<owrt-2203-builds> Build [#258](https://buildbot.openwrt.org/openwrt-22.03/images/#builders/40/builds/258) of `mvebu/cortexa53` failed.
danitool has quit [Ping timeout: 480 seconds]
tSYS has quit [Quit: *squeak*]
tSYS has joined #openwrt-devel
grid has joined #openwrt-devel
dangole has quit [Ping timeout: 480 seconds]
<KGB-0> https://tests.reproducible-builds.org/openwrt/openwrt_ath79.html has been updated. (98.6% images and 100.0% packages reproducible in our current test framework.)
<nick[m]1234> Ansuel: I saw that you already tried to fix the gihub_archive download. For me this is broken. If I use in on my host machine and than on a x86 openwrt docker, the hash is just different. I already tried setting different modes. Any suggestions?
rua has quit [Quit: Leaving.]
goliath has joined #openwrt-devel
shoragan has joined #openwrt-devel
bluew has quit [Ping timeout: 480 seconds]
ukleinek has joined #openwrt-devel
ukleinek has quit [Quit: Reconnecting]
ukleinek has joined #openwrt-devel
<Ansuel> nick the hash should not be related to the host
<Ansuel> might be that some mirror have an old tar
<Ansuel> or github changed the tar settings
<\x> Ansuel: you now can update ipq60xx boardfile right?
<\x> regdb update*
<nick[m]1234> Ansuel: yes. I also thought that. However, when trying to bump the olsrd version (or even just select another pkg_source_date), the download fails. You can even see than in the CI, that download is sucessfull for some targets and for some not.
<Ansuel> \x maybe
<Ansuel> nick do you have some link for CI so i can quickly repro it?
<\x> Ansuel: i can try later tonight, ill send you my MR7350 boardfile, but i want regdb master since a friend updated the proper ones for our country recently.
<nick[m]1234> Ansuel: I don't have the old ci values. Here is the PR I'm currently working on: https://github.com/openwrt/routing/pull/977
<Ansuel> \x saddly i won't do regdb mod and i will stick to what qcom provide
<\x> oh, i thought its like you use linux regdb
<nick[m]1234> Ansuel: can you tell me if the PKG_MIRROR_HASH changes when running "make package/olsrd/{clean,download,check} -j1 V=s FIXUP=1" ?
<Ansuel> nick are you sure it's a target thing and not just luck ?
<Ansuel> luck in the sense that a different mirror is selected and that have a different version?
<Ansuel> i see you also bump the release and the version so it should not be a mirror thing
<nick[m]1234> you mean a different github mirror? I thought it is now downloaded via archive api from github
<nick[m]1234> Ansuel: I'm not sure what's the cause of this behavior
<Ansuel> nick do you have one target that fail and one that works?
<Ansuel> so i can test this on my local buildroot?
<nick[m]1234> Ansuel: I thought maybe it has some time race condition whatever. So I looked at the mtime again.
nixuser has joined #openwrt-devel
<Ansuel> from CI i can see every test failed with mismatched hash
<nick[m]1234> Ansuel: yes please test it. I'm very interested if it works at your setup.
<Ansuel> in my local buildroot it works correctly
<Ansuel> repacked git produce the same hash
<nick[m]1234> and FIXUP cmd?
<Ansuel> check FIXUP=1 ?
<nick[m]1234> yes
<nick[m]1234> sha256sum dl/olsrd-2023-06-12-a9b3f1ac.tar.xz d8c80228b4a79fed12d6f967008da1e54ef893e6b252a394597907b2796ef211 dl/olsrd-2023-06-12-a9b3f1ac.tar.xz
<Ansuel> nope same
<Ansuel> wtf???
