=== zumbi is now known as Guest67303
=== mru_ is now known as mru
=== jk-- is now known as jk-
=== zumbi is now known as Guest16855
=== michaelh1|away is now known as michaelh1
=== raster is now known as Guest40083
=== fgiff_ is now known as fgiff
amitdkvingu, hi07:55
vinguhi amitdk07:56
amitdkvingu, I am looking into sched_mc configuration flags.. for sched_mc=2 my flags = 559, where to find the meaning of this07:57
vinguamitdk: flags bit description are available in include/linux/sched.h07:59
amitdkvingu, thankx .. i will check07:59
=== irar__ is now known as irar
=== Guest97181 is now known as mabac
hrwok, toolchain people: how to force gcc to check one directory for target libs? I gave up.08:34
=== doko_ is now known as doko
hrwI want cross gcc-linaro to check /usr/arm-linux-gnueabi/lib/ by default but it always goes to {prefix}/arm-linux-gnueabi/lib/ 08:35
davidgilukslangasek: Hi08:40
=== chrisccoulson_ is now known as chrisccoulson
mkedwardskiko: fixing tinderbox's UX is not on my short list of things to do in life <g>09:08
=== benjiG1 is now known as benjiG
tusharbrsalveti: hi09:27
tusharbrsalveti: Do you have time for a question?09:28
=== jk- is now known as jkmeep
=== jkmeep is now known as jk-
olseni'm trying to get a linaor image on my overo board following the instructions here:http://wiki.gumstix.org/index.php?title=Linaro09:33
olsenrunning the command  linaro-media-create command with the 20110920 build i get : AttributeError: 'NoneType' object has no attribute 'group'09:35
olsenthis is the exact command i'm using:  sudo linaro-media-create --rootfs ext3 --mmc /dev/sdb --binary linaro-n-ubuntu-desktop-tar-20110920-0.tar.gz --hwpack hwpack_linaro-overo_20110920-0_armel_supported.tar.gz --dev overo09:39
ramanamorning. 09:45
hrwhi ramana 09:47
ramanahi hrw09:49
=== davepigott_ is now known as davepigott
=== Guest16855 is now known as zumbi
=== pm215__ is now known as pm215
Cosmo`a brief question ...10:22
Cosmo`what's the actual difference between the mainline kernel and the linaro version?10:22
=== michaelh1 is now known as michaelh1|away
faboCosmo`: latest release is 3.0.4 release + 2 patches10:49
Cosmo`so these are patches that aren't in the mainline that someone from the linaro group has come up with10:50
Cosmo`to better support arm platforms10:50
faboCosmo`: the easiest way is to go through the git repo10:50
Cosmo`yes i'm doing this now, i've not used git before so it's taking a while to work it out10:50
Cosmo`i *think* i've got what i want now10:50
Cosmo`which is linaro kernel
faboCosmo`: it could be patches cherry-picked outside of Linaro10:50
Cosmo`so the difference is just that there's patches in the linaro kernels that aren't in the mainline kernels10:51
Cosmo`but may well eventually make it upstream into the mainline kernel10:51
faboCosmo`: that aren't in the mainline kernels *yet*. it's more precise, as everything we do is upstreamed.10:52
Cosmo`everything eh? i see10:53
faboCosmo`: git diff v3.0.4 linux-linaro-3.0-2011.09-0 should tdo the trick10:53
Cosmo`ah, i didn't know git could do that10:54
Cosmo`took me long enough to work out how to pull down a tag ...10:54
faboCosmo`: you can be interested by reading http://www.linaro.org/linaro-blog/2011/09/09/the-evolution-of-the-linaro-kernel/10:54
Cosmo`that is very interesting, thank you10:55
faboyou're welcome10:55
Cosmo`i'm mainly just trying to get the correct version of the kernel that the arm-linux guys have used their patch with10:55
Cosmo`which, as far as i can tell, is linaro
amitkvingu: I'm trying to understand this sentence on the sched_mc wiki: Finally, a load balance will be tried between the busiest group and the cpu if the average load of the local group is below the domain one and if there is an obvious imbalance (more than imbalance_pct %) between these 2 groups.11:25
amitkvingu: can you rephrase?11:25
=== ggrandou_ is now known as ggrandou
