58
jk-you should be able to do everything with `ip`, but as ogra_ says, there is still stuff in Ubuntu that assumes ifconfig09:58
ogra_seems you are missing nmet-tools from your install09:58
ogra_*net09:58
* mru always uses 'ip' for everything09:59
ericm|ubuntuogra_, the whole linaro rootfs is packed as a tarball09:59
* ogra_ isnt 100% sure, but i think NM uses ifconfig 09:59
mruifconfig can't even configure an interface with multiple IP addresses10:00
ogra_ogra@horus:~$ apt-cache show net-tools|grep -i task10:00
ogra_Task: minimal10:00
ogra_well, missing that package is pretty serious ... seems you dont have the minimal task installed10:00
* ogra_ wonders if linaro nowadays uses their nano images to build on top, that would explain minimal missing10:01
ericm|ubuntuogra_, that could be10:01
dpcI'm trying to run the Linaro 11.09 on Qemu VX with a root FS mounted via NFS. I'm not getting any prompt and there's no messages from userspace. Adding --verbose to kernel CMDLINE sugests that upstart is starting everything quite OK, and usplash is failing.  How do I fix this?10:40
tixydavepigott: do you have a Versatile Express in the Linaro offices that could be used in Lava?10:43
davepigotttixy: I spoke to ryanharkin earlier about him bringing something in on Wednesday possibly. However, I have two v expresses connected at the moment.10:48
davidgilukdavepigott, Are they both A9 vexpress?10:49
tixydavepigott: I'm speaking with ryan at moment, what coretiles do you have? A9x4 ?10:49
davepigottdavidgiluk: tixy: pass. Let me see if I can find out10:50
* davidgiluk noticed on ARMs website that vexpress is now available with an A5 tile and A15 soft macros, and it might be interesting to have one of each around for perf stuff10:51
tixydavepigott: Can you check if they're 'real' CPUs or slow FPGA prototypes10:53
davepigotttixy: If I'm at a command prompt, how can I find all this out? I have a feeling (but may be wrong) that they're FPGA prototypes10:54
tixydavidgiluk: we've got A5 coretiles of ARM (and will get A15 stuff when it's available)10:54
davidgiluktixy: Ooooh - how do I run stuff on the A5 ?10:55
davepigottdmart: Do you know what the configuration of the two v-expresses in the val lab are?10:55
tixydavidgiluk: When we've produced a working A5 kernel tree10:55
davidgiluktixy: Ah ok :-)10:56
dmartdavepigott: how do you mean?11:08
davepigottdmart: Are they FPGA prototypes, and are they A9?11:09
dmartdavepigott: real A9x411:09
davepigotttixy: ^11:09
dmart400MHz11:09
dmart1G RAM11:09
davepigottdmart: Thanks. :)11:10
tixydavepigott: dmart: Thanks11:10
davidgiluktixy: Hmm actually I'm curious, what changes need to be made for A5 to the kernel - I thought the idea on ARMv7 was that the MMU stuff was getting more stable 11:14
mrumuch of the cp15 stuff is architectural in v7 but not all of it11:17
davidgilukah ok11:17
dmartI think that the impact for A5 is pretty minimal -- however, the vexpress memory map changed for all new core tiles after the original A9x4 tile -- therefore some board support changes are needed to support the A5 core tile11:20
mrua5 is based on a9 iiuc11:20
mruso much of the specifics are probably the same11:21
tixydavidgiluk: what dmart and mru said :-)11:21
davidgilukah ok11:22
pm215A5> are we (ie linaro) doing anything interesting with it?11:25
mruI don't see how we could be doing anything specific to it other than kernel support stuff11:26
mruit's just another armv711:26
pm215we might, eg, have some particular A5 board in mind. or qemu support :-). Just wondering what the context was.11:26
mruthere might be interesting boards with an a5 of course11:27
mruthe it wouldn't be the a5 that made them interesting11:27
mrunot to me, at least11:27
pm215it's not quite the same as an A9 (eg it has VFPv4, not VFPv3)11:27
mruyes, but the difference isn't spectacular11:28
hrwpm215: but A5 was meant as cheap armv7 (kind of arm926 successor) it will be rather slower then A911:28
ryanharkinpm215: the ARM LT will be supporting VExpress to make it useable for the community, but that isn't A5 specific, it's the same job for A9, A5 and A15 when it arrives. As others said, A5 is very similar to A9, with a few small differences in memory map and BSP (timers are different for example).11:28
mrua5 is an in-order mpcore11:28
davidgilukpm215: My only interest is seeing if the string routine performance graph shapes/relative performances are the same11:28
mrua5 is primarily a low-power core11:29
pm215very similar> I know this, this is why I was wondering whether we were doing anything with it beyond saying 'it's like an A9 so we don't need to worry much about it'. Thanks for the ARM LT info.11:30
mruthe toolchain people might want to support a5 instruction scheduling11:30
dmartAFAIK, the A5 patches are already headed upstream.  It's pretty much that a new CPU type declaration and hwcaps definitions for the new hwcaps definitions for VFPv411:32
ryanharkinpm125: I don't expect there will be any A5 specific work happening from ARM LT side, mostly making sure it works in hardware packs, and so on, making sure our trees supports the whole infrastructure for A5, including DS-5 and and UEFI.11:32
dmarts/pretty much that/pretty much just/11:32
dmartthat was pretty gerbled -- hopefully you can guess what I mean11:32
pm215presumably we can have one kernel that supports both A9-vexpress and A5-vexpress?11:33
davidgilukI think from toolchain point of view we probably at least want to make sure that any optimisations we do for other A's don't make the A5 crawl11:33
