slangasek | jcrigby: have opened bug #591968 with my notes to date on this issue (they were starting to fall out of my head) | 00:17 |
---|---|---|
ubot2 | Launchpad bug 591968 in coreutils (Ubuntu Maverick) (and 1 other project) "coreutils 8.5-1ubuntu1 FTBFS on armel, powerpc, sparc (test-linkat assert fail) (affects: 1) (heat: 6)" [Critical,Triaged] https://launchpad.net/bugs/591968 | 00:17 |
NCommander | ogra: so you fixed jasper? | 00:19 |
bullet9mm | go to bed...good night! :) | 00:36 |
GrueMaster | NCommander: ogra is either in bed or traveling. | 00:38 |
=== bjf[afk] is now known as bjf | ||
jcrigby | slangasek: ok thanks | 00:47 |
=== bjf is now known as bjf[afk] | ||
* mozzwald is away: off to grandmothers house I go | 01:29 | |
* cwillu_at_work pokes around in rcn-ee's general direction | 02:14 | |
jcrigby | slangasek: still around? | 02:14 |
slangasek | jcrigby: yep | 02:14 |
jcrigby | just added a comment to the bug | 02:14 |
jcrigby | I won't repeat it here | 02:15 |
slangasek | jcrigby: hmm, is that strcmp() actually failing for you? It didn't for me | 02:19 |
slangasek | only the SAME_INODE() check fails | 02:19 |
slangasek | (sorry, the line number is wrong in the logs I posted to the bug report because I inserted some other debugging) | 02:19 |
slangasek | should be line 73 in the original file | 02:19 |
jcrigby | I only have your strace data. Strace does not work in qemu user emulation mode | 02:21 |
slangasek | OK | 02:22 |
jcrigby | from the strace data, I would predict that strcmp would fail | 02:22 |
jcrigby | given the ... on the end of the strings | 02:22 |
jcrigby | string | 02:22 |
slangasek | I think that's an abberation in the strace output; if it was really non-null garbage data, I would have expected it inside the double quote | 02:22 |
slangasek | rather - it may be signalling that there's more data there, but the string appears to be properly null terminated | 02:23 |
jcrigby | can you try using calloc instead of malloc in areadlink_with_size and see if the problem goes away? Just to humor me:) | 02:25 |
slangasek | sure, lemme find it | 02:26 |
slangasek | you want calloc(1, buf_size)? | 02:26 |
jcrigby | its in lib/areadlink-with-size.c | 02:27 |
jcrigby | yes, that should do it | 02:27 |
slangasek | rebuilding | 02:27 |
jcrigby | Meanwhile I'm going to figure out what strace really means by "string"... | 02:28 |
slangasek | jcrigby: sorry, have to afk now and the build is still running (I took the lazy way out to rebuild, which means it's rebuilding everything and will take a while) - I'll report results when I return later this evening | 02:36 |
slangasek | btw, it looks like the code does append the trailing \0? | 02:38 |
slangasek | if (link_length < buf_size) | 02:38 |
slangasek | { | 02:38 |
slangasek | buffer[link_length] = 0; | 02:38 |
slangasek | return buffer; | 02:38 |
slangasek | } | 02:38 |
jcrigby | slangasek: oops | 02:41 |
jcrigby | slangasek:ok sorry about the wild goose chase. I have noticed one other oddity. The test uses the size returned from lstat in the call to areadlink_with_size which should be 17 according to the strace data. areadlink_with_size adds 1 to it and calls readlink. In the amd64 case we see 18 which makes sense. However, the armel strace shows 1025 which means that 1024 was passed to areadlink_with_size. | 03:03 |
cwillu_at_work | what would "unknown location" be implying? | 04:19 |
cwillu_at_work | er, "unknown relocation" while loading a kernel module | 04:19 |
jcrigby | slangasek: I see this in the build log: CC test-linkat.o | 04:24 |
jcrigby | In file included from test-linkat.c:44: | 04:24 |
jcrigby | test-link.h: In function 'test_link': | 04:24 |
jcrigby | test-link.h:47: warning: implicit declaration of function 'stat' | 04:24 |
jcrigby | test-link.h:98: warning: implicit declaration of function 'mkdir' | 04:24 |
jcrigby | test-linkat.c: In function 'check_same_link': | 04:24 |
jcrigby | test-linkat.c:65: warning: implicit declaration of function 'lstat' | 04:24 |
jcrigby | test-link.c does not get errors. Its includes are different. Can you try including sys/stat.h in test-linkat.c? | 04:27 |