rsalvetitusharb: sure12:00
dzinrsalveti: feeling better?12:01
tusharbrsalveti: I was trying to build a hwpack with 3.1-rc upstream kernel ...12:01
tusharbrsalveti: I am using the origen - 3.0 as reference ...12:02
tusharbrsalveti: it boots up fine, but stops while looking for the /dev/disk/by-uuid/xxx ...12:02
rsalvetidzin: yup, thanks :-)12:02
tusharbrsalveti: I could not find any listing under /dev/disk/by-uuid too ...12:02
tusharbrsalveti: Do you know of any kernel config option which enables this listing?12:03
tusharbrsalveti: I am able to see /dev/mmcblk0p1,2,3 etc. 12:03
rsalvetitusharb: have to look12:03
tusharbrsalveti: sure .. I will try to push the code to glo/people/tushar for you to have a look at the config file .. 12:04
tusharbrsalveti: if you get any clue, that would be great ...12:04
tusharbrsalveti: thanks anyways ...12:04
rsalvetitusharb: I know current package works fine12:04
rsalvetifrom 11.0812:04
tusharbrsalveti: yes .. right .. 12:04
tusharbrsalveti: it was just an experimental effort to boot ubuntu using the latest mainline kernel without any LT enablement patches ...12:05
asacwe have tip hardware packs from our CI server ci.linaro.org/kernel_hwpack/panda/ ... unfortunately the hwpack name doesnt tell you which tree that is built from.12:06
asacseems we have lt-origen as well12:06
tusharbasac: thanks .. I will look into it ...12:07
asactusharb: intersted in panda?12:08
asacor what?12:08
tusharbasac: was looking for origen build12:08
asactusharb: ok. seems we build it for all trees we track ... thought for a moment it was only one tree and then i could have told you what it is :)12:09
asaclook in the tabs here: https://ci.linaro.org/jenkins/view/Linux%20Next%20CI%20Builds/12:09
asacthere are various trees12:09
asacunfortunately you have to look at the log to find out the hwpack name for the build you want to try12:09
tusharbasac: thanks for the pointer12:10
asactusharb: Linux Master or Linux Linaro 3.0 tabs might be most interesting for you ... and maybe arm-soc for-next 12:11
asaclinux-next is currently broken for some other reasonn :)12:11
asactusharb: let me know how useful those are12:14
asactusharb: we will fix the name to be more meaningful in future12:14
rsalvetitusharb: did you try using the config from 11.08 package?12:17
rsalvetitusharb: and which image are you using?12:20
rsalvetidon't know if it's related with the kernel or udev12:20
vinguamitk: ok, let me try to rephrase12:21
amitkvingu: I've made other minor changes, feel free to review12:22
amitkvingu: and perhaps expand on the terminology secton12:23
vinguamitk: the sched_group to which the cpu belong must have a avg load that is lower than the sched_domain one. We will not pull load if the cpu is already above the average load of the sched_domain12:24
ibirisalf_: https://wiki.linaro.org/WorkingGroups/Middleware/Graphics/Notes/2011-08-3112:25
alf_ibiris: thanks!12:26
dokolool, michaelh1|away : anybody to care about bug 832830 ?12:27
ubot2Launchpad bug 832830 in qemu-linaro "qemu-linaro version 0.14.50-2011.06-0-0ubuntu1 failed to build in oneiric" [High,Confirmed] https://launchpad.net/bugs/83283012:27
vinguamitk: if the previous condition is true and the load difference between the busiest group and the local group is greater than the threshold, the cpu will pull tasks12:27
vinguamitk: ok, i will read your update12:27
amitkvingu: ok, that makes sense. Perhaps you want to rephrase that statement to12:29
pm215doko: slangasek does the ubuntu packaging for qemu12:30
dokopm215, does he? he's not with Linaro anymore12:30
pm215that's ok, the ubuntu package isn't a linaro output :-)12:31
dokopm215, as long as you have a fixed package in linaro-overlay ;-P12:31
pm215anyway, if this looks like a problem with configure and it's still present in 2011.09 then I'll have a look at it. if it's a packaging bug then slangasek is your man12:32