dmartramana, mru: do you have an opinion on what the __linux__ define means?  I think in practice it means "being build for a linux-hosted userspace target environment" -- in which case there is no correct way to make use of this macro when building kernel code11:34
mrudmart: agree11:34
mruthat's how gcc defines it iiuc11:34
dmartpm215: in principle, yes.  The DT will describe the coretile, and the board code should also be able to autodetect it, once the memory map is known.  For now though, we have the problem that those core tiles use different memory maps.  This means that static I/O mappings etc. need to get cleaned up before both boards can work in a single kernel -- I believe Pawel Moll is currently working on sorting that mess out.11:36
dmartmru: That was my understanding -- just want to get tools guys to comment :)11:38
ramanadmart: In theory that's what it means .11:38
davidgilukdmart: I agree that's what it means, but I've never heard of anyone building a linux kernel on linux with anything other than a user space compiler11:39
dmartThe android team have been trying to build with a bare-metal compiler (i.e. arm-elf-eabi or something), which is causing is to hit this problem11:42
ramanadavidgiluk: A5 shouldn't be that badly supported already. There is some tuning already in there. 11:42
davidgilukdmart: Hmm oh well, that's the 1st time I've heard of someone doing that; why are they trying to do that?11:43
ibirisrsalveti: https://bugs.launchpad.net/linaro-ubuntu/+bug/816638 seems to be still an issue can you confirm?11:43
ubot2Launchpad bug 816638 in linaro-ubuntu "Pulseaudio consumes 100% of the cpu when trying to play a sound with natty's linaro LEB and 3.0.0-1402-linaro-lt-omap " [High,Fix released]11:43
dmartramana, mru: For that reason, I'm not too happy with the idea of putting -D__linux__ in KBUILD_CPPFLAGS11:44
=== plars is now known as plars-holiday
dmartThis will work in the vast majority of cases, but so does the existing setup -- IMHO, we make things different, but not definitely better11:44
ramanamru: try mcpu=cortex-a5 with a newish gcc-linaro-4.6 11:45
ramanathere is still some work that can be done around neon.11:47
mruI've heard a5 neon schedules quite differently from others11:48
davidgilukdmart: My worry would be globally changing all uses of  __linux__ is bound to break something, and unless there is a good reason to use that bare metal compiler, then what's the point?11:48
mrubut then again, any worthwhile neon code is hand-written anyway11:48
mrudavidgiluk: using __linux__ in the kernel is plain wrong11:49
berodavepigott: The good reason to use the bare metal compiler is that -- on the Android side of things -- the "real" compiler doesn't work yet (we're hoping to fix this for 10.10) and that we'd like to be able to compile things with the upstream toolchain (which is also bare metal)11:52
berodavidgiluk: ^^^ (dav[TAB] didn't cut it ;) Too many dav*s)11:53
davepigottdavidgiluk: ^11:53
davidgilukbero: Ah OK that's reasonable - personally I think it's better to fix the compiler than change what's been done in Linux for a long long time (even if it's technically wrong)11:53
mruthere is nothing wrong with the compiler11:54
mrufixes should go where the error is11:54
mruand here the error is in the kernel code11:54
davidgilukmru: I meant fix their user space compiler that they know is broken, not frig anything11:54
mruthat's unrelated11:54
berodavidgiluk: The problem is that we can't fix the upstream compiler11:54
davidgilukbero: Ah11:55
ramanadavidgiluk: A baremetal toolchain shouldn't claim that it is linux hosted :)11:55
ramanait's user error. 11:55
davidgilukramana, That's not what I was suggesting11:55
berodavidgiluk: And it would be wrong to patch the bare metal compiler to define __linux__ as well, given it could also be used to build BSD or whatever11:55
davidgilukbero: Agreed11:55
davidgilukramana: I was suggesting fixing their user space compiler, not frigging their bare metal one - but as bero says they're stuck because it's upstream11:56
ramanadavidgiluk: oh that's because they aren't using -android-linuxgnueabi11:56
ramanaI think. 11:56
beroramana: True11:57
beroramana: Except it should be -linux-androideabi11:57
ramanabero: yeah I can never get that order right. Don't speak enough android. -linux-gnueabi is good enough for my work. 11:59
beroThe order kind of makes sense, given the -gnu is, more than anything else, a libc indicator (there's also linux-uclibc etc), and Android has its own libc -- but it's still somewhat confusing12:00
* pabhishek is back (gone 02:30:58)12:01
dmartdavidgiluk: this isn't a suggestion to change all uses of __linux__, but just the ones in the kernel.  It's actually a relatively small set of cases.12:03
mrumost of them with the else side beingd decidely bsd12:05
mruthe code in question can currently only build if __linux__ is defined12:05
mruchanging that to __linux_kernel__ and always defining that can't possibly break anything12:06
beromru: Sounds like a good plan to me12:08
mrubero: there are only a handful of such places in the kernel12:09
mruok, maybe two handfuls12:09
mrumost of them in the aic7xxx scsi drive12:09
mrur12:09
dpcAnyone knows if Linaro VX build should be working just fine with Qemu?12:16
davidgilukdpc, If by VX you mean versatile express, then yes I run it in Qemu12:17
davidgilukdpc: There was a bug in the qemu git as of about a week ago that meant it didn't start up, but normally it does (not sure that fix has gone in yet)12:19
* mru hopes nobody is relying on qemu for testing anything12:19
