freemangordon | Wizzup: again, we shall find what is really broke, I doubt this is the only problematinc package | 08:28 |
---|---|---|
freemangordon | weird, in VM apt log I see no triggers for shared-mime-info | 08:58 |
freemangordon | Wizzup: is it possible to create some quick-n-dirty image in the builder that installs hildon-update-category database only | 09:12 |
freemangordon | hmm | 09:15 |
freemangordon | maybe my trigger is wrong | 09:15 |
freemangordon | ugh | 09:22 |
freemangordon | what happens is that update-mime-database is called for empty /usr/share/mime/packages directory it seems | 09:24 |
freemangordon | lemme test something | 09:24 |
freemangordon | Wizzup: what I guess happens is that shared-mime-info triggers are triggered before packages is being fully configured | 09:49 |
freemangordon | no idea why that happens, but this is what I suspect | 09:49 |
freemangordon | parazyd: any idea? ^^^ | 09:50 |
freemangordon | Wizzup: please respin image build when you are around | 10:09 |
parazyd | freemangordon: I don't know how the triggers work nor what the ones you linked are supposed to do. | 10:12 |
freemangordon | parazyd: yeah, ok, docs a bit vague | 10:14 |
freemangordon | I suspect that shared-mime-info postinst is being called before backage is being fully expanded (or whatever the term) | 10:15 |
freemangordon | *package | 10:15 |
freemangordon | that seems to happen in the builder only | 10:17 |
parazyd | aha | 10:18 |
freemangordon | maybe apt is called with some special parameters? | 10:18 |
parazyd | There's a possibility that the chroot divert is blocking this from happening. | 10:18 |
freemangordon | blocking? | 10:19 |
parazyd | https://github.com/parazyd/libdevuansdk/blob/master/zlibs/helpers#L165 | 10:19 |
parazyd | But no though | 10:19 |
parazyd | This is just invoke-rc.d | 10:19 |
freemangordon | mhm | 10:19 |
freemangordon | it is rather a bug in dpkg, invoking postinst for triggers before package is expanded | 10:20 |
freemangordon | that's what I suspect | 10:20 |
parazyd | apt is also always called with DEBIAN_FRONTEND=noninteractive | 10:20 |
parazyd | I can't think of anything else | 10:20 |
freemangordon | maybe we can workaround that by having shared-mime-info being installed in an earlier stage | 10:21 |
freemangordon | is that possible? | 10:21 |
freemangordon | though I now changed hildon-update-category-database to have pre-depends on shared-mime-info | 10:22 |
freemangordon | hoping that shared-mime-info will be configured by the time hildon-update-category-database install starts | 10:22 |
freemangordon | parazyd: BTW, do I have login on jenkins? | 10:23 |
freemangordon | in the UI that is | 10:23 |
parazyd | I believe so. Let me check. Should I make you one if you don't? | 10:23 |
freemangordon | don;t know, what I can do from there? | 10:24 |
freemangordon | can I run image builds? | 10:24 |
parazyd | Yes | 10:24 |
parazyd | We can put the pkg here: https://github.com/maemo-leste/image-builder/blob/master/beowulf.config#L32 | 10:24 |
parazyd | I'm not sure of the implications | 10:24 |
freemangordon | there shouldn't be any, iiuc | 10:25 |
freemangordon | lets first try with my latest changes, if that doesn;t help we can try and move | 10:25 |
parazyd | ok | 10:25 |
freemangordon | ok, so please either create account for me or start leste-image-droid3-dev image build | 10:26 |
mighty17[m] | freemangordon: does glmark2-es2-*/glxgears run for you in the pvr mesa? | 10:27 |
freemangordon | never tried glxmark, but glmark2-es2 runs just fine | 10:28 |
freemangordon | both drm and x11 ones | 10:28 |