dokomost likely libraries have switched to a multiarch location, and the configure is explicitly scanning directories than letting the compiler doing it's job 12:33
* davidgiluk can try the build12:34
rsalvetidoko: slangasek is still maintaining it, even at the overlays12:34
* davidgiluk has a fast oneiric box to hand12:34
=== pfefferz is now known as blarg
dokorsalveti, thanks12:34
rsalvetimabac: did you change l-i-t to support x-loader/spl better as we discussed already?12:34
=== blarg is now known as __oo
dokorsalveti, janimo: btw the arm test rebuild is finishing 22 to go ...12:35
=== __oo is now known as pfefferz
davidgilukdoko: Is that just to provide some fodder for the porting jam?12:35
mabacrsalveti, yes. I'm just about to verift that my panda can boot with the changes.12:37
janimorsalveti, is usb boot coming to omap4/uboot?12:37
mabacrsalveti, I'll push the branch to lp so you can have a look at it12:37
rsalvetidoko: great, good to know, will try to check in details later today12:37
rsalvetijanimo: we'll see, that was the idea12:38
rsalvetijanimo: I believe john is working on it this week12:38
janimorsalveti, thanks12:38
pm215doko: I suspect it is the bug fixed by this commit: http://git.linaro.org/gitweb?p=qemu/qemu-linaro.git;a=commit;h=20fa53ece42bec6ce5db801bead125277b26ab8a 12:38
rsalvetimabac: great, and hopefully we'll be able to use all hwpacks with v2 now12:38
pm215doko: fixed in 2011.0712:39
* davidgiluk has the 0.15.50-blah version building at the moment12:40
mabacrsalveti, let's try it. :) I'd like someone to try with Origen especially since there's some paths that are unique to Samsung boards and that code has moved around quite a lot now12:40
rsalvetimabac: I can help on that, but would need to understand better the changes first12:41
mabacrsalveti, sure. I'll provide examples of the changes to the hwpack configs. there isn't that much to change12:42
rsalvetimabac: ok, great12:42
* davidgiluk wishes he knew why dpkg-buildpackage was ignoring DEB_BUILD_OPTIONS=parallel=1012:43
pm21510> that is a fast box :-)12:44
mabacthe change that I merged today will require you to add 'mmc_id=0:1' to the configs. the Samsungs and ST-E boards use other values though12:44
davidgilukpm215: It's an 8 core with 16GB - we erm found it12:44
pm215you don't actually need to finish the compile, the failure is from configure12:45
davidgilukpm215: Oh, well package seems to be happily building at the moment12:45
pm215if you could confirm that (a) 0.14.50-2011.06-0-0ubuntu1 fails (b) doing the one line configure fix corrects this, that would be interesting info to add to that bug report12:47
hrwdoko: any hint how to make gcc-linaro-cross check /usr/arm-linux-gnueabi/lib instead of /usr/lib/gcc-linaro/arm-linux-gnueabi/lib for libs? headers were easy - edit of cross-include.diff12:47
looldavidgiluk: There are various ways to pass parallel information to packaging; the way you used requires support in the packaging itself, which varies packgae by package12:48
rsalvetimabac: sure, that's fine :-)12:49
looldavidgiluk: However, you can try passing -j10 to dpkg-buildpackage or debuild to parallelize calls into debian/rules; this might break or not, depending on whether debian/rules is correctly written or not, but it doesn't need explicit support by the package12:49
davidgilukpm215: OK, let me just get this build finished12:49
davidgiluklool: Ah ok, I thought the the DEB_BUILD_OPTIONS was equivalent to the -j12:50
=== pfefferz is now known as Guest42774
olseni'm running a linaro-n-ubuntu-desktop on a gumstix overo - do i have to install the modules and run depmod to get the wireless working?12:52
route128ping lool 12:56
=== Guest42774 is now known as pfefferz-x
=== pfefferz-x is now known as _pfefferz
=== _pfefferz is now known as pfefferz_
davidgilukpm215, OK, I can confirm 0.14.50-2011.06-0-0ubuntu1 fails to build12:58
=== pfefferz_ is now known as help