* mru remembers what happened when codesourcery did that12:19
ericm|ubuntufabo, npitre - what is the expected plan of a linux-linaro-3.1?12:20
hrwI love lack of unaligned access support in qemu12:20
dpcdavidgiluk: I'm running linaro 11.09 kernel + nano FS over NFS and I don't get any userspace outputs.12:22
dpcIt was the same in 11.08.12:22
davidgilukdpc: My experience of setting up FS over NFS is that it can be tricky - things like mountall etc tend to get confused; I mostly use an SD image; if you add init=/bin/sh to the arguments do you get a prompt?12:23
dpcLet me try this.12:26
tgall_foorsalveti, morning ...  looks like both of the panda lt hwpacks didn't get built/propogated to snapshots12:30
pm215didn't start up> yes, Blue Swirl committed the fix for that to upstream git a few days ago12:30
* tgall_foo is just starting to pull images to start the day12:31
pm215[bug never made it into qemu-linaro git]12:31
davidgilukpm215: Yes I found it out the hard way; a few hours of bisecting to find it was a patch that he'd posted a few hours before12:33
pm215does it really take hours to bisect a segfault-at-startup bug?12:35
davidgilukpm215: Didn't seg for me12:35
pm215ah12:36
davidgilukpm215: booted the kernel and just hung before init12:36
pm215I'm not going to pull any new changes from upstream for qemu-linaro 2011.10, incidentally (since we're into release week already)12:36
dpcOK, it works.12:36
pm215I'll just make sure we have the patches we want (mostly the A15 insn support) in12:37
* pabhishek is away: I'm busy12:50
=== cjb``` is now known as cjb
mabacfabo, is this something you can work your magic on? [rt.linaro.org #133] linaro.kanbantool.com:443 is unreachable from git.linaro.org - it seems that there are a lot of holidays going on, so I guess this might not get much attention today.13:01
dzinasac: mumble?13:04
faboericm|ubuntu: mounir should know better than me. I'll be back to you as soon as I get an answer.13:05
ericm|ubuntufabo, ok13:05
fabomabac: nobody at vanguard atm13:06
mabacfabo, ok thanks for trying. it can wait until tomorrow then.13:07
=== agreen is now known as agreen-away
rsalvetitgall_foo: probably because the rt is not yet done13:23
* tgall_foo returns13:24
tgall_foorsalveti, that's a bummer  but such is life I guess13:24
rsalvetiand a bunch of fails...13:25
rsalvetifabo: check offspring13:25
tgall_fooyeah I noticed that as well ...  /me is just starting to look at that13:25
faborsalveti: I know, I pinged IS this morning (my morning)13:26
rsalvetifabo: ok, cool13:27
tgall_foofabo, anything to do on our end ...   /me just started my day,  looks almost like the plug was pulled in the middle of the desktop build....13:30
fabotgall_foo: Ng just said he's going to take a look. we can't do much more than wait.13:30
fabotgall_foo: do you have the backlog? ericm|ubuntu complained earlier about missing ifconfig 13:32
tgall_foojust scrolling through the ubuntu-desktop build ...   the log ends abruptly quite late in the build  13:32
=== agreen-away is now known as agreen
tgall_fooyeah it's pretty much right after all the local hooks have fired and it's doing it's last little bit of clean up before making a tarball13:34
feweHi13:35
feweI am considering buying a new computer to build Android and other distributions faster13:35
feweWhat type of config should I use?13:35
ericm|ubuntutgall_foo, dexter from freescale is using linaro-o-ubuntu-desktop-tar-20110929-0.tar.gz, and even no GUI was up13:35
fewe(i am actually considering buying a Mac Pro, maybe the one with 8 cores)13:36
mrufewe: as many cores and as much ram as you can afford13:36
* mru would not suggest a mac as a build system13:36
ogra_unless you are building debs for PPC :)13:37
ericm|ubuntuogra_, why is that? the new Mac available are all x86 based now13:37
mrufor building stuff really fast, I'd get a dual or quad-socket xeon with 48G ram or so13:37
ogra_ericm|ubuntu, i didnt say which mac ;)13:37
ericm|ubuntuogra_, heh13:38
ogra_(nor did mru btw)13:38
mruif on a budget, a single 8-core with 24G ram13:38
faboericm|ubuntu: older images add a knwon issue (rootfs is 73MB). the fixed image is ~470MB.13:38
fabos/add/had/13:38
davidgilukmru: The Mac Pro's are dual socket xeons typically (note not book)13:38
mrudavidgiluk: do they support 48GB ECC ram?13:39
fewe24 G RAM?13:39
fewe:O13:39
feweisn't that a bit too much?13:39
mrunope13:39
fewewow13:39
davidgilukmru: I think so - they're built for heavy video wranglers13:39
mruI have a quad-core i7 with 12G and I could certainly use more13:39
fewehow much time for a build?13:39
fewecomplete build13:39
mrudavidgiluk: even so, a non-mac is going to be cheaper13:40
feweif you know the number13:40
ericm|ubuntumru, or you want a building farm - connect several cheap PCs with distcc13:40
mruand you wouldn't want to run macos on it anyway13:40
mruericm|ubuntu: I've found distcc to be not entirely reliable13:40
ericm|ubuntufabo, ok - I'll check w/ the latest13:40
ericm|ubuntumru, it's working quite OK actually13:40
mruexcept when it fails13:40
fewemru: ok13:40
ericm|ubuntumru, the problem is really it's not very convenient when adding one more machine13:40
davidgilukmru: They say 64GB RAM max (and 2x6 core)13:40
fewemru: do you have a template configuration that i can check out?13:41
mrufewe: no, sorry13:41
tgall_foodavidgiluk, and ecc if you want it13:42
davidgilukfewe: Of course you can get PCs as big as your money will let you go - 80 core and 2TB of ram if you can stretch to it - I'd love to see what that was like for a make -j :-)13:47
zygawow, what are you building?13:47
