adam_free2air | i'm having a rpi-img-builder issue with PART_UUID being incorrect(??) building devuan beowulf/testing under latest Ubuntu Focus, the PART_UUID in part-uuid.txt does not correspond to the value of the partition returned by ROOT_PARTUUID=$(blkid -o export -- "/dev/mmcblk0p2" | sed -ne 's/^PARTUUID=//p') run after blowing the SD card. | 16:29 |
---|---|---|
adam_free2air | resulting in builds not finding their root partitions. | 16:30 |
adam_free2air | the UUID code seems pretty well the same for all target OSes, anyone come across this? | 16:33 |
adam_free2air | currently building from commit 2b338973618fe75f9ead0950e553b8ffcaa67db4 Devuan: added rev 1 to disable pi-bluetooth in firstboot script | 16:36 |
adam_free2air | log is here https://ci.free2air.net/job/Images/job/rpi-img-builder/job/rpi-4-devuan-chimaera-5.10.y-arm64/42/consoleFull | 16:37 |
adam_free2air | the log includes *** cat ./userdata.txt config. | 16:39 |
c0rnelius | how many ls /dev/loop* do you have? | 16:40 |
adam_free2air | yes, i noticed the loop devices sometimes remain open. when i looked when it was built, i think about 18 ... | 16:41 |
Tenkawa | or just run sudo losetup | 16:41 |
Tenkawa | you have to ckean those up first | 16:41 |
adam_free2air | i saw that at rebooted the box, although i'm not sure i've rebuilt since. | 16:41 |
Tenkawa | er clean | 16:41 |
c0rnelius | Someone else mentioned this a few days and then closed the issue as they resolved it them selves. They did a full reinstall, but normally a `make cleanup` fixes it. But if it doesn't rebooting should do the trick. | 16:41 |
Tenkawa | lsof/fuser/umount stay by my side when that happens | 16:42 |
Tenkawa | lol | 16:42 |
adam_free2air | i'll get jenkins to rebuild and check again. | 16:42 |
adam_free2air | s/at/and/ | 16:43 |
c0rnelius | 18 loops? Thats probs why it's getting confused. It is suppose to close the loops after the build but sometimes it doesn't. Not sure why that? | 16:44 |
adam_free2air | c0rnelius: using jenkins to checkout & build so super clean build env each time. | 16:44 |
adam_free2air | seems there are 2 issues with jenkins approach and current config. 1 if builds (currently for all rpi's devuan beowulf & testing) fail, it can fill disks and burp in unexpected ways. 2 losetup cleanup may affect builds through cumulative disconnect issues. | 16:48 |
adam_free2air | also, with jenkins, still experiencing a state where i have to rerun the docker hack - this *may* be down to disk or losetup issues? not sure. | 16:49 |
c0rnelius | What is Jenkins exactly? | 16:50 |
adam_free2air | once build, i'll check image and try booting on target rpi4 | 16:50 |
adam_free2air | c0rnelius: https://www.jenkins.io/ a build automator of sorts. | 16:52 |
c0rnelius | Remember if you every have an image or rootfs fail, always run `make cleanup`. | 16:52 |
c0rnelius | ever* | 16:52 |
c0rnelius | I usually run it twice for good measure. | 16:52 |
adam_free2air | c0rnelius: each build under jenkins operates entirely from a clean slate, from git clone onwards in my rpi-img-builder config | 16:53 |
c0rnelius | it in theory "clears the loops" and removes all the old partition directories and misc files that are created during stage1 and stage2. | 16:53 |
adam_free2air | ok. | 16:53 |
adam_free2air | with jenkins building 6 images every evening, this might need a cronjob to check and clear. | 16:55 |
adam_free2air | or an external post-build jenkins cleanup script, i guess. | 16:55 |
adam_free2air | actually, or i invoke a make cleanbuild *before* each build - thanks! | 16:57 |
c0rnelius | Thats what I would do. | 16:57 |
adam_free2air | yup. doh! | 16:57 |
c0rnelius | make run `make cleanup` twice and then start the next build. | 16:57 |
c0rnelius | it run* | 16:58 |
adam_free2air | twice just to make sure? %^\ | 16:58 |
c0rnelius | Doesn't hurt :) | 16:58 |
c0rnelius | losetup is kinda dumb sometimes | 16:58 |
adam_free2air | yeah. i've come across some of its dumbness before ... | 16:59 |