=== help is now known as pfefferz_
mabacrsalveti, https://code.launchpad.net/~mabac/linaro-image-tools/use-spl-for-xloader/+merge/76401 I'll let you know when I think it's done13:02
davidgilukpm215: Configure certainly works with that patch, just building13:02
davidgilukpm215: Hitting some unused variables warnings->errors in usb-ehci.c13:04
rsalvetimabac: great, will give this branch a try 13:04
loolroute128: I'm on13:04
pm215davidgiluk: they should all have been fixed in later releases too13:04
loolroute128: actually, not sure what you're pinging me on, is this about the confcall?13:04
route128lool: question about trustzone stuff13:05
pm215davidgiluk: qemu-linaro-0.15.50-2011.08 built ok, right?13:05
looldavidgiluk: The DEB_BUILD_OPTIONS=parallel=10 or 22,23,24,25,26,27,30,68,69]          13:05
looldavidgiluk: sorry13:05
looldavidgiluk: DEB_BUILD_OPTIONS=parallel=10 or DEB_BUILD_OPTIONS_PARALLEL=10 are supported by a small snippet in debian/rules or by build systems like cdbs or dh (didn't verify this) while -j is passed to rules13:05
davidgilukpm215: 0.15.50-2011.08 is still going because it didn't get the -j 13:05
loolroute128: Ok; on a call right now, but can followup after call13:06
route128lool: just looking for details about what we are doing in the short term (it's on the roadmap q42011-q12012)13:06
davidgilukpm215: OK, 0.15.50-2011.08-0ubuntu3 just finished building fine, but it's the 0.14.... that's in Oneiric13:07
route128lool: np - ping me later when you have a chance13:07
mabacrsalveti, great. here's a panda config http://bazaar.launchpad.net/~mabac/+junk/hwpack.natty.linaro-panda/view/head:/hwpacks/linaro-panda13:07
rsalvetimabac: thanks13:07
pm215davidgiluk: yes, but it's slangasek's call whether he'd rather backport bugfixes or move forward to a newer qemu-linaro13:09
davidgilukpm215: One of the set but not used seems suspicious; it's set pid= to one of two constants but uses ehci->pid instead later, never uses the local pid variable - I can see in 0.15.50 that it does get used13:10
pm215we fixed a whole pile of set-but-not-used bugs, yes. one or two were actual bugs.13:10
mabacrsalveti, my Panda booted using that branch and config just now13:10
amitkrob_lee_linaro: morning, your weekly report is not up :)13:15
davidgilukpm215: There are indeed a whole pile of them in 0.14.x - I've blindly commented out half a dozen or so13:27
pm215I thought oneiric didn't make "set but not used" a fatal error13:28
davidgilukpm215: Seems to 13:30
davidgilukpm215: And frankly given the number of real looking bugs it's finding it's a good thing13:33
patrikrydasa_san, Maybe you should subscribe to the bug https://bugs.launchpad.net/bugs/825327 (related to prelinking and libgui)13:42
ubot2Ubuntu bug 825327 in linaro-android "Investigate apriori issues so as to enable prelink for Android" [Medium,New]13:42
davidgilukhttp://www.nvidia.com/content/PDF/tegra_white_papers/tegra-whitepaper-0911b.pdf interesting paper that a coleague pointed me at - Nvidia's 5 core chip; all A9 but one is made for running slower with less power13:44
amitkasymmetric MP, scheduler head aches :)13:44
davidgilukamitk: Still easier than having to write separate chunks of code to run on a different core13:46
pm215the idea I think is that the scheduler proper doesn't really need to care, something at a lower level flips to the companion core when there's only one low-cpu-use job around13:48
pm215I wonder if "companion core" is a deliberate portal reference...13:51
jk-"weighted companion core"13:51
hrwhope that it will be handled by userspace daemon13:52
anmarndec: sebjan amitk agreen : you have time to do the call now?14:02
ndecin 30 min yes.14:02
amitkanmar: on weekly call now, not for another 1 hr14:02
ndeci am in a call already14:02
ndeci am not free in 1 h...14:02
ndec4 UTC was fine for me !14:02
amitkamitdk: call?14:02
anmarndec: sebjan agreen amitk how about 16:00 UTC then? 14:03