zyga64GB ram, 2x6 core?13:47
fewe:D13:47
zygawill it run unity ;-) ?13:47
fewedavidgiluk: what do you have?13:47
davidgilukfewe: An 8 core, 16GB of RAM - but I don't build full distros; it'll build me a cross gcc in about 4 minutes13:48
mruzyga: he mentioned distro building and android13:48
mruthose things eat ram for breakfast13:48
feweDo you have a data sheet to check it out?13:48
mruif you go for intel CPUs, I recommend an all-intel system13:50
mruif there's even anything else to be had these days13:50
* mru had a system with nvidia chipset once, bad idea13:50
davidgilukfewe: I think the one I'm using is the previous model of http://www-03.ibm.com/systems/x/hardware/rack/x3550m3/index.html13:51
feweIs this something you rent, like with amazon, or is something you bought? it's pretty expensive13:52
davidgilukfewe: I work for IBM so well it just appeared :-)13:53
feweok 13:53
fewe:)13:53
fewecan you make one appear in my office as feel please? that would be nice13:54
fewe:D13:54
feweas well*13:54
mruI don't think I'd want one of those at home actually13:54
davidgilukmru: It'll keep you warm in the winter13:54
davidgilukmru: But, no rackmount stuff at home is a bit noisy13:55
mruI have enough machines to stay warm anyway13:55
davidgilukfewe: Some of the cloud stuff is quite good - you can rent fairly chunky machines by the hour13:56
fewedavidgiluk: yes i was also considering this but as a last resort13:57
feweand when i see the prices of the computer that i need13:57
fewei think i'd better go with this solution13:57
rsalvetiericm|ubuntu: that's quite an old build 13:58
rsalvetiin theory today's image should be good for a first try13:59
rsalvetibut offspring is off :-(13:59
rsalvetiibiris: that was fixed on a newer hwpack14:00
davidgilukfewe: Well obviously you can spend a fortune on a computer; but my own personal machine is an i7 - now those aren't too bad and aren't that expensive14:01
rsalvetithe released hwpack is one built at 092914:01
mru1000 $currency should get you a half-decent machine14:01
davidgiluknod14:01
fewedavidgiluk: personally i have the last mac book pro, and it builds Android in 1h - 1h30min14:02
mrubut then you'll have to make do with consumer-grade stuff14:02
jussimru: except if $currency == yen :P14:02
fewedavidgiluk: and i saw on linaro build service that the mean build time is around 40 minutes14:03
mru@currencies = (pound, euro, dollar);14:03
fewedavidgiluk: so if i have to spend 10K$ to have a good build machine which will only reduce the build time by half14:03
fewenot worth it14:04
davidgilukfewe: Probably true depending how often you do the build14:04
fewemy busiest day was 4 times14:04
feweMy average is 2 builds / day14:04
feweplus when i am building i can work on something else14:05
ericm|ubuntursalveti, 'k - will try14:22
davidgilukhttp://www.theinquirer.net/inquirer/news/2115929/ubuntu-1110-support-arm-chips-fight-red-hat14:22
JesseBarkertgall_foo, ping?14:48
tgall_fooJesseBarker, pong!14:49
JesseBarkertgall_foo, doing a little blueprint "weeding".   Can I kill the work items for you on this:14:49
JesseBarkerhttps://blueprints.launchpad.net/linaro-multimedia-project/+spec/multimedia-linaro-validate-instrumented-player14:49
JesseBarkertgall_foo, looking to tie up any smartt related loose ends.14:50
* tgall_foo scratches his and head wonders what that is14:50
tgall_fooyeah close that out14:51
JesseBarkertgall_foo, thanks!  I'll mark it all done.14:51
mrubtw, does anyone in linaro do anything with valgrind?14:52
mruor should I just fix the bugs myself?14:52
tgall_foodavidgiluk, uweigand ^^^^ valgrind ?14:53
JesseBarkerasac_, can you mark https://blueprints.launchpad.net/linaro-multimedia-project/+spec/multimedia-linaro-validate-instrumented-player as "implemented"?14:54
ericm|ubuntujcrigby, I used your debian* directories from linux-linaro-oneiric to build my 3.1 kernel, I'm having the follwing errors:14:55
ericm|ubuntujcrigby, make[2]: Nothing to be done for `/home/ycmiao/linux-linaro/Makefile'.14:55
ericm|ubuntu  HOSTCC  scripts/basic/fixdep14:55
ericm|ubuntu/bin/sh: scripts/basic/fixdep: not found14:55
ericm|ubuntumake[4]: *** [scripts/basic/fixdep] Error 12714:55
ericm|ubuntujcrigby, any hint why this would happen?14:55
asac_JesseBarker: done14:55
rsalvetiericm|ubuntu: did you just copied the directories or applied the sauce that creates it?14:56
JesseBarkerasac_, thanks, man.  Just trying to tidy up my blueprint garden :-)14:56
davidgilukmru: Which bugs have you hit - I've not looked at valgrind myself on ARM, but there is a bug with a whole pile of ARM bugs attached to it14:56
ericm|ubuntursalveti, copied14:56
davidgilukmru: And some patches14:56
mrudavidgiluk: most recently, vcvt to/from fixed-point not implemented14:56
rsalvetiericm|ubuntu: could be that there are also some other patches to fix/improve the build system14:56
uweigandmru, tgall_foo, davidgiluk: I fixed a couple of valgrind bugs on ARM a while ago, I think they should be upstream now ... that was good enough for me to valgrind gdb, so at that point I stopped looking into it14:57
asac_JesseBarker: very good ... do you need general powers on that project?14:57
rsalvetiericm|ubuntu: jcrigby would know better, but it's holiday at us today14:57
ericm|ubuntursalveti, 'k - I'll check w/ him tomorrow then14:57
davidgilukmru: http://bugs.kde.org/show_bug.cgi?id=266035  is the one I know; but doesn't look like any vcvt stuff14:57
ubot2KDE bug 266035 in general "Support running Valgrind for Android on ARM" [Major,Unconfirmed]14:57
mruI've fixed a few bugs myself in the past14:58
