=== doko_ is now known as doko [00:45] Ok, to save me some embarassment later on, how explicately can I publish my branch to kernel.u.c? (I have an account on zinc ...) [05:19] morning osmosis [05:19] mornin [05:22] how goes it osmosis [05:23] im hating 3ware raid cards. Im going to switch to linux software raid. trying to figure out a migration plan [05:24] osmosis, ouch. I'm learning that the PowerPC kernel blows up if you add too many platforms at once ;-) === asac_ is now known as asac [11:02] hi, it is a bit short for final 8.10 release,but there is already 2.6.27.4 out [11:05] Kano, Which patches are there that aren't in .6.27-7.14 ? [11:06] it is based on 2.6.27.2 when you look at the git log [11:06] so interdiff between 2 and 4 [11:06] 7.14 did only disable ath5k [11:08] Let me ask that differently. Which patch of those is release-critical for Ubuntu 8.10? [11:08] you can update it later if you like [11:08] some deadlock fixes are there however [14:20] Is it currently possible to cross-compile the kernel? [14:20] NCommander, Yes, but it's generally only done for bootstrapping purposes. [14:21] persia, well, I want to provide daily images of the ports kernel, and since PPAs are not really that useful for -ports, I'm looking towards cross-compiliation for the dailies unless someone can cough up hardware [14:22] You *don't* want to do that. cross-compiled kernels tend to be fragile. [14:22] persia, not in my experience, I cross-compiled kernels for both my ARM and MIPS boxes [14:23] Then you have different experience than I, although I've not cross-compiled a kernel in a 3-4 years, so my experience may be out of date. [14:24] persia, its possible, but usually the biggest cause of a DOA kernel due to cross-compiling is a faulty toolchain [14:24] Oh, mine weren't DOA, just not as fast as the native ones. [14:24] Odd [14:25] persia, it would just be nice to make sure the new ports kernel actually builds on all architectures without having to manually kick off builds [14:25] Port xen :) [14:26] I could do that for PowerPC [14:26] Xen for PowerPC already exists, just not for G4s [14:26] With enough lobbying, you might be able to get LP to support ppc PPAs then. [14:26] Someone told me awhile back Canonicals PPC buildds are G4 XServes [14:27] oh, xen-powerpc requires 64-bit hardwares upport [14:27] Ah. Well, you probably have to support that then. [14:28] yeah [14:52] ciao a tutti [14:52] hi [14:52] is there any italian? [14:53] marlock, there is an ia64 port of Ubuntu if thats what your asking [14:56] !it [14:56] ack stupid lack of bots [14:56] I give up [14:56] morning jdong [14:56] Well, there are italian kernel hackers in the channel sometimes, just not now. [14:56] morning :) [14:56] jdong, how goes it? [14:57] NCommander: meh hanging in there, catching up on homework for next week [14:57] jdong, do you happen to use any of the Ubuntu port architectures? [14:57] no, I'm not cool enough to have outda^H^H^H^Hexotic hardware [14:58] * NCommander watches PowerPC, SPARC, and IA-64 users descend on jdong [14:58] all 5 of them? [14:58] actually powerpc has about 3,000 users [14:59] I know, I'm just messing with you :) [14:59] * NCommander wished that the ports got more spick and span so to speak then they currently do [20:32] hey BenC === thunderstruck is now known as gnomefreak === thunderstruck is now known as gnomefreak [22:47] * NCommander has discovered git cherry-pick is his new best friend [22:52] NCommander, Hey. You might be interested in this one. I was waiting to see if there would be a 7.15 kernel, to upload linux-rt, but given he timing, I think 7.14 will be the final. When I recompiled linux-rt against 7.14 (rather than 7.12), it failed ABI check. [22:52] 7.15? [22:53] I assume thats the version of the rt patches [22:53] No, the main kernel. 2.6.27-7.14 is the current version (last I checked). I heard rumours of a possibly newer one, but it doesn't appear to have been uploaded. [22:54] Oh! [22:54] linux-rt is 2.6.27-3.7 (fewer uploads means fewer ABI changes). [22:54] Right, the X.Y version refers to X is the ABI, Y is the upload [22:54] I didn't expect an ABI change between 7.12 and 7.14, but maybe it's related to the patches in -rt. [22:55] persia, this close to release, breaking the ABI is to be avoided if possible. If -rt is broke the API, we need to transition linux-rt-lrm and lbm (I dunno if it has either one of those) [22:55] Just lrm, but that's easy enough. [22:56] Probably is easier to just transition lrm and be done with it vs. trying to unbreak the rt ABI (if possible) [22:56] On the other hand, I'd rather not break