anmarndec: thanks I will count you in. lets see what the others say :)14:03
ndecyou can always count on me ;-)14:04
anmarndec: indeed :-)14:04
=== shadeslayer_ is now known as shadeslayer
davidgilukpm215, qemu-linaro-0.14.50-2011.06-0 eventually built with that patch and a heck of a lot of hacking around the warnings14:44
pm215warnings> are these all "variable set but not used" ?14:48
davidgilukpm215: Most, there is one static function declared but not used as well14:52
pm215davidgiluk: last time I did an oneiric compile "set but not used" was not a fatal error -- are you using a non-stock gcc or has that changed?15:04
pm215from a linaro perspective I'm not too worried about this anyway -- this is all fixed in later qemu-linaro releases15:05
davidgilukpm215: It seems to be giving me 4.6.1-9ubuntu215:05
* davidgiluk hasn't updated it for a week or so15:06
GrueMasterplars: You may get a spurious email from iso.qa on imx51 netboot images.  Ignore it.  cjwatson is running a test.15:08
pm215davidgiluk: I have 4.6.0-5ubuntu1 ; I guess it must have changed (hopefully deliberately)15:08
IGEPv2_userI'm newbie with linux Ubuntu.  I need to prevent root autologon on serial console. I used google, but didn't find working solution. Can anybody help me with this? Thank you15:11
davidgilukpm215: http://packages.ubuntu.com/oneiric/gcc-4.6 says it's currently at 4.6.1-9ubuntu215:13
route128lool: anywhere I can find the the trustzone plans/activities? or who I should talk to? not finding it on the wiki15:36
route128plars: ever heard of Jenkins? is it something used in LAVA?15:36
zygaroute128, we us jenkins15:37
route128zyga: ah interesting thanks!15:37
janimojcrigby, the tree you built the mx5 2.6.38 kernel for Ubuntu is at git://git.linaro.org/people/jcrigby/linux-lt.git ?15:38
zygaroute128, if you wish to know more just ask15:38
route128zyga: I was asked yesterday on a call with a prospective member if LAVA is using it15:39
zygaroute128, currently we are using it for various things15:40
zygaroute128, kernel-ci is based on jenkins15:40
zygaroute128, some testing jobs are triggered by jenkins15:41
zygaroute128, lava itself has no ability to observe source code repos or act on pre-defined schedule15:41
hrwhave a nice rest of day15:41
loolroute128: Ups, sorry15:44
loolroute128: I don't think we're doing much with respect to trustzone right now15:44
loolroute128: it's been in the discussions around boot architecture, and everybody considers it as a requirement, but it's likely going to be implement by pieces of firmware before linux starts which vary from soc to soc and that linaro typically doesn't develop15:45
=== vicky` is now known as vicky
=== vicky is now known as vickyj
=== irar_ is now known as irar
=== salgado is now known as salgado-lunch
=== irar_ is now known as irar
rsalvetiJesseBarker: are we having the unity sync call?16:30
JesseBarkerno, not today16:30
rsalvetiJesseBarker: ok, will go back and try to fix sgx then :-)16:31
tgall_foofabo, thanks for forwarding that ltj note from Bill16:38
Cosmo`pm215 ... I have done a lot of interesting stuff today16:44
Cosmo`i've stepped through the linux booting with the ICE enough to work out that the mmu is initialising properly and getting happily out of the compressed code16:45
pm215that's a good start16:45
Cosmo`and it's actually getting to the calls of printk that show varous things16:45
Cosmo`so i'm assuming it means that the console device is knacked16:45
Cosmo`but i'm not sure how to go about fixing that16:45
Cosmo`i've learnt more today than any other day for a LONG time16:46
pm215console is one of the things most likely to work because just about everybody uses it16:46
pm215so it's almost certainly either (a) you didn't configure the serial driver in or (b) command line argument to set console not right16:46
Cosmo`this is interesting16:46
Cosmo`because i'm using the same config as a working kernel16:46