cwillu_at_work | hmm, I can't modprobe anything | 04:45 |
* mozzwald is away: time for some zeez | 04:56 | |
slangasek | jcrigby: oh, nicely caught | 06:35 |
slangasek | jcrigby: I saw those link warnings but wrote them off, thinking that the prototypes would Just Work by default | 06:36 |
slangasek | jcrigby: so, the tests pass now - do you want to put that into a bzr branch that I can merge & upload? | 06:38 |
ogra | NCommander, were your debian-cd changes merged or uploaded in any way ? (why did you close the workitem) | 07:46 |
NCommander | ogra: the work is done, there is a seperate work item for merging I thought | 07:52 |
* ogra doesnt see one that talks about merging | 07:53 | |
ogra | i see one that talks about debian-cd and one that talks about the post processing | 07:53 |
ogra | and one about bziping | 07:54 |
NCommander | ogra: huh, maybe I'm loosing it | 07:54 |
ogra | anyway, leave it like that | 07:54 |
* NCommander is having one of those weeks | 07:54 | |
NCommander | ogra: I'm filing MIRs for x-loader/x-loader-omap4 and uboot-omap | 07:54 |
NCommander | ogra: much sanier for the code if they're in main since apt-selection doesn't seem to work in universe, and seems non-trivial to fix | 07:55 |
ogra | it has to, how else would distros like edubuntu or xubuntu roll their images | 07:55 |
ogra | but i agree, MIR is definately better | 07:55 |
NCommander | ogra: its more a matter that when control passes into the boot/post-boot scripts, the necessary environment for apt-selection goes "poof" as far as I can tell | 07:56 |
NCommander | and its proving difficult to get that to work. | 07:56 |
NCommander | I'll keep poking at it, but we should have these in main anyway so only if we do a 10.04.1 do we need to care about actually retrieving these files from universe | 07:57 |
ogra | well, there has to be a way that edubuntu and xubuntu use for the alternates | 07:57 |
ogra | but dont bother now, MIRs are fine | 07:57 |
ogra | i'm still experimenting with jasper, it could be that we need to generate ext2 by default and convert after resizing to ext3 | 07:58 |
ogra | bug 5583317 is really bad | 07:58 |
ubot2 | ogra: Error: Bug #5583317 not found. | 07:58 |
* ogra pokes the bot | 07:58 | |
ogra | bug 583317 | 07:58 |
ubot2 | Launchpad bug 583317 in genext2fs (Ubuntu) "genext2fs creates revision 0 filesystems instead of revision 1 (affects: 1) (heat: 156)" [Undecided,Confirmed] https://launchpad.net/bugs/583317 | 07:58 |
ogra | pfft | 07:59 |
ogra | though thats apparently something to be fixed in genext2fs | 07:59 |
NCommander | ogra: maybe we should use loop mounting | 07:59 |
ogra | for now i added an extra ext2fsck to livecd-rootfs but i'm not sure that will not make livefs.sh explode | 08:00 |
ogra | since it exits with 2 | 08:00 |
ogra | lets keep that as last resort, i'm not a fan of loop mounting | 08:00 |
NCommander | ogra: I think you can pipe to true, to force a return of 0 | 08:00 |
NCommander | or osme other stupid trick | 08:01 |
NCommander | or trap an exit code of two | 08:01 |
ogra | well, it needs testing anyway but i wont have the time to run a livefs build on my lappie the next tw days | 08:01 |
ogra | luckily nobody uses the ext3 code so i could just add it to the function for now, if it doent work we can always revert | 08:02 |
ogra | using ext2 might also have speed advantages | 08:02 |
ogra | in jasper | 08:03 |
ogra | the resizing takes between 6 and 7 mins | 08:03 |
ogra | NCommander, btw, not sure you saw it in your backlog http://people.canonical.com/~ogra/jasper/ | 08:03 |
ogra | there is an image to test if you feel bored enough :) | 08:03 |
ogra | though there is a lot that doesnt work yet, the resizing does at least | 08:04 |
* ogra wonders why his NM applet tends to disappear from the panel | 08:07 | |
NCommander | ogra: _\o/_ | 08:09 |
NCommander | ogra: I stll have tons to screw around with the d-cd backend code, but we should probably do an initial merger soon | 08:09 |
NCommander | ogra: do we care about 10.04.1 for omap3 at all? | 08:09 |