freemangordon | not sure if I tried WL one | 10:28 |
freemangordon | there is no point trying glxmark IMO, it will run through llvmpipe if it runs at all | 10:28 |
parazyd | ok, I triggered it | 10:29 |
parazyd | You don't have an account btw. | 10:30 |
parazyd | pm'd | 10:30 |
freemangordon | ok, if this build is still broken, I'll ping you to make the change in the image builder scripts | 10:38 |
freemangordon | thanks! | 10:38 |
freemangordon | a side note - I think we shall enable openrc parallel boot | 10:39 |
parazyd | mhm I think that's safe now. | 10:40 |
freemangordon | mhm | 10:40 |
freemangordon | who shall do that? | 10:41 |
parazyd | It's easy enough to use sed from image-builder, but I'm not sure if Wizzup wants to handle this from a package. | 10:41 |
freemangordon | lets wait for him then | 10:42 |
freemangordon | hmm, what is wrong with the word "master" so jenkins is moving away from it?!? | 10:47 |
parazyd | Yeah... | 10:54 |
freemangordon | no, I really don;t understand. I mean white/black lists, good, someone may feel offended (though, maybe we shall remove the colors themselves then :) ), but "master"? | 10:56 |
freemangordon | unless it has some meaning unknown to me | 10:56 |
freemangordon | maybe some native speaker can explain | 10:57 |
freemangordon | I mean - do we still have slavery so a slave can feel offended? | 10:57 |
freemangordon | parazyd: it seems shared-mime-info is already installed in stage 2(iiuc) it seems it is pulled as a dependency | 10:59 |
freemangordon | no, it is "thirdstage" | 11:00 |
freemangordon | hmm, is that normal: "scandir: Value too large for defined data type" | 11:01 |
parazyd | 11:01 <freemangordon> hmm, is that normal: "scandir: Value too large for defined data type" | 11:21 |
parazyd | Where does this appear? | 11:21 |
parazyd | Regarding shared-mime-info, I think it's not part of the usual debootstrap | 11:22 |
parazyd | What I could do, is add an array that we can fill with package names and have them installed during the initial debootstrap. | 11:22 |
parazyd | That could help with a few specific ones | 11:22 |
freemangordon | parazyd: here https://phoenix.maemo.org/view/Images/job/leste-image-droid3-dev/6/console | 11:23 |
freemangordon | maybe some openrc warning | 11:24 |
parazyd | hmm I haven't seen that before | 11:24 |
freemangordon | hmm, don;t see that in the VM image build | 11:25 |
wunderwungiel[m] | Hej | 11:51 |
wunderwungiel[m] | Hello | 11:51 |
freemangordon | parazyd: still broken :( | 11:54 |
wunderwungiel[m] | Is Matrix or IRC chat more active? | 11:55 |
wunderwungiel[m] | ? | 11:56 |
Wizzup | we don't have matrix | 12:09 |
Wizzup | if it exists, it's bridged to irc | 12:09 |
freemangordon | Wizzup: no matter what I do, droid3 images are broken re mime | 12:11 |
freemangordon | it looks like there is no /usr/share/mime/packages/freedesktop.org.xml during image build | 12:12 |
freemangordon | or, /usr is bindmounted somewhere else | 12:12 |
freemangordon | like, if you move /usr/share/mime/packages/freedesktop.org.xml somewhere else and execute update-mime-database, you will get exactly the same /usr/share/mime/ content as on the build image | 12:13 |
freemangordon | I don;t see how libhildonmime can possibly break this | 12:13 |
Wizzup | this is not really my terrain | 12:15 |
Wizzup | it sounds like arm-sdk is perhaps just broken then | 12:15 |
freemangordon | hmm, what about this: Err:1 https://maedevu.maemo.org/leste beowulf InRelease | 12:16 |
freemangordon | at the end of the image build | 12:17 |
freemangordon | Temporary failure resolving 'maedevu.maemo.org' | 12:17 |
