Daanct12 has quit [Read error: Connection reset by peer]
minimal has quit [Quit: Leaving]
donoban-woops has joined #openwrt-devel
donoban has quit [Ping timeout: 480 seconds]
Daanct12 has joined #openwrt-devel
skynet2 has quit [Ping timeout: 480 seconds]
<russell-->
i think i figured out a way: build a kernel with CONFIG_KERNEL_KALLSYMS turned off, so the kernel size is under 3MB, on the target: modify /lib/upgrade/ubnt.sh's ubnt_get_target_kernel() function replacing kernel2 with kernel1, copy the compat_version 2.0 firmware with a kernel <= 3MB to the target, sysupgrade -v -n $file, get the warning, then add the -F to force. should reboot into the
<russell-->
wonderful world of 6MB kernel land.
n3ph has quit [Ping timeout: 480 seconds]
MAbeeTT4 has joined #openwrt-devel
MAbeeTT3 has quit [Ping timeout: 480 seconds]
felix has quit []
felix has joined #openwrt-devel
Daanct12 has quit [Ping timeout: 480 seconds]
Daanct12 has joined #openwrt-devel
<russell-->
not sure the ubnt.sh changes is even necessary, i forgot on one, factory partition was set to boot from kernel2, booted okay anyway
<dwfreed>
it probably failed to boot kernel2, and then automatically switched to kernel1 after
<russell-->
dwfreed: or it wrote to kernel2 and u-boot just jumped to the offset and booted from there, a followup sysupgrade rejiggered things back to the front
<russell-->
in anycase, as long as the kernel is <=3MB, it seems to work either way with -F
<enyc>
Hrrm; alas I don't know way round apk to create said page
<mrkiko>
what's about the .list files in /lib/apk/packages ? Are we keeping them forcompatibility or are they part of apk working? I ask because I find strange that in a .list file the .list file itself is mentioned, like it was part of the package
<ynezz>
KanjiMonster: hi, instead of supporting both opkg/apk, shouldn't we get rid of the opkg/USE_APK once the transition to apk is done?
robimarko has joined #openwrt-devel
<ynezz>
KanjiMonster: I mean, opkg is dead, so I understand the current situation with support for both opkg/apk just a support for smoother transition during rollout
<robimarko>
aparcar: It seems that EXTRA_DEPENDS in the current format is breaking APK depends
<robimarko>
ERROR: info field 'depends' has invalid value: dependency format is invalid
<robimarko>
For example on uspot which has: EXTRA_DEPENDS:=ucode (>= 2023-11-07)
rua is now known as Guest9192
rua has joined #openwrt-devel
Guest9192 has quit [Ping timeout: 480 seconds]
rua has quit [Remote host closed the connection]
<ynezz>
robimarko: needs to be converted? EXTRA_DEPENDS:=ucode (>= 2023.11.07)
schmars[m] has joined #openwrt-devel
<mrkiko>
ynezz: any chance you could propose a PR or something? Would be nice... :)
<mrkiko>
or, MR in gitlab parlance I guess...
<robimarko>
ynezz: It seems that its not EXTRA_DEPENDS, it doesnt like something in depedns
<robimarko>
Ah, it is cause: depends:ucode>=2023-11-07
<robimarko>
APK probably doesnt like that but I removed EXTRA_DEPENDS
<robimarko>
I forgot to remove it from uspotfilter though
<robimarko>
What is even purpose of EXTRA_DEPENDS?
<robimarko>
Yeah, removing it from uspotfilter as well removes the error
<mrkiko>
It would bealso nice if apk complianed if I ask it info about a package does not exist. "apk info nonexistentpackage" return nothing to stdout/stderr, but an exit status 1
<Habbie>
openwrt/rootfs on docker hub is pre-apk it appears
<Habbie>
Last pushed 7 days ago by aparcar
aparcar[m] has joined #openwrt-devel
<aparcar[m]>
Habbie: yea it's a mess, I'll see how to improve this rootfs situation since it's not really useful anyway
<Habbie>
what's not useful?