NCommander | Now that I have maverick locally, I can choose not to care | 08:10 |
ogra | we do | 08:10 |
NCommander | ogra: so we want pre-installed images for 10.04.1? | 08:10 |
ogra | there are a bunch of bugs we need to milestone before the meeting | 08:10 |
hrw | morning | 08:10 |
ogra | no, but we want some kernel SRUs for 10.04.1 in omap3 | 08:10 |
NCommander | ogra: ok, so I'll leave the boot scripts for lucid alone, and just merge those into the maverick tree | 08:11 |
ogra | like the USB NIC udeb misses a lot of drivers | 08:11 |
ogra | yeah | 08:11 |
NCommander | ogra: works for me, that makes my life simpiler. I'm going to add a work item on the MIRs for the bootloader | 08:11 |
ogra | we wont change images, but there are some SRU bits we want, we need to add them to the release meeting report tomorrow | 08:11 |
ogra | dont ! | 08:12 |
ogra | it will rise the WI count ! | 08:12 |
NCommander | ogra: https://bugs.edge.launchpad.net/ubuntu/+source/x-loader/+bug/592045 | 08:18 |
ubot2 | Launchpad bug 592045 in x-loader-omap4 (Ubuntu) (and 3 other projects) "[MIR] x-loader/x-loader-omap4/u-boot-omap3/u-boot-omap4 (affects: 1) (heat: 10)" [High,New] | 08:18 |
NCommander | ogra: ok, no work item | 08:19 |
ogra | great, thanks | 08:21 |
ogra | we'll mention the bug in the report, that saves us the WI | 08:24 |
JamieBennett | /join #ubuntu-devel | 08:52 |
ogra | heh | 08:53 |
=== lag_ is now known as lag | ||
JameswStubbs | Hello | 10:29 |
JameswStubbs | Can someone please point me to some documentation for creating an initrd for an arm boot? | 10:30 |
JameswStubbs | What's the difference between iop4xx, orion5x, iop32x | 10:38 |
JameswStubbs | ? | 10:38 |
bullet9mm | hi all!! | 10:44 |
JameswStubbs | Hello | 10:44 |
asac | ogra_cmpc: will you guys make good step forward on the thumb2 porting this cycle? | 10:54 |
asac | (I assume not, but wanted to check if that is priority for your team) | 10:54 |
=== ericm|ubuntu is now known as ericm-afk | ||
JameswStubbs | is armv5tej the same as armv6? | 11:52 |
hrw | no | 11:56 |
hrw | armv6 runs armv5te binaries but not reverse | 11:56 |
JameswStubbs | So a distrobution compiled for armv5te won't work with an armv6 kernel? | 11:57 |
hrw | will work | 11:57 |
JameswStubbs | Ah, good, thank you | 11:57 |
hrw | armv6 binaries do not work with armv5te hardware | 11:57 |
hrw | etc | 11:57 |
JameswStubbs | Thanks | 11:58 |
hrw | JameswStubbs: if you are rather x86 guy then remind i386<486<568<686 and then armv4 < armv4t < armv5te(j) < armv6 < armv7a | 11:58 |
JameswStubbs | So arm like x86 has backwards compatibilty with older archs | 11:59 |
hrw | yes | 11:59 |
hrw | but in arm world it is more clean | 11:59 |
hrw | in x86 i586 != i586 if cpu are from diferent vendors | 12:00 |
hrw | amd k5 has 3dnow which intel pentium-mmx do not have. but both are i586 | 12:00 |
JameswStubbs | Do you prefer arm to x86? | 12:01 |
hrw | or intel pentium pro (first i686) contra Core i7 (latest i686). first has only mmx, second has lot of addons... | 12:01 |
hrw | JameswStubbs: for my gadgets I prefer arm. for small low power I prefer arm. for my desktop I prefer x86-64 | 12:01 |
JameswStubbs | This is my first experience with arm on embedded, atm it seems to be far more difficult to get a busybox than x86 | 12:02 |
hrw | my laptops are x86 and x86-64. but arm powered one would be nice | 12:02 |
hrw | JameswStubbs: what hw? | 12:02 |
JameswStubbs | iPhone 3g | 12:03 |
JameswStubbs | armv6 | 12:03 |
hrw | s3c6410 iirc | 12:03 |
JameswStubbs | Yes produced by samsung | 12:04 |
JameswStubbs | Has imgtech power vr althought gpu drivers are not yet reverse engineered | 12:04 |
hrw | powervr... | 12:05 |
JameswStubbs | PowerVR (MBX Lite) | 12:13 |
JameswStubbs | hrw using rootstock now with iPhone's kernel, should be able to emulate using qemu | 12:53 |
hrw | qemu requires qemu kernel | 12:55 |
JameswStubbs | Can you not use a different arm kernel? | 12:55 |
hrw | arm kernels are bound to machine or set of machines | 12:57 |
hrw | there is no such thing as 'one kernel image works on any arm device' | 12:57 |