JesseBarkerasac_, that might be useful.  When krtaylor and I negotiated the transfer of smartt, I don't think we followed all of the loose ends (the project itself now lives under the gfx project group).14:58
asac_JesseBarker: nevermind... seems i dont have powers either in general :) ... just as approver here14:58
mruadding instructions isn't hard, just tedious14:58
asac_JesseBarker: so i cannot help you anyway ;)14:58
tgall_foorsalveti, actually it's more of an optional holiday today than anything ... in these parts we still have school and everyone I know of is working 14:58
JesseBarkerasac_, well, if I find anything else, I'll bug someone with powers :-)14:58
asac_good14:58
pm215valgrind> IME Julian implements instructions on-demand but is generally happy to do so14:59
rsalvetitgall_foo: oh, ok, interesting 14:59
JesseBarkertgall_foo, rsalveti, no mail, no banks, no union labor.  everyone else is probably working.14:59
rsalvetihaha :-)15:00
=== ericm|ubuntu is now known as ericm-Zzz
pm215somebody on the qemu list is asking what the recommended source for building versatile/realview/vexpress kernels is -- any suggestions? (in particular, is mainline kernel.org ok?)15:10
davidgilukpm215: There are a set on arm.com's site15:12
pm215I know that. But are they the currently recommended sources? :-)15:13
=== zyga is now known as zyga-afk
zyga-afkI'll be back later15:57
hrwhave a nice rest of day16:00
xranbythe reason why linaro oneiric images fail to run Metacity , gnome-terminal and friends http://paste.ubuntu.com/705466/16:02
xranbyogra_: ^16:03
tgall_fooxranby, ah ...  related I seem to recall on friday there was some discussion here about gconf being highly resource intensive16:04
davidgilukwth did it get that path from16:04
xranbydavidgiluk: not sure yet16:04
xranbyim no gconf expert16:04
davidgilukme neither16:05
tgall_foodavidgiluk, that very much looks like an evil hardcode got into the packge16:05
tgall_foosmells of multilib16:05
pm215also a classic "Failed ...: Success" error message :-)16:07
davidgiluktgall_foo, But wouldn't that break on x64 boxes?16:08
tgall_foodavidgiluk, not with multilib installed16:08
davidgiluktgall_foo, Sorry, wouldn't that break onx 64boxes that hadn't installed a whole ton of 32bit stuff?16:08
davidgiluktgall_foo, I don't think Oneiric installs a lot of 32bit stuff by default on 64bit hosts16:09
ogra_xranby, hmm, the new gnome-terminal (as the rest of gnome3) uses gsettings/dconf ... it shouldnt even bother to look for any gconf bits16:11
mrugconf, making the windows registry look good16:12
tgall_fooxranby, this linaro build you have,   any idea what the date on it is ?16:12
xranbytgall_foo: i am using exactly sudo linaro-media-create --rootfs ext4 --mmc /dev/sdc --binary linaro-o-ubuntu-desktop-tar-20111008-0.tar.gz --hwpack ../Skrivbord/hwpack_linaro-lt-mx5_20110929-0_armel_supported.tar.gz --dev mx53loco16:14
xranbyso the rootfs from 100816:14
xranbyin combination with last release's hwpack for imx5316:15
tgall_foohmm  yeah that should be exactly the same as ubuntu's desktop image ogra_  .. just fyi, we use your meta pkg16:15
ogra_we dont :P16:15
ogra_(ubuntu uses tasks )16:16
tgall_fooogra_, o?  maybe I misheard but I thought at budapest it was decided you were going to16:16
ogra_we stopped using metapackages in ubuntu in 2006 iirc16:17
ogra_probably 200716:17
tgall_foowell you'all should quit updating all your seed packages then ;-)16:17
tgall_fooogra_, for the purposes of live-build, where are you keeping your task lists ?16:21
ogra_tgall_foo, in the seeds ... germinate generates tasks from the seeds which live-build installs then16:25
ogra_alongside (and inside the task) there is the metapackage that is needed for system upgrades and user convenience16:25
tgall_fooah ...yeah, very patato. patato then  ... in the grand scheme the effect will be the same then16:29
ogra_i dont think it will, i'm sadly not the best person to explain tasks vs metapackages though but the results differ quite a bit 16:30
xranbyhttp://paste.ubuntu.com/705479/  the  /usr/share/dbus-1/services/org.gnome.GConf.service   file contains  Exec=/usr/lib/i386-linux-gnu/libgconf2-4/gconfd-216:31
xranbyi also noticed that xfce4 (that i installed manually) contains a lot of similar bugs16:31
tgall_fooogra_, no they don't, if they did then live-build would be broken16:32
ogra_?? 16:32
ogra_live-build just does what you tell it 16:32
xranbydavidgiluk: thats the file that contains the wrong info ^16:33
tgall_fooyes but,   the output of germinate is the input to a meta package ... ergo,   using germinate to generate your task-list   vs using a meta package created with germinate will create the same effective in live-build16:34
ogra_tgall_foo, it doesnt care wether you install ubuntu-desktop^ or ubuntu-desktop 16:34
tgall_foowell it does for the purposes of it'd best exist in the archives you've told it about of you'll error out16:35
ogra_the first one desont need to exist as a package16:35
ogra_its a task16:35
tgall_fooand ubuntu-desktop as a meta pkg does exist because you guys are creating it, and it's a direct reflection of the same seed you're using to gen your task.  So least for my evil and nefarious purposes I'm happy :-)16:36
xranbyi have filed a bugreport for you https://bugs.launchpad.net/linaro/+bug/87189216:46
ubot2Launchpad bug 871892 in linaro "org.gnome.GConf.service contains i386 pathname" [Undecided,New]16:46
tgall_foothanks xranby 16:52
=== agreen is now known as agreen-away
