<Strykar>
anyone recall a package that just downloads the binary from upstream and packages it for openwrt?
gch9812138 has joined #openwrt-devel
gch981213 has quit [Ping timeout: 256 seconds]
gch9812138 is now known as gch981213
<philipp64>
Strykar: that sounds problematic, since paths/directories might be laid out differently, but they’re frequently compiled into the binaries…
gch981213 has quit [Read error: Connection reset by peer]
<aparcar[m]>
mangix: doesn't it miss a mirror hash?
PtitGNU has quit [Quit: Quassel terminated!]
<aparcar[m]>
mangix: also, do you want to become maintainer of that package? I'm not sure what's the process for "non core developers" and being a maintainer but I'd rather have a name there
dana44 has joined #openwrt-devel
kakaka has quit [Ping timeout: 240 seconds]
jmccrohan has quit [Ping timeout: 272 seconds]
jmccrohan has joined #openwrt-devel
kakaka has joined #openwrt-devel
dorf has joined #openwrt-devel
dana44 has quit [Ping timeout: 245 seconds]
gch9812135 has joined #openwrt-devel
gch981213 has quit [Ping timeout: 260 seconds]
gch9812135 is now known as gch981213
<mangix>
philipp64: what do you think about removing perl-device-usb?
gch9812137 has joined #openwrt-devel
gch981213 has quit [Ping timeout: 246 seconds]
gch9812137 is now known as gch981213
gch9812136 has joined #openwrt-devel
gch981213 has quit [Read error: Connection reset by peer]
gch9812136 is now known as gch981213
dorf has quit [Remote host closed the connection]
gch9812132 has joined #openwrt-devel
gch981213 has quit [Ping timeout: 264 seconds]
gch9812132 is now known as gch981213
<mangix>
aparcar[m]: I don't want to maintain that package. I don't use it. It's also a dev package
<mangix>
ffainelli seems to be gone now
black_ant has joined #openwrt-devel
black_ant has quit [Changing host]
black_ant has joined #openwrt-devel
gch981213 has quit [Read error: Connection reset by peer]
gch9812137 has joined #openwrt-devel
gch981213 has joined #openwrt-devel
gch9812137 has quit [Ping timeout: 260 seconds]
gch9812137 has joined #openwrt-devel
gch981213 has quit [Ping timeout: 260 seconds]
gch9812137 is now known as gch981213
gch9812133 has joined #openwrt-devel
gch981213 has quit [Ping timeout: 260 seconds]
gch9812133 is now known as gch981213
Darkmatter66 has joined #openwrt-devel
Darkmatter66 has quit [Read error: Connection reset by peer]
Darkmatter66 has joined #openwrt-devel
Darkmatter66 has quit [Ping timeout: 264 seconds]
ivanich has joined #openwrt-devel
gch981213 has quit [Read error: Connection reset by peer]
gch981213 has joined #openwrt-devel
Borromini has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
danitool has joined #openwrt-devel
feriman has joined #openwrt-devel
<russell-->
mangix: you mean [florian]
gch9812138 has joined #openwrt-devel
gch981213 has quit [Ping timeout: 264 seconds]
gch9812138 is now known as gch981213
Darkmatter66 has joined #openwrt-devel
feriman has quit [Quit: WeeChat 3.0]
swex has quit [Quit: No Ping reply in 180 seconds.]
damex has quit [Quit: No Ping reply in 180 seconds.]
<enyc>
jow: hrrm it seems the libc variant makes a lot of difference to getaddrinfo() priorities/intelligence, which may be nobbling busybox ntpd seemingly
<jow>
enyc: yes, for some reason glibc decided to turn its stub resolver into a kitchensink
<jow>
ldir: my first hunch would be a dangling symlink
<enyc>
jow: I think suitably tested patch to busybox is a sensible pragmatic approach, but without being openwrt-dev-centric [as somebody learning my way back into compsci and so-on myself] I don't know how best to get it attended to, I'd like to test variants/builds which is something I am well placed to do.
<ldir>
which looks sensible & correct. So maybe there's something really odd in my config.