freemangordon | could it be dns change breaking that? | 12:17 |
Wizzup | yes, but it cannot cause this | 12:19 |
Wizzup | it is part of this: | 12:19 |
Wizzup | apt-get update | 12:19 |
Wizzup | apt-get clean | 12:19 |
Wizzup | updatedb | 12:19 |
Wizzup | update-ca-certificates | 12:19 |
Wizzup | updatedb is just locate related | 12:19 |
Wizzup | ca certs clearly aren't causing the mime problems | 12:19 |
freemangordon | sure | 12:19 |
Wizzup | and it's the last step | 12:19 |
freemangordon | we have one more error though | 12:20 |
Wizzup | I was just inclined to blame libhildonmime because to my knowledge we were seeing ok images before | 12:20 |
freemangordon | "scandir: Value too large for defined data type" | 12:20 |
Wizzup | let me see | 12:20 |
Wizzup | hm | 12:20 |
Wizzup | seems to be openrc/sysvinit related | 12:21 |
freemangordon | mhm | 12:21 |
Wizzup | I think those should be unrelated, but we can go back to images that did work and look at the log | 12:22 |
freemangordon | Wizzup: how hard is to create a simple image that installs hildon-update-category-database (and deps) only? | 12:22 |
Wizzup | I don't know, parazyd would know | 12:22 |
freemangordon | as now it takes 2 hours to test changes | 12:22 |
freemangordon | Wizzup: so, only -dev images are broken, rifght? | 12:23 |
Wizzup | no | 12:23 |
freemangordon | *rigth | 12:23 |
Wizzup | everything is broken | 12:23 |
Wizzup | except for apparently your vm image | 12:23 |
freemangordon | VM is not | 12:23 |
Wizzup | n900 is also broken | 12:23 |
freemangordon | no, VM image is not | 12:23 |
freemangordon | also, the new hildon-update-category-database is only in -devel | 12:24 |
freemangordon | so it cannot affect stable images | 12:24 |
Wizzup | last non-devel change is 2021-09-30 it seems | 12:24 |
freemangordon | int the vm image it is the old version | 12:25 |
freemangordon | 3.x | 12:25 |
Wizzup | 3.1.0? | 12:25 |
freemangordon | mhm | 12:25 |
Wizzup | I mean I can verify and boot the n900 image but someone already did that and reported purple and black | 12:25 |
Wizzup | so I am not sure if that makes sense | 12:25 |
freemangordon | no | 12:25 |
Wizzup | and we don't have -devel n900 images | 12:25 |
freemangordon | but, does it install from beowulf-devel? | 12:26 |
Wizzup | no | 12:26 |
Wizzup | of course not | 12:26 |
Wizzup | only -devel images fetch from beowulf-devel | 12:26 |
freemangordon | unless I miss something, 4.x of libhildonmime is in -devel only | 12:26 |
Wizzup | yes | 12:27 |
Wizzup | so something else perhaps broke it | 12:27 |
freemangordon | no way then this to break shared-mime-info | 12:27 |
Wizzup | we'd have to go back in time to see which is the last good image | 12:27 |
freemangordon | yeah | 12:27 |
freemangordon | VM one is still good :) | 12:27 |
Wizzup | but I'm at a loss here and it's been years since I managed to even run arm-sdk locally | 12:27 |
Wizzup | so I'm hoping parazyd can fix this | 12:27 |
freemangordon | and it doesn;t have "scandir: Value too large for defined data type" error | 12:27 |
Wizzup | https://bugzilla.redhat.com/show_bug.cgi?id=141167#c33 | 12:29 |
Wizzup | not super related | 12:29 |
freemangordon | n900 image log: Setting up hildon-update-category-database (3.1.0+2m7.1) ... | 12:29 |
Wizzup | could be the backing filesystem or something | 12:29 |
Wizzup | freemangordon: right no surprise there | 12:29 |
Wizzup | parazyd: are you available to look at this? we want to give an image to a lima dev who said he's wanting to boot it on a pinephone and try to fix our problem or give hints | 12:30 |
parazyd | Not today, sorry | 12:30 |
Wizzup | :( | 12:30 |
