=== mamarley is now known as Guest69748 === mamarley_ is now known as mamarley [15:23] Hey there [15:24] I've been having issues for ages with sound output and I know that using the kernel 4.2.0-35 from a previous release (even with current userspace) just works [15:24] So I'm trying to bisect the kernel git to find the faulty change [15:24] It appeared somewhere between the release using 4.2 and the one that moved to 4.4 [15:25] My problem is that git bisect tries to test some changes in 4.1 and for some reason, that kernel won't boot (stuck at loading init ram image) [15:26] Do you have any guesses why it would do that? I'm testing random commit manually now since I can just "skip" commits with bisect :( [15:26] And there are like 349 commits left... [15:28] Oh, and basically, my problem is that sound doesn't work properly anymore as in some sound sources will work, some won't. The device seems to be stuck in some kind of exclusive mode since I can't play multiple sounds together [15:28] And since this is an HTPC running Kodi, it's easy to check if a kernel works or not: I don't have sounds in the menus but *some* medias will play sound fine (not all) [16:34] jsalisbury, ^ [16:35] apw, ack. [16:35] Orphis, do you happen to have a bug opened in Launchpad for this issue? I could help you bisect and build kernels for you to test. [16:35] jsalisbury: I haven't, I hoped to find a similar one, but it looks like I can't [16:36] So I figured a single user haveing an issue with maybe some hardware that isn't that common isn't going to help resolving it [16:37] Orphis, we can try. you can open a bug by running the following in a terminal: "ubnutu-bug linux" or you can do it thoght the web interface [16:37] I bisecting right now, but since I couldn't boot 4.1 kernels apparently, I just picked a random 4.2 version from the git bisect visualize list, tried it (it works) and now I have more revisions to test :-/ [16:37] I guess I should do it from a bad kernel [16:38] The problem is that it just takes forever to rebuild the vanilla kernel with the ubuntu options [16:39] Orphis, it's good that 4.2 is good, you can now work forward until you find the first bad version. [16:39] Orphis, we have pre-built mainline kernels. It might be worth testing these ones: [16:39] It just weird, git describe shows 4.1rc and 4.2rc changes [16:39] From what bisect is trying to test [16:39] 4.2 final: http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.2-wily/ [16:40] and 4.3 final: http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.3-wily/ [16:40] Orphis, if 4.3 final has the bug, you could try some of the 4.3 release candidates from the same page: [16:40] http://kernel.ubuntu.com/~kernel-ppa/mainline/ [16:40] What's the commit granularity? [16:40] Orphis, once those are narrowed down, I can build kernels for you [16:41] I can build kernels too, the HTPC is a 3 year old i5, it's quite decent :) [16:41] Orphis, between 4.2 final and 4.3 there are like 13K commits [16:42] Though, it's no 24 core machine :P [16:42] jsalisbury: Do you know a way to automatically trim down unused modules from my kernel config based on what I'm currently using? [16:42] That would cut down the build times considerably probably [16:43] Orphis, you could run dmesg and lspci and see what's being used and remove what's not being used. That might run the risk of other errors though. [16:44] Yeah, I know... Not easy [16:44] Or consistent [16:46] I'll try this revision that bisect suggest then try your kernels [16:46] Still compiling so it will take a bit [17:09] jsalisbury: Should I use the 4.3 mainline or skip to rc1 directly? [17:14] jsalisbury: I can provide you with a bisect log if you are interested [17:14] https://pastebin.com/cYThEiag [17:15] Orphis, maybe try -rc1 to start with. If it is good, try rc2 or later like rc3/4 [17:53] jsalisbury: So, 4.3-rc1 doesn't work [17:54] Orphis, ok, it looks like we shoudl bisect between 4.2 and 4.3-rc1 [17:54] Yup, that's what I've been doing :p [17:54] Orphis, it might also be worth while to test current mainline to see if the bug was already fixed [17:55] Well, it's not fixed in 4.10, from latest release [17:55] Orphis, that's available here: [17:55] http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.12-rc2/ [17:55] I can try that too though [17:55] Orphis, ack. If it is fixed in 4.12-rc1, it's probably best to perform a "Reverse" bisect to identify the fix [17:56] Well, if it's fixed in 4.12, I'll just use that until Ubuntu releases a new version I guess [17:58] So much faster to download a kernel than building it... [17:58] Nop, doesn't work [18:16] Does Ubuntu 16.04.0 GA kernel ensures stable kabi for its lifetime? [18:36] Mehul: quite the opposite [18:48] jsalisbury: How should I test kernels when it's stuck at "Loading initial ramdisk..."? Is there any way to get debug output on screen for this step? [18:53] Even though I edit the boot parameters and remove quiet and splash, I still don't see anything :-/ [19:10] Orphis, yeah, those can be difficult to get around. sometimes you have to just "git bisect skip" or review commits to see if there is a fix in later versions [19:10] I have no idea of what's going on at all [19:10] I can't get any output [19:27] Orphis, I can build you a test kernel if you send your latest bisect log [19:28] jsalisbury: https://pastebin.com/NCScMejM [19:29] I'm trying commits manually now since skip only stays in the 4.1 branch [19:29] Well, not really 4.1 but that's what describe says [19:30] The "git bisect visualize" output in gitk isn't easy to grasp, a lot of tangled merges [19:30] I'm thinking it could be an issue with gcc-6 that I have on my system or something similar [19:31] Compiling v4.2-rc7-648-g3be66711b6 now to eliminate a few changes... [21:33] b ubuntu-kernel