c0rnelius | this is all the script does - https://raw.githubusercontent.com/pyavitz/rpi-img-builder/master/scripts/clean | 17:00 |
c0rnelius | which is what the stage1 script is suppose to do at the end of the build. But sometimes its tarded. | 17:00 |
adam_free2air | OK. First rebuild attempt after losetup screw up and build failed. It failed in the way I recognise the docker hack fixes. Re-run docker hack and rebuilding. | 17:30 |
adam_free2air | fyi: | 17:31 |
adam_free2air | 17:16:20 ls: cannot access './*.img': No such file or directory | 17:31 |
c0rnelius | strange you need to keep doing that? | 17:31 |
adam_free2air | 17:16:20 RAWIMG= | 17:31 |
adam_free2air | 17:16:20 ZIPIMG= | 17:31 |
adam_free2air | 17:16:20 sudo zip | 17:31 |
adam_free2air | 17:16:20 adding: -PK- | 17:31 |
adam_free2air | 17:16:20 �S��������- (deflated 0%) | 17:31 |
adam_free2air | 17:16:20 PKPK- | 17:31 |
adam_free2air | 17:16:20 �S�-PK/MArchiving artifacts | 17:31 |
adam_free2air | 17:16:20 ‘rpi*.zip’ doesn’t match anything | 17:31 |
adam_free2air | 17:16:23 ERROR: Step ‘Archive the artifacts’ failed: No artifacts found that match the file pattern "rpi*.zip,rpi*.img". Configuration error? | 17:31 |
adam_free2air | 17:16:23 Sending e-mails to: adam@dyne.org | 17:31 |
adam_free2air | 17:16:23 Finished: FAILURE | 17:31 |
c0rnelius | do you have all the depends installed? | 17:31 |
c0rnelius | is this cross? | 17:32 |
adam_free2air | c0rnelius: can't explain it yet, but it seems to clear the issue. | 17:32 |
adam_free2air | yes. amd64 running fully updated Ubuntu Focus | 17:32 |
c0rnelius | run `make ccompile` | 17:33 |
c0rnelius | make sure all the depends are there. | 17:33 |
adam_free2air | yes. they're there. normally builds are successful. | 17:33 |
adam_free2air | unless build dependencies for focus were added in the last few commits. | 17:33 |
c0rnelius | I get errors building devuan testing sometimes, but it is normally because of the mirrors. | 17:34 |
adam_free2air | :) yeah, i know those errors too :/ | 17:34 |
adam_free2air | building 6-8 builds nightly, i stumble across most things. | 17:34 |
adam_free2air | just reran make ccompile - yeah, it's all there. | 17:38 |
c0rnelius | I don't know, it's been working fine for me? But in all honesty I don't cross all that much and I don't use automation. | 18:00 |
c0rnelius | My x86 buildbox is almost done building a kernel for my laptop, I'll run a build and see what happens when it finishes. | 18:01 |
adam_free2air | cross? you mean you don't build on amd64 for rpi targets? | 18:01 |
c0rnelius | Not all that often no. | 18:01 |
adam_free2air | ok | 18:01 |
adam_free2air | what's your defualt build env then? | 18:02 |
c0rnelius | Sometimes I'll build kernels, but native is just way faster. I have an RK3399 and Odroid N2+ I use for making most of my images. | 18:02 |
adam_free2air | ok. | 18:02 |
c0rnelius | I then have a custom made armel img for the Pi4 which I use for those builds | 18:03 |
adam_free2air | as you may remember, i tried starting to build all these images on a devuan rpi4 but it failed too much of the time. so i thought the canonical build env was ubuntu focus on amd64, so i switched to that. | 18:04 |
c0rnelius | We also added clang recently as well which is pretty handy as the kernels don't care about gcc and the headers seem to stay clean across releases. | 18:05 |
adam_free2air | i saw that. | 18:05 |
c0rnelius | alright, running a quick test cross - `make 2711 && make rootfs && make image` | 18:10 |
c0rnelius | lets see if I get that stupid multiarch error | 18:10 |
adam_free2air | my re-run after previous burps is still running ... https://ci.free2air.net/job/Images/job/rpi-img-builder/job/rpi-4-devuan-chimaera-5.10.y-arm64/44/console | 18:11 |
c0rnelius | looks like it made past the normal error point on that script | 18:12 |
c0rnelius | Hmmm. I got that dumb multiarch error. | 18:12 |
c0rnelius | There has to be a better way to fix this problem than using docker | 18:13 |