freemangordon | well, I can try and run armsdk here, but, which repo to clone? | 12:31 |
freemangordon | or it hasn;t moved recently? | 12:31 |
parazyd | git clone https://github.com/maemo-leste/image-builder --recursive | 12:31 |
parazyd | Run it in a VM where an account has nopasswd sudo | 12:32 |
freemangordon | leste VM? | 12:32 |
parazyd | Just Devuan or something | 12:32 |
freemangordon | leste is devuan, no :) | 12:33 |
parazyd | Sure :D | 12:33 |
freemangordon | Wizzup: maybe initial bootstrap is > 2GB | 12:35 |
Wizzup | sure, but what modern fs would not support that | 12:36 |
freemangordon | some 32 bit one? | 12:36 |
freemangordon | :) | 12:36 |
freemangordon | and that would explain why 64bit VM doesn;t have the issue | 12:36 |
freemangordon | lemme see what is the situation on PP | 12:36 |
Wizzup | https://phoenix.maemo.org/job/leste-image-pinephone-dev | 12:39 |
Wizzup | no scandir error | 12:39 |
freemangordon | uncompressing | 12:39 |
Wizzup | ok | 12:39 |
freemangordon | mhm | 12:40 |
freemangordon | /usr/share/mime is fine | 12:40 |
freemangordon | Unpacking hildon-update-category-database (4.0.0+2m7) | 12:41 |
freemangordon | so, it is something on 32bits | 12:41 |
freemangordon | not related to libhildonmime | 12:41 |
freemangordon | I am going to revert changes in it and force-push | 12:42 |
freemangordon | ok? | 12:42 |
freemangordon | Wizzup: ^^^ | 12:42 |
mighty17[m] | `ls /sys/bus/mmc/drivers/mmcblk | 12:42 |
mighty17[m] | bind mmc0:1117 mmc1:0001 uevent unbind` | 12:42 |
mighty17[m] | any idea why mmc4:001 is not in here? | 12:43 |
mighty17[m] | `samsung-espresso3g:/sys/bus/platform/drivers/omap_hsmmc# ls | 12:43 |
mighty17[m] | 4809c000.mmc 480b4000.mmc 480d5000.mmc bind uevent unbind` we have it here | 12:43 |
mighty17[m] | `480d5000.mmc` but no mention of it in dmesg or anywhere | 12:43 |
Wizzup | freemangordon: revert what | 12:43 |
mighty17[m] | only happens on a cold boot, any clues? | 12:44 |
Wizzup | freemangordon: ok, maybe something else is up then, I can try to think about that for a moment | 12:45 |
freemangordon | revert changes I made in libhildonmime | 12:45 |
freemangordon | Wizzup: iiuc, it is some cpio archive, maybe it cannot be more than 2GiB on 32 b its | 12:46 |
freemangordon | Wizzup: so, are you ok with me reverting libhildonmime changes and overwriting the history? | 12:48 |
Wizzup | freemangordon: sure I guess, but I wonder if triggers should really return non-0, but idk | 12:48 |
freemangordon | ok, I'll leave it as it is then | 12:49 |
freemangordon | I'll just pull those changes in master | 12:49 |
Wizzup | parazyd: where is this sysvinit -> openrc transition in the arm-sdk | 12:50 |
Wizzup | parazyd: maybe this is part of the openrc package? | 12:51 |
Wizzup | freemangordon: I don't think cpio is used at this stage, but it's hard to confirm | 12:52 |
Wizzup | anything I grep for I can't find it | 12:53 |
Wizzup | it loads something called zuper or something | 12:53 |
freemangordon | mhm | 12:54 |
Wizzup | seems to be random zsh stuff | 12:54 |
Wizzup | I can't find the entry point lol | 12:55 |
Wizzup | build_image_dist doesn't even occur anywhere | 12:55 |
Wizzup | maybe more gets fetched on demand? | 12:56 |
freemangordon | install dependencies atm, need some time :) | 12:56 |
Wizzup | just grepping for cpio also shows just an env var | 12:56 |
Wizzup | magic | 12:56 |
Wizzup | same for fsck and anything else basically | 12:57 |
Wizzup | mkfs just doesn't occur | 12:57 |
Wizzup | parazyd: where is this missing code? | 12:57 |
Wizzup | lol ok I'll wait for you fmg | 13:00 |
Wizzup | I can't even figure out what kind of fs is being used | 13:00 |
