<aparcar[m]>
anyone knows debugging if a luci-app doesn't show up in the menu?
danitool has quit [Ping timeout: 480 seconds]
Tapper has joined #openwrt-devel
Rentong has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
Rentong has quit [Ping timeout: 480 seconds]
rmilecki has joined #openwrt-devel
Tapper has joined #openwrt-devel
<aparcar[m]>
do we have any tracking which ABI version changed between 19.07.6 and 19.07.7 and 21.02.x?
<aparcar[m]>
jow: ?
<aparcar[m]>
or hauke or zorun ?
Tapper has quit [Ping timeout: 480 seconds]
Tapper has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
floof58 has quit [Ping timeout: 480 seconds]
valku has quit [Quit: valku]
floof58 has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
Tapper has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
nitroshift has joined #openwrt-devel
Tapper has joined #openwrt-devel
Rentong has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
Tapper has joined #openwrt-devel
Rentong has quit [Ping timeout: 480 seconds]
<rsalvaterra>
hauke: Wait, the backports are generated with SmPL? Coccinelle semantic patches?
<jow>
blocktrron: the usual motivation for moving stuff to packages is bypassing the openwrt base review process
<jow>
indeed for uqmi the trigger seemed to be a feature ptch that wasn't merged or that didn't receive timely feedback
<jow>
in principle I am fine with it but I've noticed multiple times that stability of moved packages regressed
<jow>
broken stuff got merged, immediate bumps to new upstream versions breaking existing functionality etc.
<jow>
nbd: what's the proper way to deal with too large replies in ubus? I've multiple reports of certain ubus rpc calls not working under certain conditions because the payload grew too large
<jow>
apparently rpcd right now crashes when dealing with such messages, haven't debugged it in detail yet
<jow>
wondering about the proper approach to deal with that for now
<nbd>
how large is the reply?
<nbd>
and what kind of call is it?
<f5>
is 21.02 going to released this weekend or has that changed?
<jow>
nbd: not entirely sure yet, but something in 50-70KB ballpark
<jow>
a file listing (array of objects with stat info)
<nbd>
in that case it might be a bug in libubus or rpcd
<nbd>
we used to have a 64KB limit, but i think that limitations was removed at some point
<nbd>
so if it crashes, please make a coredump and get me a backtrace
<nbd>
i'll help you debug it
<hauke>
rsalvaterra: yes, we use SmPL to adapt to many of the API changes
<ynezz>
just wondering if you've spotted anything wrong with this PR, that you didn't pushed it to 21.02 or you just forget it :)
<ynezz>
s/forget/forget about/
takimata has quit [Quit: wat?]
takimata has joined #openwrt-devel
<ynezz>
hauke: sorry for the noise, I've just realized, that I was checking different branch, found that commit ea308e2f383c071a710973aa7825530de80c08d6
rejoicetreat has joined #openwrt-devel
<ynezz>
and it seems to be part of v21.02.0-rc3, so I'm wondering why it's still borken
rejoicetreat has quit []
felix has quit []
felix has joined #openwrt-devel
danitool has joined #openwrt-devel
rejoicetreat has joined #openwrt-devel
rejoicetreat has quit []
Acinonyx has joined #openwrt-devel
dangole has joined #openwrt-devel
Acinonyx_ has quit [Ping timeout: 480 seconds]
dangole has quit [Remote host closed the connection]
Tapper has quit [Ping timeout: 480 seconds]
dangole has joined #openwrt-devel
<ynezz>
and ea308e2f383c071a710973aa7825530de80c08d6 is not the solution anyway, the imagebuilder is broken again
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nlowe has joined #openwrt-devel
dangole has quit [Ping timeout: 480 seconds]
nlowe has quit [Remote host closed the connection]
Borromini has joined #openwrt-devel
Rentong has joined #openwrt-devel
Rentong has quit [Remote host closed the connection]
Rentong has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
Rentong has quit [Ping timeout: 480 seconds]
Tapper has joined #openwrt-devel
_lore_- is now known as _lore_
rejoicetreat has joined #openwrt-devel
rejoicetreat has quit [Remote host closed the connection]
Borromini has quit [Ping timeout: 480 seconds]