JameswStubbs | The iphone kernel works on the iphone :/ | 12:57 |
JameswStubbs | Android is already booting | 12:57 |
JameswStubbs | The kernel is made just for the iPhone | 12:58 |
sivang | hi all | 13:15 |
JameswStubbs | Hello | 13:15 |
sivang | so there's an official arm port? canonical supported? | 13:20 |
* sivang is curious who does the arm port for Marvell | 13:20 | |
hrw | sivang: depends which marvell cpu | 13:21 |
hrw | sivang: there is official ARM port but armv7a only (cortex-a8/a9 core) | 13:21 |
sivang | hrw: coretx-a8 is TI's no? | 13:22 |
sivang | hrw: or are marvell also producing it? | 13:22 |
hrw | sivang: nope | 13:22 |
sivang | hrw: N900 uses TI's right? | 13:22 |
hrw | sivang: Cortex-A8/A9/A5 and M0/M3/M4 are names from ARM Ltd. | 13:22 |
hrw | sivang: yes, it has TI OMAP3530 | 13:23 |
sivang | hrw: so it's their specs | 13:23 |
sivang | okay | 13:23 |
sivang | thanks | 13:23 |
sivang | hrw: and the port is officiall supported by Canonical due to OEM's demand? | 13:23 |
hrw | sivang: ARM creates core, vendors buy license and make a silicon | 13:23 |
hrw | so Cortex A8/A9 you can get from TI, Freescale, Marvell, ST-Ericsson, NVidia and few others | 13:23 |
hrw | sivang: I was not here when port was created so cant comment | 13:24 |
sivang | hrw: okay, cool, thanks. | 13:24 |
armin76 | sivang: considering they've done it without using public hw, i'd say yes | 13:24 |
armin76 | nevertheless ask ogra, lool and asac | 13:24 |
sivang | armin76: what do you mean using public hw? | 13:25 |
armin76 | hrw: qualcomm, how could you forget so important one? :P | 13:25 |
armin76 | sivang: one you can buy | 13:25 |
armin76 | sivang: as in smartbooks, nettops, you know... | 13:26 |
armin76 | stuff you can buy, not development stuff | 13:26 |
* sivang wonders if ogra_cmpc remebers him | 13:26 | |
sivang | or asac for that matter | 13:26 |
sivang | :) | 13:26 |
armin76 | i can also tell you that here there are a lot of ppl from vendors, so that reinforces my thought :) | 13:27 |
sivang | i see | 13:31 |
asac | sivang: lucid has a official marvel dove port | 13:57 |
lag | Has anyone seen this error before? omapdss DISPC error: SYNC_LOST_DIGIT | 13:59 |
lag | It renders my Panda unusable | 13:59 |
sivang | asac: nice, I see. | 14:05 |
asac | lag: we are waiting for the pandas ;) ... so cant tell! | 14:06 |
hrw | asac: ogra has panda | 14:06 |
asac | ogra also has eucalyptus ;) | 14:07 |
* lag shouts "OGRA" | 14:08 | |
lag | ogra_cmpc: Can you give me a shout when you return please? | 14:08 |
asac | ogra_cmpc: what triggeres the flash-kernel thing? | 15:01 |
asac | is that in postinst of linux-image? | 15:01 |
asac | or a trigger? | 15:01 |
=== bjf[afk] is now known as bjf | ||
GrueMaster | asac: In answer to your question re:flash-kernel, I believe it is the postinst script that triggers it. I know it runs whenever I install a test kernel. | 16:56 |
asac | yeah i know its always run. just wanted confirm its not using a trigger | 17:03 |
asac | thanks | 17:03 |
=== hrw is now known as hrw|gone|lt | ||
GrueMaster | asac: actually, it might be a triggered event. I looked at /etc/kernel-img.conf and it lists postinst_hook=flash-kernel. I also don't remember flash-kernel requiring any parameters, so this could be it. | 17:09 |
asac | i dont think its triggered because it didnt get triggered when installing a custom kernel .deb | 17:09 |
asac | from rcn | 17:09 |
GrueMaster | Hmmm. The custom test kernel packages I always get from clooney or ericm always initiated flash-kernel somehow. I admit I don't know enough about the deb system to fully understand all the mechanisms used. | 17:11 |
asac | right. i think they use the "official debian" packaging | 17:17 |
asac | while rcn's lucid kernels do something else (supposed to be used without initram i guess) | 17:17 |
GrueMaster | ah, fun. | 17:18 |
=== bjf is now known as bjf[afk] | ||
armin76 | asac: i am getting a panda too, right? :D | 20:13 |
=== bjf[afk] is now known as bjf |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!