/srv/irclogs.ubuntu.com/2010/03/15/#ubuntu-arm.txt

=== dyfet` is now known as dyfet
=== DanaG1 is now known as DanaG
hrwhi09:37
=== ramana-away is now known as ramana
loolpersia: Hmm I don't think qemu-debootstrap works in pbuilder-dist -- it needs an arch arg11:44
loolpersia: And I doubt it works in mk-sbuild either11:45
loolEh I see the same biarch stuff in there :)11:45
persiaI tested both before I shipped, and successfully built armel and powerpc chroots on an amd64 machine.11:47
persiaSo, unless something else changed since I pushed, it works, even if it's not supposed to work.11:48
persiamk-sbuild is my primary workspace, with pbuilder-dist attempting to keep up out of courtesy for pbuilder users.11:48
persiaWith the latest mk-sbuild, I consider pbuilder obsolete.11:48
persia(because of support for type=file chroots).11:49
cooloneyplars: for the suspend/resume regression, which kernel works on your side?11:50
cooloneyplars: i failed to find any kernel on my side.11:50
loolpersia: Ah I'm wrong indeed; pbuilder and sbuild pass --arch to debootstrap11:53
loolSo it just works11:53
loolI was expecting one had to care for --arch to be passed as well, but that's already the case11:54
persiamk-sbuild has always passed --arch.  It was added to pbuilder-dist some time ago (gutsy or hardy, I think).11:56
persiaOne does have to care for --arch, but not quite so carefully.11:56
persiaNote that neither of these tools are capable of correctly building chroots for unsupported Ubuntu distributions right now.  I'm not sure how much that matters, but figure it's worth mentioning.11:57
loolpersia: Would you have some time to review some early documentation?12:08
persialool: What sort of documentation?12:09
loolpersia: https://wiki.ubuntu.com/UbuntuDevelopment/Ports12:09
loolpersia: On development with qemu / for armel12:09
persiaAh, cool.  I definitely have time to review that :)12:10
loolpersia: This is very much tools focused rather than tasks-focused; I wish I'd document "How to build an armel package" stories12:10
persiaAlso, on a related note, I'd like to get your input on how we should adjust pbuilder-dist and mk-sbuild to work in Debian.  I know you've done a lot to bring nomenclature in line, but I'm not that familiar with the names in Debian.12:10
loolpersia: I'm afk for most of today, but I wish you go straight to making fixes or mention what should be fixed to me12:11
persialool: For straightforward stuff, I'll just edit.  For discussion / debateable items, I'll send you email.12:11
loolpersia: pbuilder-dist > I think this functionality should be merged in pbuilder itself; I have some strict views on how it would be done, but didn't finish the (large) task of cleaning things up to make this possible12:11
loolmk-sbuild > I'm not a heavy sbuild user, but I wish to change that; the sbuild team is quite active upstream touhg12:12
persialool: You'll merge *all* of pbuilder-dist into pbulder?12:12
loolpersia: Pretty much12:12
loolpersia: In fact, I think I already cover most of it12:12
persiaFo mk-sbuild, I mostly just need some pointers to how to invoke qemu-debootstrap in Debian.12:12
loolWas qemu-debootstrap merged already?12:13
persiaTo where would it be merged?12:13
persiaThe package in which it lives in Ubuntu doesn't exist in Debian, nor does anything obviously similar.12:13
loolI think I sent it to qemu12:15
loolBecause it basically requires qemu and the binfmt stuff12:15
loolWhich is in qemu-user-static in Debian12:15
loolhttp://bugs.debian.org/cgi-bin/bugreport.cgi?bug=57295212:15
ubot4Debian bug 572952 in qemu "qemu-debootstrap wrapper" [Wishlist,Open]12:15
zumbi_lool: that only works on armel12:16
zumbi_that is why i wanted to test newer release 0.1212:17
loolzumbi_: What only works on armel?12:17
zumbi_to see how that works for mips, pcc, etc... *re Bug357295212:18
zumbi_sorry, i have another keyboard layout bug 57295212:18
ubot4zumbi_: Error: Bug #572952 not found.12:18
persialool: Thanks.  I'll tack that.12:18
loolzumbi_: I created a powerpc chroot using qemu-ppc when writing qemu-debootstrap12:18
loolSo I think it works with our qemu-ppc static binary at least12:18
persiazumbi_: I've also tested it with powerpc successfully.12:18
persiazumbi_: Please test with arbitrary architectures, and let us know how it doesn't work.12:19
zumbi_uhm.. i should then test again, we tried it a couple weeks ago12:19
persiaAlternately, if you ran into an older version of mk-sbuild or pbuilder-dist that has the "only supported for armel" warning, please upgrade.12:19
zumbi_uhm.. note that i use debian stuff12:19
persiaAh.  I haven't ported my stuff to work on debian yet, and lool's work is pending the recently mentioned bug.12:22
persiaBut if 572952 lands for squeeze, the necessary ubuntu-dev-tools stuff will as well.12:25
zumbi_yes, i would love to see it applied, also working for all arches, not only armel, but that is qemu lack of enough support for emulate other arches12:27
persiaI thought powerpc worked in Debian as well.12:32
persiaOr, rather, powerpc guests.12:32
persiaI'm less sure about other stuff.12:32
saeedericm_: ping13:13
ericm_saeed, pong13:13