Fijxu has quit [Ping timeout: 480 seconds]
<Habbie>
ok yes. today it is not useful, as it expects Packages.gz :D
Fijxu has joined #openwrt-devel
<aparcar[m]>
i'll try to fix this soonish, point is instead of just pulling openwrt/rootfs:x86 it should instead just do docker load -t testcontainer https://downloads.openwrt.org/targets/...rootfs.tar.gz and then use that. this way it's always fresh and we don't build daily docker containers which is a terrible waste of resources
<ynezz>
robimarko: looks like runtime/install time dependency (not build time dependency), so it assures, that you're using the package with expected ucode version (version which supports feature needed by that package)
<Habbie>
aparcar[m], oh! yes that makes sense to me, i used docker import for your apk rootfs.tar.gz
<ynezz>
robimarko: depends:ucode>=2023.11.07 -- that is apk compatible version
<aparcar[m]>
Habbie: do you mind adopting the ci in packages.git to do very much that?
<Habbie>
aparcar[m], are you, and i'm only trying to be clear, "habbie would you volunteer to switch the packages.git github workflows from 'docker pull' to 'docker import'" ?
<Habbie>
+asking
<Habbie>
(because the answer is "i can at least have a look")
<aparcar[m]>
yeas sorry i'm multi threading here sorry
<Habbie>
no worries. the only thing i don't want is to go do something and waste everybody's time because i've misunderstood something :)
<Habbie>
understood, will have a go at it
<Habbie>
ah, that's actually in the shared workflows
caskd has quit [Ping timeout: 480 seconds]
<aparcar[m]>
should be relatively straight forward, don't depend on openwrt/rootfs in the dockerfile but instead import the tar.gz before and then depend on it instead 🙂
<aparcar[m]>
thanks for your interested
<Habbie>
yes, understood. i see it moved to shared 4 days ago :)
<Habbie>
so this is just for the testing -after- build, i see
<Habbie>
aparcar[m], (3) it sounds like i need to map x86_64 to x86/64, but that's a small table or matrix extension
<Habbie>
aparcar[m], beyond those details, the task looks simple enough :) no rush, i'll find hints in here from you or somebody else at some point
<schmars[m]>
youre looking for /releases/24.10-SNAPSHOT maybe? the snapshot dirs are hidden
<schmars[m]>
and for the target=>architecture mapping theses a script in oprnwrt.git/scripts, called dump-targets-info.pl or sonething like that. i use that to generate my buildbot jobs
<schmars[m]>
(sorry i might be lacking context)
<Habbie>
checking
<Habbie>
schmars[m], the context is finding the right rootfs.tar.gz for packages.git to run tests in (inside docker)
<Habbie>
i see 23.05-SNAPSHOT also exists, so that seems perfect
<ynezz>
tl;dr Seems there are some packages with a huge description field. e.g. external-protocol has description of more than 2kB. And I can confirm that this triggers this issue.
Stat_headcrabbed has quit [Quit: Stat_headcrabbed]
goliath has quit [Quit: SIGSEGV]
<aparcar[m]>
Habbie: any luck so far? Do you need my help?
<Habbie>
aparcar[m], schmars[m] helped me with the details i got stuck on, but i'm out of time for today :) will continue later. no help needed right now :)
<aparcar[m]>
excellent, please let me know this seems to be a bit of a pressing thing 🙂
<Habbie>
i have run some of the parts of the yml locally, doing import and then a build that doesn't pull, etc.
<aparcar[m]>
thanks for your work
<Habbie>
oh pressing how?
<Habbie>
oh. packages.git PRs are still testing against opkg i guess?
<Habbie>
building, even
<Habbie>
hmm no. building with apk
<Habbie>
No *.ipk files found :)
rmilecki has joined #openwrt-devel
<Habbie>
aparcar[m], just to be clear, i might have a few hours for this this week, i might not
<aparcar[m]>
well I always have my fingers crossed! (hence the garbage code)
<aparcar[m]>
I might have a look at it or annoy Ansuel with it, I don't want to confuse to many people with APK
<mrkiko>
but hey, APK seems great. The whole "virtuals" concept seems interesting. And the program feels very smooth
<schmars[m]>
and it's actively maintained <3
<Habbie>
that said, we are using v3 while alpine is not, but i'm sure they'll switch one day :)