Strykar has quit [Quit: /quit]
Strykar has joined #openwrt-devel
Rentong has joined #openwrt-devel
Rentong has quit [Ping timeout: 480 seconds]
Rentong has joined #openwrt-devel
Rentong has quit [Ping timeout: 480 seconds]
Rentong has joined #openwrt-devel
Tusker has quit [Quit: Time wasted on IRC: 14 hours 34 minutes 50 seconds]
Rentong has quit [Ping timeout: 480 seconds]
Rentong has joined #openwrt-devel
valku has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
dsmith-work has joined #openwrt-devel
<jow>
ynezz: well the SDK in phase2 will pull the latest HEAD of openwrt.git to build the packages there
<jow>
ynezz: marking a package nonshared actually worsens the problem because then it is frozen and will never get rebuilt
<jow>
ynezz: not until the next release, that is
<jow>
the actual fix would be not changing the library abi (multiple times!) during the RC phase
<jow>
without abi tracking this would've been silent crash bugs or (assuming it affected libubus) soft bricks every time
<jow>
for everyone using opkg on target or packaging images with the IB
JohnA has joined #openwrt-devel
<jow>
however, as for an actual solution - I could only think about somehow retaining older versions on the download server
<ynezz>
sound like a plan
<jow>
this would be a process outside of buildroot/sdk though, something we'd need to arrange on the dl server. Maybe from a rsync xfer hook that detects if an abi versioned package .ipk is about to be deleted and which then moves this package to an archive repository and updates the index accordingly
<jow>
doesn't actually need to be an extra archive repo, since the names are unique (basename + abi suffix) they can coexist in the same repo
<jow>
that would also imply moving indexing logic to the dl server (something I wanted to do for a long time already, to be able to switch from full repo transfers to incremental .ipk upload queues)
<jow>
or at the very least duplicating it there
<jow>
means taking the scripts/ipkg-make-index.sh out of buildroot and running it on the server
<jow>
(again)
<jow>
that entails signing headaches though
<jow>
so the recalculated index need to be re-signed as well
<jow>
which means the dl server needs to ask the build master for signing
<jow>
...
<jow>
or simply don't break the abi :)
<jow>
another semi-cheap fix idea
<jow>
tweak the non-master branch builders to not take the latest -SNAPSHOT SDK for building packages but the last rc or maint version
<jow>
not sure offhand whether this might result in other coherency issues
<jow>
but if there's any then likely less of them compared to the current situation
<ynezz>
SDK doesn't matter, does it? buildbot clones/pulls the source directory
<jow>
it matters for phase2
<jow>
phase2 pulls latest SNAPSHOT SDK which pulls latest HEAD of feeds
<jow>
while hm, thinko, nvm
<jow>
I guess the simplest solution is making libiwinfo not-nonshared
<jow>
it is only nonshared due to the former madwifi and/or broadcom-wl support which is target specific
<jow>
now that is is only doing mac80211 more or less there should be no need to have it nonshared anymore
<ynezz>
so revert 885910225d86722811ad4d3610d9e5d076e5614a and 72cc44958ef4e0df1a152178514c92899d6a957a ?
<jow>
we could also tweak non-master phase2 builders to patch the SDK feeds.conf to not pull latest HEAD of the core (openwrt.git) feed but the latest tag
<jow>
that should solve the current problem at hand at least for users of the latest rc/maintenance release
<jow>
since phase2 packages will get built against nonshared lib versions corresponding to the latest rc/maintence release tag
<ynezz>
yes, this seems like the easiest fix
<jow>
git -c 'versionsort.suffix=-' ls-remote --tags --sort='-v:refname' https://git.openwrt.org/openwrt/openwrt.git | sed -ne "s#[0-9a-f]\+[[:space:]]\+refs/tags/\(v21\.02[.0-9rc-]\+\)\$#\1#p" | head -n1
<jow>
"21" and "02" in the pattern need to be interpolated from the configured branch version
Rentong has quit [Remote host closed the connection]
<jow>
the result appended with ";" to the "src-git core ..." url of feeds.conf
<ynezz>
thanks, pushed to master and 21.02, have a nice weekend!