ogra_tgall_foo, sorry, was afk for a moment, indeed, if you dont care that your package system on the install handles the installed packages differently from ubuntu, then just use the metapackage, task packages are marked differently in the package database (iirc its automatically vs manually installed or some such, as i said above, i'm probably no good person to explain the difference)17:18
davidgilukogra_, Can you give an example of a task package?17:19
ogra_davidgiluk, apt-cache show gnome-terminal|grep ^Task17:19
ogra_there are no task "packages" :) its a control field in the binary deb17:20
davidgilukah17:20
ogra_tom is right though that the contents of the ubuntu-desktop metapackage and the task are the same though17:21
ogra_but the handling differs17:21
davidgilukso how do you install a task erm not-package ?17:22
davidgilukis that tasksel ?17:22
ogra_you add a caret ... 17:22
ogra_apt-get install ubuntu-desktop^17:22
ogra_or use tasksel ... tasksel install ubuntu-desktop 17:23
davidgilukhmm ok; well that's a nice on to know17:23
ogra_for proper details of someone who really knows that stuff, you could squeeze that info out of slangasek or cjwatson in orlando at the cost of a beer i bet ;)17:24
ogra_s/of/from/17:24
davidgilukah, I'm not going to Orlando17:25
ogra_:(17:28
=== zyga-afk is now known as zyga
=== cjb`` is now known as cjb
Dzap2hi, is anyone from the development team here?19:26
tgall_fooDzap2, best to just ask your question ... there's lot of teams represented here and we all do development19:28
Dzap2ah ok19:28
tgall_foorsalveti, success ...   have things building with the updated live-build3-a3419:29
Dzap2so a little background on what my senior design project is: the main goal of my project is to build a humanoid robot platform, the customer specified a few things for the electronics side of this project, which was an OS, has expandability for future projects, and off the shelf components if possible.19:31
Dzap2the reason im here is because ive been looking at the st ericsson snowball for the PC brain, and i kno developers dont like to be asked when features are going to be supported, but WiFi is something needed for my project, and i noticed that the linux and android releases for the snowball does not have complete functional support yet...19:33
Dzap2the snowball has sensors that most other embedded pcs or real pc motherboards dont have, such as the gyro and accelerometer, which would be very useful to my project if those are all integrated on the board and i dont have to buy a separate board with those sensors. what worries me is the incomplete OS support19:36
Dzap2any suggestions? should i look at something else for my project?19:37
tgall_fooDzap2, be patient for an answer, some of those here in channel that have a snowball might already be gone for the day (IE over in europe)  ...  you might consider reposting here earlier in the day based on the average european schedule19:39
Dzap2ah, i see, i can try again tomorrow morning19:40
=== michaelh1|away is now known as michaelh1
michaelh1Dzap2: hey.  Not sure about your bot, but you're likely to need much more than the Snowball has on it.19:42
Dzap2well, yea, the EEs figure 3 IR sensors and 2 touch sensors19:43
michaelh1Personally I'd go for a well established board like the IGEPv2 and add one of the many control boards you can buy to it.19:43
Dzap2which will be interfaced via an arduino and the arduino to the PC19:43
michaelh1That way you don't have to worry about the level of board support.19:43
Dzap2my team were thinking of a "PC" board (anything that can support an OS), one arduino, and one servo controller.19:44
Dzap2ive never heard of the IGEPv219:44
michaelh1OK.  So the Arduino uses serial for comms.  The IGEP has a 3.3V serial port out on headers.  That would work well with the new Cortex-M3 based Arduino but will need a level shifter for the AVR based ones.19:44
=== ohad__ is now known as ohad
michaelh1The IGEP is a OMAP3 like the beagleboard but (IMO) a better shape and layout.19:45
Dzap2ah i see19:45
Dzap2are there any high performance boards that are well established?19:45
michaelh1Have you looked at the Gumstix range?  I thought they had a few good I/O boards.19:45
Dzap2the robot platform is so that a future design team can take our existing components and start adding to it19:45
michaelh1The OMAP3 is pretty fast.  A dual core 1 GHz A9 is ~2.5 times faster.  How much speed do you need?19:46
Dzap2one thing already mentioned for the future is a camera on the robots head and beaming video live back to a remote computer via wifi19:46
Dzap2and after the camera project, the customer said stereoscopic camera vision >_<19:46
michaelh1The cameras that I've seen are normally USB or Ethernet based.  Stereoscopic isn't really a problem.19:47
Dzap2im on the platform project, future teams will do additions, so i have to provide a base where the future teams wont have to reselect the PC or servo controller, for example19:47
michaelh1Nice thing about USB is that the stream is already compressed so, providing there's no image processing, the CPU load is low as you forward the stream on.19:47
mrunot all usb cameras compress the image19:48
michaelh1mru: video cameras?19:48
mruyes19:48
mruobviously low-res19:48
michaelh1Yeah.  A back of the envelope is VGA 30 fps RGB is 26 MB/s which is hard for USB.19:50
Dzap2right, we figure VGA is a good guess of what our customer is going to want in the future19:50
michaelh1Dzap2: re: a A9 with good BSP, sorry, I don't know.  The Panda was first and furthest along but is harder from a I/O point of view.19:50