Wizzup | bind mount or otherwise | 13:00 |
freemangordon | iirc it was loop | 13:01 |
Wizzup | in any case I am pretty sure the cpio stuff is not being used and we can disable it as well | 13:01 |
Wizzup | idk it's nowhere to be found | 13:01 |
Wizzup | I think all the magic is in some libdevuansdk | 13:03 |
Wizzup | but it's not checked out | 13:03 |
Wizzup | yeah there it is | 13:03 |
freemangordon | zsh: command not found: build_image_dist | 13:05 |
freemangordon | :( | 13:05 |
Wizzup | the usage of this is a bit awkward | 13:06 |
Wizzup | did you load the sdk stuff | 13:06 |
Wizzup | check the readme | 13:06 |
Wizzup | you need to source sdk and stuff | 13:06 |
freemangordon | yes | 13:07 |
freemangordon | I did | 13:07 |
Wizzup | maybe try build_ + tab tab | 13:07 |
freemangordon | ah :_ | 13:08 |
freemangordon | :) | 13:08 |
Wizzup | do we not allow root login on our new images? | 13:54 |
Wizzup | toor doesn't work over ssh | 13:55 |
Wizzup | clearly everything must stand in our way of debugging atm :) | 13:55 |
Wizzup | even changing password doesn't work | 13:56 |
Wizzup | does anyone recall wth we have to do to get access? :D | 13:56 |
freemangordon | user:user | 13:57 |
freemangordon | I found it the hard way :) | 13:57 |
Wizzup | :( | 13:57 |
Wizzup | ok, at least new mesa is on the pinephone image | 13:58 |
freemangordon | ugh, build-dep results in 509 MB | 14:04 |
Wizzup | for? | 14:04 |
freemangordon | shared-mime-info | 14:04 |
Wizzup | mhm | 14:06 |
freemangordon | hmm, I should have disabled manual pages | 14:06 |
freemangordon | the one I built fails to generate mime cache too | 14:15 |
Wizzup | what does strace say? | 14:17 |
freemangordon | I cannot strace in chroot | 14:18 |
Wizzup | I think you can disable that via sysctl | 14:18 |
freemangordon | I will rather put printfs() | 14:18 |
Wizzup | k | 14:19 |
freemangordon | Wizzup: well, if you provide what to set via sysctl, I am finr | 14:20 |
mighty17[m] | tmlind: https://github.com/Unlegacy-Android/android_kernel_ti_omap4/commit/e01587a794fa2ee14d3559a7d919af7e386a03e4 is this commit needed for mainline? | 14:20 |
Wizzup | freemangordon: try sysctl -w kernel.yama.ptrace_scope=0 | 14:20 |
wunderw | Hello :) | 14:20 |
wunderw | I'm writing from N900 | 14:20 |
* mighty17[m] tries to figure out why sdio doesnt get loaded on cold boot and gets when booted from twrp :((( | 14:20 | |
Wizzup | wunderw: confirmed the issue you saw on the new image is a broader problem we're fixing now | 14:20 |
wunderw | Good :) | 14:21 |
wunderw | Installed IRC plugin for conversations and now I enjoy IRC on Fremantle :D | 14:21 |
Wizzup | yeah, we'll have that in leste eventually soon | 14:21 |
Wizzup | (well, deadline is 1 month from now) | 14:21 |
Wizzup | ;) | 14:21 |
freemangordon | /usr/bin/strace: ptrace(PTRACE_TRACEME, ...): Function not implemented | 14:22 |
Wizzup | freemangordon: is /proc mounted | 14:22 |
freemangordon | no :( | 14:23 |
Wizzup | in chroot: mount -t proc none /proc | 14:23 |
Wizzup | or use --rbind from outside chroot | 14:23 |
wunderw | Is this chat only for Leste or mainly for N900? | 14:23 |
freemangordon | didn;t help | 14:23 |
Wizzup | wunderw: not sure what you mean but for leste | 14:23 |
wunderw | Ok | 14:23 |
Wizzup | freemangordon: ok, then I don't really know why | 14:23 |
Wizzup | maybe qemu related or something | 14:24 |
freemangordon | will continue with prints | 14:24 |
freemangordon | mhm | 14:24 |
Wizzup | do you strace from within chroot? | 14:24 |
freemangordon | yes | 14:24 |
freemangordon | you mean I shall strace from outside? | 14:24 |