adam_free2air | well, at the moment, at least it fixes it. | 18:15 |
* adam_free2air hates and despises docker,btw. | 18:16 | |
c0rnelius | not a big fan myself | 18:16 |
c0rnelius | I used it for a while when making imgs, but its to much of a pain the ass. | 18:16 |
adam_free2air | it's for lazy people who do not care for provenance. | 18:17 |
c0rnelius | I started using Hirsute on my arm64 buildbox's, but haven't tried it yet on the x86_64. I wonder if the same shit happens with qemu? | 18:19 |
c0rnelius | Know what..? I got spare drives I'm gonna install it and see. | 18:21 |
adam_free2air | OK! | 18:32 |
adam_free2air | i thought it's on the github repo readme that amd64 ubuntu focus was the standard build env. | 18:36 |
c0rnelius | It is as its the LTS and in the past it worked fine. | 18:36 |
c0rnelius | This bug just kinda started out of no where in the last few months. | 18:36 |
adam_free2air | ok | 18:37 |
c0rnelius | I really hate the ubuntu server install :) | 18:39 |
adam_free2air | not an ubuntu fan myself | 18:39 |
adam_free2air | ex-gentoo ricer, avoided arch, and somehow landed in fedora land for now ... | 18:41 |
c0rnelius | I hate Ubuntu its bloated trash mostly, but its supports more compilers. Which is why I reluctantly started using it for the builders. | 18:42 |
adam_free2air | ok. new build looks good. now to test ... | 18:42 |
c0rnelius | Although the custom imgs I made for my two arm64 boxes run really quick on it. But they are super minimal. | 18:43 |
adam_free2air | i make my builds listening to radio isla negra, one of the oldest internet radio stations still online http://66.228.60.216:8002/stream | 18:45 |
c0rnelius | I do like it's cmd line installer interface though. I'd use it instead of Dialog if I knew what it was. | 18:45 |
c0rnelius | Kool, I'll have to add it to my list on my Pi Musicbox | 18:47 |
c0rnelius | normally it just plays random channels I have in Pandora, but I have a few streams I play with mpv on it. | 18:48 |
c0rnelius | added | 18:51 |
c0rnelius | isla(){# isla negra | 18:51 |
c0rnelius | ssh musicbox 'mplayer -nocache -prefer-ipv4 http://66.228.60.216:8002/stream' | 18:51 |
c0rnelius | } | 18:51 |
c0rnelius | ICY Info: StreamTitle='Billie Holiday - Strange Fruit'; :D | 18:52 |
adam_free2air | ICY info: Stream Title="Karen O;Michael Kiwanuka - YO MY SAINT" | 19:18 |
c0rnelius | that was a pretty good song | 19:24 |
c0rnelius | reminded me of Dengue Fever | 19:24 |
c0rnelius | adam_free2air: Looks like a fresh install of Hirsute passed the qemu error test on x86_64 | 19:27 |
c0rnelius | The fella that posted an issue about Focal made the same comment. That a fresh install fixed his problems on Focal. | 19:28 |
c0rnelius | Not sure what the Ubuntu boys did to mess that up so badly. :\ | 19:29 |
c0rnelius | Its just sitting through the nightmare of installing and setting up Ubuntu server. | 19:31 |
c0rnelius | Seems like once you have a working install, just never do a `sudo apt update && sudo apt upgrade -y` :) | 19:32 |
adam_free2air | "Seems like once you have a working install" - what's that then? :-/ | 19:33 |
Tenkawa | most of the time a upgrade/dist-upgrade -y is never recommended anyway because -y will arbitrarily do things without your consent. the same as -n will | 19:34 |
Tenkawa | it will forcibly remove/add things | 19:35 |
Tenkawa | (for -y and ignore for -n) | 19:36 |
c0rnelius | For sure seems to be an x86 prob. My aarch64 installs never have issues. | 19:36 |
Tenkawa | fortunately there doesn't actually appear to be an ignore actually... so its just the installing you have to worry about. | 19:39 |
Tenkawa | but yeah I've gotten into trouble using -y before | 19:40 |
Tenkawa | added like a couple hundred extra packages | 19:40 |
c0rnelius | I need a faser x86 box... Sucker is to slow. | 19:44 |
c0rnelius | I should go check the college/university trash/recycle bins. You would be shocked at what the idiots throw in there. | 19:45 |