<jow>
ldir: not sure the change is correct
<ldir>
I bow to your knowledge :-)
<jow>
ldir: `:=` assignments are expended when defined while `=` ones are when used
<jow>
ldir: but anyhow, maybe it is correct after all. I wonder why the chmod fails in your case, since the packaging should happen under fakeroot
<jow>
but maybe that broke again on OS X
Acinonyx has joined #openwrt-devel
<ldir>
chmod can't change a file that doesn't exist
<jow>
the root cause is not permitted
<jow>
in line 5
<jow>
I think the no such file or directory error is due to the fact that we add /etc/syslog.conf as conffile but do not package it, that should be nonfatal tough
Acinonyx_ has quit [Ping timeout: 240 seconds]
<jow>
ldir: I'd say fakeroot is broken
<jow>
otherwise chown root:root would be allowed
<jow>
did you recently upgrade OS X ?
<ldir>
yes I've gone up to big sur, but I've done a whole load of other changes too. I was able to get builds through until the commit I pointed at.
<jow>
I see
<jow>
apprently that change un-broke the logic for setting PKG_FILE_MODES which now triggers chown/chmod of /bin/busybox
<jow>
note that busybox is the only package in core using PKG_FILE_MODES
<ldir>
What can I look at to assist further?
<ldir>
you mentioned 'fakeroot' - no idea what that is
<jow>
and only a handful from the feeds use it, so maybe your package selection simply contains no package performing fakeroot assisted chown/chmod
<jow>
essentialyl it is a preload mechanism, it preloads a library before invoking the shell / a command
<jow>
that library intercepts any filesystem related system calls and for some operations pretends that they worked while not doing actual changes on the underlying fs
<jow>
this is useful to chown/chmod files before tar'ing them while lacking the permissions to actually chown root in the underlying fs
<jow>
anyhow, to start debugging I'd probably edit /Volumes/CaseSense/wrt/staging_dir/host/bin/fakeroot
<zorun>
jow: hi, I'm looking at adding persistent connections to opkg; is it acceptable to add a dependency on libuclient?
<jow>
drop some "set -x" near the top
<zorun>
currently, opkg simply calls "wget" (which is uclient-fetch on openwrt)
<jow>
make package/busybox/{clean,compile} V=sc
<jow>
see what fakeroot is(n't) doing
<zorun>
the main possible issue I see is on the host, I'm not 100% sure libuclient is usable here
<ldir>
it's a symlink to libfakeroot-0.dylib which exists.
<rsalvaterra>
jow: Something completely different: any idea why CONFIG_STRIP_KERNEL_EXPORTS breaks x86(-64, at least) kernel builds? It seems to erroneously purge symbols which are required by modules, but then again it doesn't break mvebu, ramips or ath79 kernels, which is weird.
<jow>
rsalvaterra: up until now I didn't even know that it is broken, so... no idea
<jow>
maybe upstream gained some obscure feature that now requires introspection during kernel build
<jow>
but that pure unenducated guess, maybe you could share a log of the breakage
<rsalvaterra>
jow: Welcome to the party, then. :P I'm also building OpenWrt for an APU2C4 for the first time.
<rsalvaterra>
jow: No breakage during build, only at runtime. But I can cook an image to show you.
<zorun>
jow: I thought about it, yes, but using a proper library seems cleaner
<jow>
during runtime? so it fails to load modules?
<jow>
or grub failing to boot the kernel?
<zorun>
but thanks for the suggestion, it might be easier to extend uclient-fetch indeed
<rsalvaterra>
jow: Modules fail to load and there's a dump of the missing symbols in dmesg.
<jow>
zorun: apart from that I am not sure if I want to trust libuclient
<rsalvaterra>
jow: For example, the module I have on the APU is WireGuard. Everything works fine apart from it.
<zorun>
jow: you mean API-wise?