Wizzup | no, I don't think that makes sense | 14:24 |
freemangordon | Wizzup: readdir returns null when trying to find the files in /usr/share/mime/packages | 14:36 |
freemangordon | ls works on that | 14:36 |
freemangordon | I am lost | 14:36 |
freemangordon | ugh, errno is 75 | 14:38 |
freemangordon | OS error code 75: Value too large for defined data type | 14:38 |
freemangordon | hehe | 14:38 |
freemangordon | https://github.com/rust-lang/cargo/issues/8719 | 14:40 |
freemangordon | https://bugs.launchpad.net/qemu/+bug/1805913 | 14:40 |
freemangordon | "The bug also affects shared-mime-info. update-mime-database uses readdir " | 14:41 |
freemangordon | :D | 14:41 |
freemangordon | https://lkml.org/lkml/2018/12/27/155 | 14:43 |
Wizzup | freemangordon: hm... so qemu bug? | 14:46 |
freemangordon | not really | 14:46 |
Wizzup | well one suggestion seems to be to have a 32bit host | 14:47 |
Wizzup | I can spin up another vm now for that | 14:47 |
Wizzup | if parazyd can set that up that is | 14:47 |
freemangordon | mhm | 14:47 |
freemangordon | we need 32bits VM | 14:47 |
Wizzup | parazyd: what do you think ^^ | 14:47 |
Wizzup | parazyd: shall I set up https://files.devuan.org/devuan_beowulf/installer-iso/devuan_beowulf_3.1.1_i386_netinstall.iso for you/ | 14:48 |
Wizzup | freemangordon: btw this is not fixed in newer kernel? or ? | 14:48 |
freemangordon | still reading LKML | 14:49 |
tmlind | mighty17[m]: no that's not needed, see the recent mainline wl1251 patches on how the mmc core can handle non-standard devices | 14:50 |
tmlind | see commit 187b164945c4 ("mmc: core: transplant ti,wl1251 quirks from to be retired omap_hsmmc") | 14:51 |
mighty17[m] | Ohk will look, thanks | 15:00 |
mighty17[m] | tmlind: that only works for wl1251? what about other chips? | 15:08 |
mighty17[m] | also doesnt explain mmc not working from cold boot :( | 15:08 |
Wizzup | freemangordon: lmk if you have another suggestion | 15:09 |
Wizzup | otherwise I will do the netinstall for 32 bit | 15:09 |
freemangordon | still reading through bugreports | 15:11 |
freemangordon | we're not the only ones to be bitten by that | 15:12 |
freemangordon | Wizzup: hmm, seems it might not help | 15:18 |
freemangordon | but better try it | 15:19 |
freemangordon | (32 bits on 32 bits) | 15:19 |
tmlind | mighty17[m]: you can add support for other chips | 15:23 |
freemangordon | Wizzup: I don't think we have a solution but to use 32 bits | 15:27 |
Wizzup | freemangordon: ok | 15:38 |
mighty17[m] | tmlind: not merged yet? also from where do we get `card->*` info? | 15:51 |
Wizzup | freemangordon: setting it up now | 16:06 |
freemangordon | ok | 16:46 |
mighty17[m] | <mighty17[m]> "tmlind: not merged yet? also..." <- anyways this is a quirk for wl1251 only? as i've never seen it for bcm4330, plus we're moving to sdhci so it should be gone as well? im pretty confused and overwhelmed xD | 16:49 |
Wizzup | freemangordon: the vm is all set up waiting for parazyd to set up jenkins | 17:57 |
Wizzup | and then we should be good to go | 17:57 |
freemangordon | Wizzup: does it make sense to clone armsdk there and spin image build? | 18:11 |
freemangordon | armsdk it is not *that* hard once you get used to it :p | 18:12 |
Wizzup | freemangordon: hm? | 18:15 |
Wizzup | freemangordon: I'd rather wait for parazyd to add it to the CI so that we can assign the 32 bit image builder to all 32 bit arches | 18:15 |
Wizzup | s/arches/image builds/ | 18:15 |
Wizzup | unless you want me to try and see if it works ok | 18:15 |
freemangordon | yeah, that was the point. up to you though | 18:16 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!