pm215command line then, maybe16:46
Cosmo`and the command line args are the same as for the working kernel16:46
Cosmo`using ttyAMA016:46
Cosmo`we tried stepping through the execution of printk but it just becomes too complicated at that point16:47
pm215the trick with printk is not to actually wade through the middle of it: if you put a breakpoint at __call_console_drivers() you can find out whether there's actually any consoles in the kernel's list and if so where it is16:49
Cosmo`right ...16:50
Cosmo`i seem to recall trying that and it not hitting it16:50
Cosmo`that might have been because i set the breakpoint before the MMU was setup16:50
Cosmo`so i think that knacks the breakpoints16:50
Cosmo`or at least it did something along those lines16:50
tgall_fooGRRRRR ... nothing like burning an hour trying to find a bug that turned out to be nothing more than a spelling error16:55
Cosmo`now i remember what i was confused pm21517:04
Cosmo`i did set a breakpoint for __call_console_drivers17:04
Cosmo`but it never hit it17:04
pm215Cosmo`: register_console() is another good place to breakpoint17:07
Cosmo`i see17:07
=== salgado-lunch is now known as salgado
tgall_foopfefferz_, ping17:11
Cosmo`doesn't seem to be hitting that either17:12
pm215Cosmo`: do you have CONFIG_EARLY_PRINTK set?17:18
Cosmo`i don't know, i'll look17:22
Cosmo`no, apparently not17:23
Cosmo`looks like it not mentioned in my config at all, pm21517:25
pm215it's in the config for the linux-arm.org kernel17:25
pm215if you don't have that then it's rather further on in the boot process before printk actually gets sent to the console, so a crash before that point looks like no-output17:26
Cosmo`at the moment i'm testing just a linaro kernel17:26
Cosmo`they both seem to show the same output17:26
Cosmo`just using realview-smp_makedefconfig17:27
Cosmo`i could build another image using the patch against 2.6.38 and then run that through the ICE and see what's going on17:30
Cosmo`might show some more stuff17:30
pm215you really want early printk if you don't have it enabled in configure and on your command line.17:31
Cosmo`what's the command line option for early printk?17:32
Cosmo`as i'll enable it in the config17:32
pm215Cosmo`: I think you can just say 'earlyprintk'17:33
Cosmo`it looks like it has to be17:33
pfefferz_tgall_foo, pong17:35
pm215that's x86; the arm version ignores all its parameters17:35
Cosmo`ah, i see17:35
Cosmo`thank you17:35
=== prpplague is now known as prpplague_afk
tgall_foopfefferz_, on android, auxv is present right ?17:38
pfefferz_not sure17:38
pfefferz_I don't know what auxv is17:39
tgall_foo how about this .. on one of your android boxen ls /proc/self/auxv17:39
pfefferz_one sec, need to program a build17:40
Cosmo`the amount of output with earlyprintk17:50
Cosmo`however this is good17:50
pm215oh, you get some console output now? cool17:51
Cosmo`because there's loads of exceptions17:51
Cosmo`but at least i'm seeing them niow17:51
pm215Hmm. stick it into pastebin and post a url?17:51
Cosmo`i can digest17:51
Cosmo`as they all seem to be the same17:52
pm215pastebin would be best17:52
Cosmo`Internal error: 801 [#8] PREEMPT SMP17:52
Cosmo`over and over and over17:52
steev_are the freescale linaro kernel guys about?17:54
janimosteev, they may be sleeping already, I think most are in China17:55
janimoericm_, ^ ?17:55
steev_ah, okay :(17:55
janimosteev, what is your question btw?17:55
steev_i just wanted to find out what it will take to get the ipu driver into mainline.  i know shaun was working on devicetree support, but now he's working on mx6, so not sure what the status of things are17:56
janimojcrigby, around? We'd need to enable the EXT4 huge file support config thingy that prevents booting ext4fs in the 2.6.38 mx5 kernel in oneiric17:56
pfefferz_tgall_foo, ls /proc/self/auxv17:56
pfefferz_produces /proc/self/auxv  17:56