minimal has joined #openwrt-devel
dangole has joined #openwrt-devel
Tapper has joined #openwrt-devel
Rentong has quit [Remote host closed the connection]
dangole has quit [Ping timeout: 480 seconds]
Rentong has joined #openwrt-devel
Rentong has quit [Ping timeout: 480 seconds]
<blogic_>
libusb update in 0798b13d7df0deafd005b14d52420c4f480d146e. breaks several 3gpp modems
<blogic_>
lsusb fails to discover them, making usbmode not pick them up as it uses libusb
<blogic_>
libusb 1.0.25 is on the doorstep but not final yet, it does fix the problem
<blogic_>
I would like to revert 0798b13d7df0deafd005b14d52420c4f480d146e in 21.02
<blogic_>
@ynezz @jow ^^^
<blogic_>
current libusb HEAD works but I would rather revert to latest known working than use a "not yet release" git HEAD
<blogic_>
@hauke ^^^
silver has joined #openwrt-devel
felix has joined #openwrt-devel
ecloud has quit [Ping timeout: 480 seconds]
<digitalcircuit>
Hmm. Merely as a test, dropping the DTSI changes for the ipq806x CPU schedule backport to 21.02 while keeping the "enable 1.4 GHz cache" change still results in a reboot, but less often. I still suspect I'm just missing something important from "master" (where things work fine) to backport to "21.02"...
Rentong has joined #openwrt-devel
ecloud has joined #openwrt-devel
dangole has joined #openwrt-devel
<digitalcircuit>
Yep, back to having actual errors in the kernel serial log: rcu_sched hang this time. Maybe I need to have stress-ng focus on hashing and memory cache operations to recreate this.
Rentong has quit [Ping timeout: 480 seconds]
<jow>
blogic_: feel free to revert
<jow>
blogic_: looks like one of those updates because there's an update
<rmilecki>
blogic_: this doesn't help? libusb: Fix parsing of descriptors for multi-configuration devices
<rmilecki>
(it's a commit subject)
* rmilecki
is going to sleep
Rentong has joined #openwrt-devel
Borromini has quit [Quit: leaving]
rmilecki has quit [Ping timeout: 480 seconds]
Rentong has quit [Ping timeout: 480 seconds]
arnd has quit [Remote host closed the connection]
rejoicetreat has quit [Remote host closed the connection]
rejoicetreat has joined #openwrt-devel
Borromini has joined #openwrt-devel
<Borromini>
johnf: ping. was your PPPoE issue solved on the 21.02 RCs? I recall you complaining about it breaking your PPPoE, and I've seen this myself on an APU2, but not on my EdgeRouter 4 (different DSL providers)
<Borromini>
and I pointed that guy to the (albeit concise) bug report FS #3909
<Borromini>
I gotta go again now but if you're on the forum feel free to PM me, same nick there as I got here. would be great if we could track that down before 21.02 goes final (I can unfortunately not test the APU2 connection since it's in production for a small business)
JohnA has quit [Ping timeout: 480 seconds]
Borromini has quit []
JohnA has joined #openwrt-devel
rejoicetreat has quit [Remote host closed the connection]
JohnA has quit [Ping timeout: 480 seconds]
<minimal>
Borromini: I have a APU2 here which I've been waiting for 21.02 to come out to set up as a PPPoE router. Haven't looked at the RCs yet but might try and find some time to check
rektide has quit [Remote host closed the connection]
aleasto has quit [Quit: Konversation terminated!]
dedeckeh has quit [Remote host closed the connection]
JohnA has joined #openwrt-devel
JohnA_ has joined #openwrt-devel
JohnA has quit [Ping timeout: 480 seconds]
Acinonyx_ has joined #openwrt-devel
Acinonyx has quit [Ping timeout: 480 seconds]
dangole has quit [Ping timeout: 480 seconds]
Umeaboy has joined #openwrt-devel
<Umeaboy>
Hi! There's a package missing called libubus20210630 that's needed by wpad-basic-wolfssl so I cannot install all the updates.