Tenkawa | what speed is it? | 19:46 |
c0rnelius | Its a shit i3. | 19:47 |
Tenkawa | ahh | 19:47 |
c0rnelius | I bet I could find a i7 in a bin with a broken screen that some dumb college kid threw away. | 19:47 |
Tenkawa | its not like you need the screen | 19:47 |
c0rnelius | Nah. My i3 is just sitting on a table. | 19:48 |
c0rnelius | broke gutted laptop | 19:48 |
c0rnelius | well, under the table really. | 19:49 |
c0rnelius | but whatever | 19:49 |
c0rnelius | My main PC i got from MICA college in Baltimore as the IT department was trashing them and I was dating a chick that worked there. | 19:50 |
c0rnelius | So she grabbed me one | 19:50 |
Tenkawa | heh | 19:51 |
c0rnelius | I use to work for Anne Arundel Community College in Maryland and they would do the same shit, just toss everything in the garbage. They actually got written up in the paper for it. Essentially, why aren't you people donating this stuff? | 19:53 |
c0rnelius | So wasteful | 19:53 |
Tenkawa | indeed | 19:53 |
c0rnelius | well looks about right - rpi-4-devuan-testing-5.10.52-v8-ext4-2021-08-07.img.xz | 20:36 |
Tenkawa | still trucking | 20:42 |
Tenkawa | du -sh gcc-11 | 20:42 |
Tenkawa | 12G gcc-11 | 20:42 |
c0rnelius | sudo apt install gcc-11 | 20:47 |
c0rnelius | Reading package lists... Done | 20:47 |
c0rnelius | Building dependency tree... Done | 20:47 |
c0rnelius | Reading state information... Done | 20:47 |
c0rnelius | The following additional packages will be installed: | 20:47 |
c0rnelius | cpp-11 libgcc-11-dev | 20:47 |
c0rnelius | Suggested packages: | 20:47 |
c0rnelius | gcc-11-locales gcc-11-multilib gcc-11-doc | 20:47 |
c0rnelius | The following NEW packages will be installed: | 20:47 |
c0rnelius | cpp-11 gcc-11 libgcc-11-dev | 20:47 |
c0rnelius | The easy way :) | 20:47 |
c0rnelius | I don't know why you make like so hard for ur self. | 20:48 |
c0rnelius | life* | 20:48 |
Tenkawa | c0rnelius: wont work for what needs to be done | 20:48 |
c0rnelius | mhmm | 20:48 |
Tenkawa | that wont fix the arm64 compiling armel problen | 20:48 |
Tenkawa | er problem | 20:48 |
Tenkawa | those pkgs dont have it built in | 20:48 |
Tenkawa | I'm building it special | 20:49 |
c0rnelius | I think a file needs to be hit with sed | 20:49 |
c0rnelius | and remove stuff | 20:49 |
Tenkawa | no... the armel multiarch needs to be "added" | 20:49 |
Tenkawa | so it doesn't try to pickup arm64 automaticly | 20:50 |
c0rnelius | the headers shift patch was created for x86 builds, not for arm64 to armel. I think something like minded needs to be done. But I'm just guessing. | 20:50 |
c0rnelius | Its something inside the scripts dir. | 20:51 |
Tenkawa | that probably needs to be done as well.. but theres no armel build capability at all right now | 20:51 |
Tenkawa | (without using a whole seperate chroot) | 20:52 |
Tenkawa | would be interesting to try just switching the shift patch and see what happens | 20:53 |
c0rnelius | Strange that the kernels install just fine on an actual board though. | 20:53 |
c0rnelius | You would think the error would be the same? | 20:53 |
Tenkawa | yeah kernels are agnostic.... just building them are't | 20:54 |
Tenkawa | er aren't | 20:54 |
c0rnelius | Make me wonder if we installed say qemu on the img its self right before the kernel install inside the chroot, would it compile the headers? | 20:54 |
Tenkawa | I tried | 20:55 |
Tenkawa | same error | 20:55 |
c0rnelius | hmm | 20:55 |
c0rnelius | Well I don't need to try that then :) | 20:55 |
Tenkawa | arm64 doesnt have those headers which makes me agree that the shift patch is needed too | 20:55 |
Tenkawa | or the one that its bombing on | 20:56 |
Tenkawa | (at least in our pkg) | 20:56 |
Tenkawa | it exists in the default | 20:56 |
Tenkawa | it seems to be connected to uapi too which is a pain | 20:57 |
c0rnelius | Well I thinks I deserve a nap today. Be back laters. | 20:58 |
Tenkawa | cheers | 20:58 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!