pm215Cosmo`: no, really, I'd like to see more output than that. (haven't seen that message before, but it's always much easier to guess what's going on when you have more context for error messages)17:57
tgall_foopfefferz_, excellent, thanks!17:57
janimoI'd gladly do the work - I have the tree checked out, but I am afraid I'd need some more guidance. The packaging a tree does not have info about workflow for adding new changes.17:57
Cosmo`right pm215, give me a second17:58
janimojcrigby, I see you have consistent changelog entries generated by a template and probably a sequence of steps before  uploading. And probably want to keep your git tree in sync with the package17:58
steev_janimo: that info would be nice as we've been told to use the linaro kernel as well17:58
janimosteev, this is what  I use now https://wiki.linaro.org/Resources/HowTo/PackageYourOwnKernel17:59
janimoit is helpful but there are aspects not covered there17:59
janimosteev, no idea about your ipu driver  question, sorry17:59
Cosmo`here you go pm215 17:59
c10udhello, i'm using linaro 11.08 with Linux linaro-desktop 2.6.38-1000-linaro-lt-mx5 #8-Ubuntu Tue Aug 16 15:08:35 UTC 2011 armv7l armv7l armv7l GNU/Linux -- I'm doing some stuff through USB (capturing pics with custom program and lib) and i see memory usage is increasing each time i snap a pic of ~3MiB. Monitoring the situation with top i see my program's memory isn't growing (nor others). Could it be a leak in the kernel?18:00
steev_janimo: oh i didn't think you would, just wanting more info on the status of things18:00
pfefferz_rsalveti, which tree are you using for 4460?18:03
rsalvetipfefferz_: tilt, but I'm still fixing it18:03
rsalvetineed fixes at u-boot and kernel18:03
Cosmo`i think pm215 it's got to be something to do with the cross compiler suite i'm using18:03
pfefferz_would you send me the exact git/branch/config?18:04
Cosmo`i'm just using the latest one from codesourcery18:04
rsalvetipfefferz_: sure, after I get it working here18:04
pfefferz_rsalveti, actually I'm just collecting some data18:04
pfefferz_rsalveti, I want to see what are all the things we need to compose the Android version of it18:05
pfefferz_so everyone can get on the same page18:05
rsalvetipfefferz_: sure18:05
pfefferz_rsalveti, I think it may be as easy as,18:05
pm215Cosmo`: ah. the interesting bit here is "Unhandled fault: alignment exception" -- the kernel is trying to do an unaligned access which the hardware is faulting18:05
pfefferz_...not the right word...18:06
Cosmo`you know i think this is a problem with the latest codesourcery crosscompiler suites18:06
pfefferz_anyway...I'm just trying to get some clarity18:06
Cosmo`as i've just found an excerpt from a mailing list with someone having the same problem with the same toolset i'm using18:06
Cosmo`he;s identified an old version which works18:06
Cosmo`so i'll try that and report back18:06
pm215Cosmo`: you should make sure the combination of your kernel config and the compiler are telling it to generate code for armv618:06
pm215(ie not armv7)18:07
Cosmo`yes, it 'should' be fine18:07
Cosmo`as the config seems correct18:07
Cosmo`let me try the other toolchain and then i'll look further into the arm7/arm6 conundrum18:07
Cosmo`thanks for your help :)18:07
pm215you can probably add -mno-unaligned-access to your CFLAGS for the kernel build18:09
Cosmo`i'll do that18:09
Cosmo`does that sit in the main toplevel Makefile?18:09
Cosmo`oh god, i should really stop, i need to go home18:10
Cosmo`but there's always one more thing to do ...18:10
Cosmo`my brain is so fried18:10
pm215I think you can just say 'make [....] CFLAGS=-mno-unaligned-access' but I could be wrong18:12
pm215anyway, that's a workaround, but we should figure out who's actually right/wrong here, the kernel or the compiler18:12
pm215if you say 'make [...] CFLAGS=...  V=1' it will print the actual gcc commands as it runs them18:13