mruthe uncompressed cams are usually yuv or raw bayer19:51
mruwhich is 8-12 bits per pixel19:51
michaelh1Ah, of course.19:51
mruso vga 30fps is just about possible19:52
mwhudsonmabac: i was trying to reproduce the sigill bug, so it's in that report19:52
mruif your isochronous usb works out19:52
mwhudsonmorning all19:52
michaelh1Heh,19:52
* mru remembers dealing with a nasty host controller that didn't support isochronous properly19:53
Dzap2it doesnt have to be a usb camera, arent there expansion ports on the beagleboard/snowball/etc?19:53
mruthe beagle-xm and panda boards have a special camera input19:53
michaelh1Dzap2: how about https://www.gumstix.com/store/index.php?cPath=40 such as https://www.gumstix.com/store/product_info.php?products_id=254 or https://www.gumstix.com/store/product_info.php?products_id=26219:53
Dzap2in any case. michaelh1 said it best, im looking for a well established board with enough performace for the future19:53
michaelh1The cameras linked hook into the OMAP3 camera interface.  Never used it myself.  Won't give you stereoscopic.19:55
Dzap2funny u mention the gumstix, michaelh1, a team one semester before us taped a gumstix and the camera to a rc helicopter and programmed it to recognize colors and send images back via wifi, and they told me their gumstix burned up19:55
mrugumstix should be fine19:56
michaelh1Interesting.  These are low power (~2 W), but if you don't let the heat out then they will fail.19:56
Dzap2but it is an option in lieu of no other options19:56
mruI've never heard of gumstix boards having any particular heat issues19:56
Dzap2i believe they said the wifi module on top of the processor board did the most damage as it got really hot, even though the helicopter's rotor blades shoudve been cooling it19:57
michaelh1I mounted my first beagleboards in a plastic container with a few heat holes.  It wasnt enough as the plastic wouldn't conduct the heat and the airflow was too low.19:57
Dzap2ive seen their burned out gumstix, so, my teams been wary of the gumstix19:58
Dzap2shit, i got class. thanks to you all for discussing this with me, ill be back on later!19:59
mwhudsonmabac: https://bugs.launchpad.net/bugs/85947319:59
ubot2Launchpad bug 859473 in linaro-ci "CI kernels causing many "Illegal Instruction"s" [High,Confirmed]19:59
michaelh1Dzap2: I'd buy a Snowball and have a play.  The BSP will come along.  It brings much of the I/O out to headers and is small, which is nice.19:59
michaelh1(oh, lost him)19:59
czajkowskialoha 20:31
czajkowski.c20:31
mwhudsonzyga: around?20:37
=== jboyer2002 is now known as jboyer
zygamwhudson, briefly21:39
zygamwhudson, what's up21:39
mwhudsonzyga: i forgot already :-)21:39
zygamwhudson, I have two tiny branches, could you have a look (server, dashboard)21:39
zygaI'm falling asleep now so no need to rush21:40
mwhudsonzyga: stuff about projects a bit i guess21:40
zygayes21:40
mwhudsonzyga: i worry that the user experience is going to get worse before it gets better :/21:44
zygamwhudson, ?21:45
mwhudsonzyga: i guess we should spend a lot of time talking dashboard/vlo UX at the connect21:45
zygamwhudson, this is about failed upgrades21:45
zygamwhudson, not about new UI :-)21:45
mwhudsonzyga: no, i wanted to talk to you about projects21:45
zygamwhudson, sure21:45
zygamwhudson, I'd love that21:46
mwhudsonzyga: ah yeah, i saw the mps about the upgrade situation21:46
mwhudsonzyga: do you know how i can reproduce it in a vm?21:46
zygamwhudson, I tried21:46
zygamwhudson, I think that not having seatbelt installed initially is enough21:46
mwhudsonzyga: and then upgrading from lava-server 0.4 to 0.5 ?21:48
zygamwhudson, correct21:48
zygamwhudson, what I cannot understand though21:48
zygamwhudson, is how one can do that21:48
zygamwhudson, as seatbelt is required by django-debian-webserver-apache221:48
zygamwhudson, I believe that it might be an early configuration failure if you install the right set of packages on an otherwise pristine system21:49
mwhudsonzyga: maybe django-debian-webserver-apache2 isn't installed?21:50
zygamwhudson, I added p-d-seatbelt to pre-depends 21:50
mwhudsonnothing depends on it in my vm21:50
zygamwhudson, it was according to the reporter21:50
mwhudsonah ok21:50
zygamwhudson, hmm?21:50
zygamwhudson, really?21:50
* zyga scratches head21:50
zygamwhudson, how about lava-server?21:50
mwhudsonit depends on django-debian21:51
mwhudsonoh ah21:51
mwhudsonand django-debian-webserver21:52
mwhudsonwhich is satisfied by django-debian-webserver-apache2 of course21:52
* mwhudson doesn't know how this end of debian works very well21:52
zygamwhudson, add to the confusion that installed != available as you need to configure python packages first, sigh21:53
mwhudsonah right21:55
mwhudsonwhen did the dependency on django-seatbelt appear?21:55
zygawhich one21:55
zygathe one on lava-server is new, from the merge request21:55
zygathe one in django-debian-webserver-apache2 was there for a long time21:55
zyga(as soon as I created seatbelt to reduce support questions on IRC)21:56
mwhudsonzyga: the latter21:56
mwhudsonhm ok21:56
zygait was around for 0.4 for sure21:56
mwhudsonah of course the db on the vm is way newer than the code21:58
mwhudsonso it has all the migrations applied21:58
zygaright21:58
zygahard to work like that21:59
zygaI'd say that a fresh deploy is more likely to show this issue21:59
mwhudsondo you remember when you upgraded lava-server on vlo? :-)21:59
zygahmm, no but why is that relevant?21:59
zygawe rarely wait for ppa to build so I always copy the full set of new packages21:59