<Ansuel> b7e1cd4b43a4d8607d6761267f23d28fb09c577d746c22ca60d2ed2fc19b1ea8 is the hash reported by CI
<Ansuel> and match mine
<dwfreed> note that github does not produce reproducible tarballs
<Ansuel> we are not using github api
<Ansuel> the script just clone the repository
<Ansuel> strip some file
<Ansuel> and make a tar out of it
<Ansuel> and the mirror hash is the tar from it
<Ansuel> maybe nick have some special option set in git config (global?) that are affecting repository cloned?
xback has joined #openwrt-devel
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
schwicht has joined #openwrt-devel
<nick[m]1234> I thought we are using the github API to download it via commit archie like : https://github.com/OLSR/olsrd/archive/a9b3f1ac6e73a39b5bd97d1e66b1e039998314f5.tar.gz
<nick[m]1234> Ansuel: my github conifg looks clean
PJin[m] has joined #openwrt-devel
PJin[m] has quit [Quit: Reconnecting]
PJin[m] has joined #openwrt-devel
<Ansuel> nick maybe for some reason download.py is failing
<Ansuel> and you are fallbacking to git-raw
<Ansuel> ?
<Ansuel> and maybe that produce different hash?
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<Ansuel> nope even git-raw produce same hash
<jow> did you compare the tarballs yet?
<jow> diff or vbindiff -u over the unzipped tar's
<ukleinek> Ansuel, jow, nick[m]1234: that reminds me of https://github.blog/changelog/2023-01-30-git-archive-checksums-may-change/
<nick[m]1234> jow: where is the tarball located?
<jow> nick[m]1234: dl/
robimarko has joined #openwrt-devel
<Habbie> if i build something that depends on openssl, i can't avoid building openssl first, right?
<nick[m]1234> jow: yes I already did a diff between local build and docker build. and I did not find any difference.
<jow> nick[m]1234: well if the checksums don't match then there must be a diff ;)
<nick[m]1234> jow: I think the checksums can also be because of metadata of the tar.xz format? Like ownership stuff and packaing time?
<jow> so you say the uncompressed tars have identical checksums but the compressed ones not?
<robimarko> Hm, we had a issue with reproducible checksums that the script produced but that was fixed
<nick[m]1234> maybe my diff is wrong
<jow> nick[m]1234: yes, it is
<jow> xzcat olsrd-2023-06-12-a9b3f1ac.tar.xz > /tmp/file1.tar
<jow> xzcat dockerbla/olsrd-2023-06-12-a9b3f1ac.tar.xz > /tmp/file2.tar
xback has quit [Ping timeout: 480 seconds]
<jow> sha256sum /tmp/file*.tar
<jow> diff -u /tmp/file1.tar /tmp/file2.tar
<jow> if those are identical, then the issue is the .xz compression
<jow> if not, the issue is with tar packing (file permissions, sorting)
<jow> also tar -tf /tmp/file1.tar > /tmp/file1.list and tar -tf /tmp/file2.tar > /tmp/file2.list
<jow> diff -u /tmp/file1.list /tmp/file2.list
<ynezz> Ansuel:
<ynezz> GitHub Actions has encountered an internal error when running your job.
<ynezz> have you seen that yet?
<jow> nick[m]1234: use diff -a -u ... to force text comparison
<jow> this should display actual diffs
<nick[m]1234> the diff doesnot show anything
<jow> of the list files?
<nick[m]1234> yes
<jow> then the issue is in the tar headers, not ordering or exlusions
<jow> use diff -a -u /tmp/file1.tar /tmp/file2.tar
<jow> or better, if you can easily install it, use `vbindiff`
<jow> nick[m]1234: not the .list files, the .tar files
<jow> empty diff for list files is good, means both archives contain the same files at least
<jow> in the same order
<jow> so the diff must be either in tar headers or even in packaged file contents themselves (but you ruled that out with your recursive diff over the unpacked source trees)
<jow> so it must be tar headers
<jow> my bet is on GNU extended header fields
<nick[m]1234> jow: I'm seeing now a diff
<jow> or symlink permissions
<jow> nick[m]1234: can you upload those two .tar files somewhere?
danitool has joined #openwrt-devel
<nick[m]1234> jow: https://file.io/oorqKwE957Zc
<Ansuel> ynezz just lovely github https://www.githubstatus.com/history
<nick[m]1234> jow: sorry, I have to go afk for a short time
<Ansuel> nothing we can do about it aside from restarting the job, not our fault
<ynezz> Ansuel: well, I've tried that already, so maybe should do that harder or rebase/force push...