Cosmo`ok, fresh uImage coming for the old version of the toolchain18:13
pm215google suggests maybe EXTRA_CFLAGS=-mno-unaligned-access actually18:14
Cosmo`i'll see how this uImage fares18:15
pm215http://comments.gmane.org/gmane.linux.ports.arm.kernel/117863 # looks like the kernel folk are aware of it, so for the moment the EXTRA_CFLAGS will do if you're compiling an older kernel with this toolchian18:20
Cosmo`today has been fun anyway18:20
Cosmo`this is the sort of crap i wanted to be doing when i started my PhD, heh18:20
mruoh, you've hit *that* bug18:21
=== prpplague_afk is now known as prpplague
Cosmo`you realise mru, that that phrase sums up my entire day18:23
mrusee also http://article.gmane.org/gmane.linux.ports.arm.kernel/13164318:25
Cosmo`well, thank you very much18:26
Cosmo`this all could have been avoided if i'd looked in the right place18:26
Cosmo`but it's been worth it, i've learnt a huge amount about how the linux kernel works today, and also got the ICE working18:26
mruthis is of course all russel king's fault18:27
mrufor insisting on *turning on* strict alignment checking18:27
Cosmo`i'm lost as to the specifics of why this broken right now18:27
Cosmo`but i'll look at it tomorrow18:27
Cosmo`right now i seriously need to go home and have a bath18:27
Cosmo`my brain feels like it's about to melt out of my ears18:27
mruit fails because some gcc versions (cleverly) emit unaligned loads and stores if the target cpu supports them18:28
mruand rmk refuses to flip one bit in a control register18:28
Cosmo`right ...18:28
Cosmo`i'll read it tomorrow and work it out18:29
Cosmo`what does rmk maintain?18:29
mrucore arm support18:29
Cosmo`it looked to me like all the arm linux stuff was pretty deprecated18:29
Cosmo`i mean the toolchains from his site are ... so old18:29
mruhe is stuck in the armv5 era18:30
Cosmo`the latest post on his website it from 200818:30
Cosmo` is18:30
Cosmo`right, now i really need to go home, i'm still in the lab ...18:30
Cosmo`thanks mru and pm215 18:30
=== michaelh1|away is now known as michaelh1
rsalvetipfefferz_: http://lists.denx.de/pipermail/u-boot/2011-September/101928.html for u-boot20:20
rsalvetipreparing the other patch set for kernel20:20
repetejoey, Hey dude20:29
repetejoey, got a recommendation on a small dev board.  I just want to use it to build a NAS.20:30
joeyhey repete!20:30
repetejoey, Hey :-)20:31
joeyI'm sure asac, michaelh1, lool and others will have some good ideas20:31
joeyI'd say panda but they are in short supply20:31
loolrepete: there aren't many with SATA though20:33
loolimx53 LOCO/quickstart has SATA and is very cheap; panda is nice and complete but it's USB2 and short supply20:34
repetelool, thx.  is that sata or esata?20:35
loolrepete: SATA, and without power20:41
mkedwardsorigen is also USB2, but otherwise potentially interesting if you want wifi20:41
repetelool, thx20:41
mkedwardswhich reminds me, I need to order up a batch of origens20:41
mkedwardshow exactly do I order up origens?20:46
mkedwardsah: origenboard.org20:47
mkedwardssold out, natch20:48
mkedwardsand inSignal seems to have gone almost completely dark since the first batch shipped; not good20:50
gcllinusw_: ping?20:52
mkedwardsand forums and wiki are pwned by spammers20:53
mruforums *are* spam21:03
mkedwardsmru: very true21:04
=== lbt_away is now known as lbt
linusw_gcl: pong - no notification on this desktop...22:59
plarsroute128: was off today, just got home23:36
plarsroute128: jenkins, as zyga alluded, is a tool mostly geared towards CI (building, running simple tests, etc)23:37
plarsroute128: it's not really a part of lava, but we do use it for the build part of CI, then hand off jobs to lava for testing23:37
gcllinusw: Are you going to be able to make it to Prague for the arm maintainership workshop at ksummit?23:37
gcllinusw: you're pretty key, so I'd like to have you there.23:38

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