ABI. Since the -rt code is unchanged, and it's only the 7.12 -> 7.14 updates, I'm not entirely sure why it fails ABI check. [22:56] Or rather, wonder if the main kernel didn't break ABI. [22:57] Can you post the output of the failed ABI check? [22:58] http://paste.ubuntu.com/63049/ [22:59] Hrm [23:00] persia, given its just xfs, its probably safe to ignore it (I can't see anything in lrm building against the XFS API) [23:00] persia, however, I won't say I'm familar with it to be sure, so I recommand you ask BenC, or another kernel guru [23:02] persia, but that would be very strange if the rt patchset changed the XFS API [23:02] * NCommander would be wondering more whats going on [23:02] Oh [23:02] persia, http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-intrepid.git;a=commit;h=21c69ea60244403e503f148cd29d89df85eb0908 [23:02] persia, your tree might be out of date [23:03] I'm not using a git tree. linux-rt build-depends on linux-source-2.6.27 [23:03] * NCommander grabs linux-rt [23:03] The XFS ABI was blacklisted [23:04] Ah, so maybe a change to config fixes it? [23:04] persia, what source package is linux-rt? (I get linux-rt-meta if I try to apt-get source it) [23:05] * NCommander doesn't want to give you bad instructions [23:06] Yeah, well, the source/binary naming for the kernels is way broken. Try apt-get source linux. The source package is called linux-rt. You need to apt-get source linux-image-2.6.27-3-rt to get it. [23:06] * persia filed a bug about this a while back. Some people have ideas they plan to bring to UDS [23:06] persia, I can give you a debdiff for it, or just simply teach you how to do the blacklist [23:07] Up to you, although I suspect you'll find that the -rt kernel source is different from other kernel sources. [23:07] persia, it uses the same build system [23:07] Indeed it does. [23:08] Is it more than CONFIG_XFS_FS=m -> CONFIG_XFS_FS=n ? [23:09] Err "# CONFIG_XFS_FS is not set" [23:09] persia, no, you need to tell the ABI check to ignore XFS [23:09] You have to add it to the perm-blacklist file [23:09] * NCommander is test building and test installing the result now [23:13] persia, http://paste.ubuntu.com/63053/ [23:14] I'm a little confused. Does this just tell the ABI checker not to check that? [23:14] persia, pretty much [23:16] So the ABI does change, but we're explicitly saying we don't care, because we don't expect anyone to have external modules compiled against XFS? [23:16] BIngo [23:16] Welcome to the world of kernel hacking [23:17] Somehow that seems dishonest, but if that's the standard way these things are done ... [23:19] persia, well, its just copying the change to the mainline kernel [23:19] Yeah, well. That only confirms the latter part. [23:26] persia, its building modules now, so another 20-30 minutes and I can confirm [23:26] NCommander, OK. Once you confirm, I'll get the ACK from MOTU Release, and push. [23:39] * NCommander didn't realize linux-rt was in multiverse [23:42] Where else would it be? [23:43] Actually, ideally it would be in universe, but there's issues with metapackages and virtual packages and the like. [23:47] persia, I thought it was in multiverse due to firmwares [23:48] No. The firmware is in the arch-independent linux-firmware packages, shared between several kernels (you do use this for ports, don't you?) [23:49] persia, we just recently made that change [23:49] persia, all the blobs are removed? [23:50] Should be. Remember, linux-rt is just a patch package, and depends on linux-source. As long as linux-source is clean (and it should be), linux-rt is clean. [23:51] NCommander, When you're done testing, could you stick your debdiff on 289683? [23:52] persia, no problem [23:52] persia, wow, debian should adopt our kernel packages [23:52] * NCommander should post that to d-devel .... [23:52] Why? [23:52] Because of the firmware? [23:53] yeah [23:54] It doesn't solve the essential issue : linux-firmware is in main. It's on the CDs. Putting it in non-free doesn't help anyone : the argument there needs to be resolved in it's own way, but that's orthogonal to the packaging change. [23:54] persia, actually, with debian, linux-firmware would just get dropped into non-free; d-i in lenny already prompts to grab firmware from non-free [23:55] Umm. It's more complicated than that. Anyway, this isn't the right forum. [23:56] persia, right, anyway, I'm finishing compiling linux-rt now (its in the final module link) [23:56] Cool. Now it just needs smoke testing :) [23:56] mmmmmmmmm, smoke [23:57] persia, disregard that debdiff I gave you [23:58] persia, I forgot to move the ABI folder