mwhudsoni could grab an older db snapshot21:59
zygathis did not happen on v.l.o21:59
zygaah21:59
zygahmm21:59
zygaprobably +2 months ago 22:00
zygabut I'm not sure22:00
* mwhudson gives up on that plan then22:00
zygamwhudson, I'll do a deploy test then22:01
zygamwhudson, without dpkg related changes22:01
mwhudsonok22:01
zygamwhudson, btw, did you see dispatcher docs?22:01
zygamwhudson, more and more of lava gets rtfd.org domain :)22:01
mwhudsonzyga: i clicked the link, not much more than that though22:01
zygamwhudson, I resumed my work on bundle docs, I'll finish this tomorrow and submit it for review22:04
zygamwhudson, plars asked me to rework the front page22:04
zygamwhudson, I added some new API to make that saner (no more context processors)22:05
zygamwhudson, I'll try to include his bootchart report, the existing boot status report and some new scheduler/board activity report22:05
zygamwhudson, is there anything specific to the UI we have now that you'd like to discuss?22:06
mwhudsonzyga: not really, i guess i'd like to start with thinking about who our users are and what they want22:06
mwhudsonzyga: have we really done this properly?  i like doing in with pen & paper which is a bit tricky to wiki22:07
zygamwhudson, I don't think we did22:07
zygamwhudson, but one did happen22:07
zygamwhudson, our focus shifted from unspecified users to linaro users22:07
mwhudsonzyga: ah right22:08
mwhudsonzyga: let's make super sure we talk about this at the connect (hi plars-holiday)22:09
zygamwhudson, agreed22:09
zygamwhudson, my fav list is 1) development process/released/deployment 2) task focus/priority 3) general UI rework and refocus on projects22:10
mwhudsonzyga: can you expand on 2?22:10
mwhudsonzyga: and i'd insert what i've been talking about as a prerequisite for 3) ;-)22:11
zygaI keep having issues with the way we work, now even more than ever, with lots of blueprints that have poor tracking features, imprecise priorities, it is getting hard for me to know that I'm working on important stuff, I feel we lost some of the fluidness of the process22:11
mwhudsonah ok22:12
mwhudsonyeah22:12
zygain the past I used to work what I knew was needed and it gave fruit, I was focused and the code was fine, recently there are a lot of ideas and things and the important things evades me22:13
mwhudsoni know i'm being a bit of a stuck record on this, but i think clarifying what we're trying to achieve will help 22:13
zygasure, and who the stakeholders exactly are22:13
mwhudsonthat too22:13
zygaon the other hand I sometimes feel that asking our users what they want is not the optimal approach22:13
mwhudsonah no, i'm not saying we should do that22:14
zygabut that's better than nothing I guess22:14
mwhudsonwell, not just that22:14
mwhudsonwe need to understand what they need22:14
mwhudsonwhich is different from asking them what they want (hat tip to the recently departed here)22:14
zygayeah22:15
zygathat'a art though :)22:15
mwhudsonit's true22:17
mwhudsonlucky our users are not -so- different from us22:18
zygamwhudson, offtopic, did you see http://www.nuitka.net/ ?22:18
mwhudsonno22:18
zygamwhudson, look at it now :)22:18
zygamwhudson, python compiler22:19
zygamwhudson, compiles all python to C++22:19
zygamwhudson, http://www.nuitka.net/blog/nuitka-a-python-compiler/ I guess this link is better22:19
mwhudsonah yeah22:20
mwhudsonzyga: i still think you need something jit like to really do anything interesting with python performance22:22
=== michaelh1 is now known as michaelh1|away
ojnHm, is anyone else seeing a looong delay when booting linux-next? I don't have my panda available right now to compare with, but tegra takes 30+ seconds for first console messages (delay seems to be from before earlyconsole is up, and before time starts ticking)22:23
=== michaelh1|away is now known as michaelh1
zygamwhudson, pypy thread on tech is interesting too :)22:27
mwhudsonyep22:27
zygamwhudson, I think that guided ahead-of-time might be better for certain workloads22:28
zygamwhudson, as it is more predictable22:28
zygamwhudson, anyway :)22:28
zygainteresting project to look at from time to time22:28
mwhudsonmichaelh1: hi, completely off-linaro-topic, but do you know stuff about the gcc vectoriser?23:05
michaelh1Some.  What would you like to know?23:05
mwhudsonmichaelh1: friends in the pypy project just ran into http://gcc.gnu.org/bugzilla/show_bug.cgi?id=5069323:05
ubot2gcc.gnu.org bug 50693 in tree-optimization "Slightly different loop body leads not vectoring loop" [Normal,Unconfirmed: ]23:05
mwhudsonmichaelh1: wondered if it rang any "oh, i'd expect that" or "oh, that's really surprising" type bells23:06
michaelh1mwhudson: my first response is that the slower loop is evil :)23:09
mwhudsonwell yeah23:09
mwhudsoncode generation is fun23:09
michaelh1mwhudson: but past that, it's using gotos instead of a loop.  The GCC vectoriser is primarialy a loop vectoriser.  GCC is quite clever and can recognise some things that are loops, even if they're not written as one.23:10
michaelh1mwhudson: yeah, I've caused similar code to be generated as well.23:10
mwhudsonah so in this case, it's probably not being recognized as a loop?23:10
michaelh1probably.  The vectoriser can also vectorise straight line code, but there's not enough in a straight line here to do anything.23:12
michaelh1Ask irar when she's about.23:12
michaelh1mwhudson: pypy is cool BTW.23:13
* michaelh1 lunches23:13
mwhudsonmichaelh1: yeah, i wish i still had time to really keep up with it23:13

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