<jow> nick[m]1234: "The transfer you requested has been deleted."
<ukleinek> nick[m]1234, jow: numeric vs. named owners I think
<ukleinek> jow: huh, my fault, sorry, "Files will be deleted after download"
<jow> ukleinek: we should pass --numeric-owner to our tar packing routines
dangole has joined #openwrt-devel
<jow> ukleinek: can you reupload them, then? :)
<jow> *already pass
<robimarko> jow: We already are passing --numeric-owner
<jow> robimarko: I know, bad formulation
<jow> ukleinek: ahh, sticky bit :)
<jow> erm brainfart, symlink permissions
<jow> so one of both repos was packed with a wrong .tar
<jow> *archives
<Habbie> any idea why 'make package/h2o/compile' is causing 'package/kernel/linux compile'?
schwicht has joined #openwrt-devel
<jow> OpenWrt buildroot uses a patched GNU tar which forces symlink permissions to 0777
<Habbie> (using the SDK)
<jow> GNU tar running on systems which support symlink permissions (macOS) will embed the specific permissions like 0644 instead of the expected 0777
<ukleinek> jow: there is also "root" missing, look at 003c9500
<jow> also belonging to the symlink header
<ukleinek> jow: ack
<ukleinek> so it's permission bits + owner + group I think
<jow> https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=tools/tar/patches/100-symlink-force-root-name.patch
<jow> https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=tools/tar/patches/110-symlink-force-permissions.patch
<jow> I guess nick[m]1234 uses macOS on the host and the archive creator invoked the wrong tar
<nick[m]1234> Im using arch Linux
<ukleinek> note that the permission bits are 1160 vs 0777, so "|= 0777" isn't enough
<ukleinek> nick[m]1234: which file system?
<owrt-images-builds> Build [#36](https://buildbot.staging.openwrt.org/images/#/builders/30/builds/36) of `master_ath79/generic` failed.
<jow> 01160 is a weird permission to begin with
<jow> what is that? sticky + owner executable + group rw + others denied?
<Ansuel> strange that we have this only now o.o we have many macos user
<Ansuel> ynezz yes i guess the entire state of the workflow got corrupted so i guess force push is the way
<jow> ukleinek: actually the difference appears to be the checksum field, not the mode (permission) one
<Ansuel> ynezz i retriggered the entire job and not only the failed one should be the same
<ynezz> Ansuel: ok, thanks, lets see
tlj has joined #openwrt-devel
wvdakker has joined #openwrt-devel
torv has quit [Ping timeout: 480 seconds]
schwicht has quit [Read error: Connection reset by peer]
torv has joined #openwrt-devel
<nick[m]1234> ukleinek: ext4
<nick[m]1234> maybe it builds again my host tar?
wvdakker has quit [Quit: ZNC 1.8.2+deb3.1 - https://znc.in]
schwicht has joined #openwrt-devel
stintel has joined #openwrt-devel
<stintel> 12|23:35:25 [OFTC] -!- #openwrt-devel Cannot join channel (Need to be identified and verified to join this channel, '/msg NickServ help' to learn how to register and verify.)
<stintel> how is this possible if I use certfp auth
<stintel> sigh
TianlingShen[m] has joined #openwrt-devel
mrkiko has quit [Remote host closed the connection]
tidalf has quit [Ping timeout: 480 seconds]
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<stintel> is anyone aware of a tool to generate Kconfig files from e.g. a JSON file?
schwicht has joined #openwrt-devel
<f00b4r0> stintel: Perl
* f00b4r0 runs
<stintel> welpo
<nick[m]1234> ukleinek: jow what are you using as buld system?
enyc has joined #openwrt-devel
<nick[m]1234> I renamed host my tar binary, so that I'm sure build process is not accessing it. However, still wrong hash... :/
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<Habbie> 10:19Z <Habbie> any idea why 'make package/h2o/compile' is causing 'package/kernel/linux compile'?