saeedericm_: regarding kexec patches13:14
saeedI send the patches to mailing list, but has been rejected13:14
saeeddo you know if both patches were needed also for imx?13:15
ericm_saeed, I guess the one for L2 is needed as well - according to cooloney's test result13:16
ericm_it's really annoying I know, I didn't look too much into L2 on ARMv7, but it's totally different than ARMv5, guess we need some neater solution for the outer cache13:16
saeedericm_: can you please check with cooloney's regarding the other patch?13:17
ericm_saeed, sure13:17
saeedericm_: now I'm debugging the highmem issue (crash when booting from sata)13:18
ericm_saeed, cool13:18
saeedericm_: can you please also check if this issue happens on imx?13:19
ericm_saeed, called him13:21
dmartNCommander: you awake yet?13:21
ericm_saeed, he tried kexec with both patches, not individually tested, so the first question is not sure which one is mandatory13:21
ericm_or both are13:21
ericm_and babbage with imx51 has only 512MB, so actually HIGHMEM wasn't actually even used13:22
saeedericm_: you can pass kernel vmalloc=256M, that will make some dram as highmem13:24
ericm_dmart, do you know if Catalin has an IRC name here on freenode? I guess Saeed and I can talk a little bit about some L2 issus we met during kexec debugging13:24
ericm_saeed, indeed - I'll see if he has time to test this later13:24
saeedericm_: thanks13:25
ericm_saeed, no problem13:25
dmartericm_: not sure whether he's usually online13:25
dmartOK, I think he will join - he's cmarinas13:26
ericm_dmart, thanks13:27
NCommanderdmart: yes13:33
dmartNCommander: hi, dunno if you saw my post on the OOo bug, but it looks like the fix we had has a problem: privateSnippetExecutor pushes some arguments on the stack, but they never get removed after cpp_vtable_call returns13:40
dmartNCommander: this is a more complete attempt http://pastebin.ubuntu.com/395620/13:43
dmartSince I have a OOo build tree now, I rebuilt libgcc3_uno.so and clobbered the installed version in lucid with it--- OOo seems to start up and "work"; is there anything else I should look for?13:44
NCommanderdmart: no, that was our usual test since we never got past start up :-)13:47
NCommanderdmart: who caught the arguments on the stack problem? (I didn't have an OOo crash, but I didn't abuse it hard)13:48
NCommanderdmart: actually, there IS an OOo test suite believe it or not. Might be worth trying to run it if you cna spare a few weeks13:48
persiaNah.  The test suite should only take 100-150 hours to run.13:49
dmartOne of the other guys on my side, by inspection (it was a dumb mistake, we push some stuff which is never popped)13:49
dmartIt means exception unwind annotations are needed after all, to clean up the stack if an exception happens, but it seems reasonably straightforward now we understand what the code is doing.13:50
dmartpersia: is the testsuite part of the OOo source package?13:50
* persia checks13:51
persiadmart: My apologies: I don't appear to be able to answer the question.  Attempting to download the ooo source from my local mirror (~50cm away) will apparently take 9 hours, and trying to unpack it directly on the mirror seems to break either with the version of dpkg on that server or with the amount of available RAM14:27
* persia looks forward to more hardware exceeding the 512M boundary14:27
persiaNCommander: Maybe you could share?14:28
NCommanderdmart: no, its an addon, and OOo has to be built in a special way14:30
dmartargh, so, 4-5 days + x14:30
dmartCan I suggest that you verify what's suggested in http://pastebin.ubuntu.com/395620/ works for you and push that (assuming it does work, of course)14:31
dmartI did some standalone testing on this, checking that the normal and exception return paths do the right thing: it appears to behave well.14:32
dmartNCommander: scrollback the last few posts (forgot to ping you)14:35
NCommanderdmart: sure14:36
dmartNCommander: should I assign the lp bug back to you in the meantime?14:37
NCommanderdmart:14:37
NCommanderyeah14:37
dmartok14:38
dmartdone14:38
prpplaguedavidm: greetings15:50
=== amitk is now known as amitk-afk
=== NCommander is now known as Guest5889
=== NCommand1r is now known as NCommander
NCommanderasac: is chromium on ARM known to be broken?20:08
NCommandercan someone who has Ubuntu on ARM running on something beside dove or imx51 test something for me?21:14
prpplagueNCommander: depends on what you need21:15
prpplagueNCommander: i'm running a beagle21:15
NCommanderprpplague: can you download and install chromium-browser from the archive (in lucid) and see if it works properly21:23
prpplagueNCommander: sorry, i am not able to do that at the momment, sorry21:23
prpplagueNCommander: no display running atm21:24
NCommanderprpplague: X11 forwarding over SSH?21:24
prpplagueNCommander: only uart console21:25
NCommanderprpplague: ouch21:25
NCommanderprpplague: thanks anyway21:25
prpplagueNCommander: i should have something up and running towards the end of the week, but i suspect that is too far down the road for you21:25
NCommanderprpplague: yeah :-/21:26

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