<Habbie> because h2o depends on openssl and OPENSSL_ENGINE is on by default
schwicht has joined #openwrt-devel
tidalf has joined #openwrt-devel
tidalf has quit [Ping timeout: 480 seconds]
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
tidalf has joined #openwrt-devel
<owrt-images-builds> Build [#38](https://buildbot.staging.openwrt.org/images/#/builders/64/builds/38) of `master_bcm47xx/mips74k` completed successfully.
goliath has quit [Quit: SIGSEGV]
<Slimey> i got the Adtran Bluesocket BSAP-1930 working using the ws-ap3715i dts as base, what is this simpleboot- stuff?
<Slimey> heh and sysupgrade tries to write a ubi part when it shouldnt be
schwicht has joined #openwrt-devel
goliath has joined #openwrt-devel
<Slimey> my board is the only one with "mpc85xx,p1010" after the device for this target, is this ok?
rua has joined #openwrt-devel
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
schwicht has joined #openwrt-devel
Borromini has joined #openwrt-devel
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
schwicht has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
Borromini has quit [Quit: Lost terminal]
tidalf has quit [Remote host closed the connection]
tidalf has joined #openwrt-devel
dangole has quit [Quit: Leaving]
<schmars[m]> robimarko i think the fix for https://github.com/openwrt/asu/issues/515 needs to be backported to 23.05 - i still have 23.05 imagebuilders failing with libubox/mtd dep error
<robimarko> schmars[m]: I dont have commit rights
<schmars[m]> belkin e8450
<schmars[m]> ah i always though you did :) thanks for the fix though! (cc Ansuel )
<robimarko> schmars[m]: You need to jog my memory, I only vaguely remember this issue
<schmars[m]> i don't actually understand the underlying issue, i'm afraid
<robimarko> I mean, how was I involved in it?
<schmars[m]> no clue how that commit fixes it, but fwiu mtd simply fails to build and is then not available as a dependency of libubox
<robimarko> schmars[m]: That commit has no relation
<robimarko> It was probably that buildbots make a new build after that commit
<schmars[m]> right, that asu issue got me all confused
<schmars[m]> yeah, issue is that 23.05 still has libubox20220927_2022-09-27-ea560134-1_aarch64_cortex-a53.ipk
<schmars[m]> and mtd wants libubox20230523
<schmars[m]> b6e0a24c492537e5bbfa015e2a3638ccc53c164b is what needs backporting to 23.05, i think
<schmars[m]> feels like dejavu?!
<Ansuel> i'm not following this?
<schmars[m]> yeah i just got all confused about someone being wrong about a fix :) sorry about the ping
<schmars[m]> last 4 lines about libubox still stand though :)
_lore_ has joined #openwrt-devel
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Tapper has joined #openwrt-devel
robimarko has quit [Quit: Leaving]
<Znevna> is anyone looking into this? https://github.com/openwrt/openwrt/issues/11650
bluew has joined #openwrt-devel
<Ansuel> chance it's caused by the bridge offload hack patch
<Ansuel> that is sched to be dropped
tidalf has quit [Ping timeout: 480 seconds]
<schmars[m]> jfc i'm so blind :) fix for the libubox thing was committed yesterday, just need to be patient for the builds :)
tidalf has joined #openwrt-devel
lucenera has quit [Quit: The Lounge - https://thelounge.chat]
lucenera has joined #openwrt-devel
rua has quit [Quit: Leaving.]
tidalf has quit [Ping timeout: 480 seconds]
schwicht has joined #openwrt-devel
<owrt-images-builds> Build [#37](https://buildbot.staging.openwrt.org/images/#/builders/30/builds/37) of `master_ath79/generic` completed successfully.
Tapper has quit [Ping timeout: 480 seconds]
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Tapper has joined #openwrt-devel
tidalf has joined #openwrt-devel