<rsalvaterra>
*the only module
<jow>
zorun: I know that its proxy and redirect handling is broken at least, and I wouldn't surprised by yet unkown bugs in http body parsing
<zorun>
ah
<zorun>
well, we do use uclient-fetch in opkg already
<jow>
true
<zorun>
but you're right, it would tie opkg much more tightly to uclient
<zorun>
maybe in the future we could use the "real" wget on target
<jow>
ldir: next try would be STAGING_DIR=/Volumes/CaseSense/wrt/staging_dir/target-x86_64_musl /Volumes/CaseSense/wrt/staging_dir/host/bin/fakeroot bash
<jow>
ldir: within that shell session, try something like touch /tmp/foo; chown root:root /tmp/foo
<jow>
see if that is working
<jow>
(/tmp/foo should appear on your system but not really belong to root, even if the chown succeeds)
<jow>
within the running bash session it should appear as belonging to root, though
<zx2c4>
rsalvaterra: whats dmesg say?
<jow>
rsalvaterra: did you enable the wireguard kmod and recompile after you initially enabled the export stripping?
<jow>
like mklibs, the export stripping needs to know all kmods that are going to be used, maybe it failed to accout for wireguard
<jow>
maybe its broken for out-of-tree kmods in general
<rsalvaterra>
zx2c4: It's not a WireGuard specific problem, I think basically every module is affected. Let me build an image do show you guys. :)
<jow>
nvm, wireguard is not out of tree anymore I guess
<rsalvaterra>
I don't think it's just OOT modules, that's the problem.
<rsalvaterra>
jow: WireGuard is still OOT on Linux 5.4.
<jow>
my guess would be that something changed with a major kernel bump
<jow>
and the exports stripping now fails to assemble the required symbol list
<jow>
try to drill down into that area
<jow>
see also 2ef0acc5fcda557fa5aaad35d27cb8cf75be96d2
<jow>
maybe the same happened again
<rsalvaterra>
Alright, I'll try to dig deeper, thanks for the hints.
<jow>
didn't upstream recently change stuff around GPL symbol exporting etc?
<rsalvaterra>
jow: I think this is an old issue (by "old" I mean at least one year old). I remember trying to compile OpenWrt 19.07 and having similar issues on the APU.
gch9812137 has joined #openwrt-devel
gch981213 has quit [Read error: Connection reset by peer]
gch9812137 is now known as gch981213
goliath has quit [Quit: SIGSEGV]
<ldir>
jow: as you've probably guessed by the delay, fakeroot doesn't work. I even installed the homebrew version and that has the same problem. Hmmmm.
<ldir>
ahh, I have an idea....
<rsalvaterra>
jow: I'd replace that grep | sed with a single awk, most likely… :P
gch981213 has quit [Read error: Connection reset by peer]
gch981213 has joined #openwrt-devel
Ycarus has quit [Quit: Ycarus]
gch981213 has quit [Read error: Connection reset by peer]
gch9812133 has joined #openwrt-devel
Ycarus has joined #openwrt-devel
goliath has joined #openwrt-devel
<Hauke>
rmilecki: I have a Buffalo WSR-2533DHP2 (mt7622) which uses NAND flash and a TRX image header, U-Boot expects a TRX header. I would like to put the rootfs and the overlay FS into UBI, how to do this best?
<Hauke>
I think the Buffalo device using brcm53xx is doing this
<rmilecki>
rmilecki: depends on what Buffalo wants in the TRX
<rmilecki>
rmilecki: standard Broadcom boards expect kernel in the TRX partition
<rmilecki>
rmilecki: so you can't have kernel in the UBI
<Hauke>
rmilecki: it is loading the kernel with "append-ubi | trx-nand | < some other headers>"
<rmilecki>
that's why we have:
<rmilecki>
1. TRX part #0: kernel
<rmilecki>
2. TRX part #1: UBI
<Hauke>
yes I would like to have this
<Hauke>
so kernel on raw flash and rootfs + overlay in ubi
<rmilecki>
so just build UBI image with whatver you want (normally squashfs + ubifs)
<rmilecki>
and then use (o)trx to build final TRX binary
<rmilecki>
see arch/arm/boot/dts/bcm47094-phicomm-k3.dts
Borromini has joined #openwrt-devel
<rmilecki>
oh, we have such a thing?
<Hauke>
ok will try brcm,trx
<rmilecki>
I think i left it for some old targets
<rmilecki>
try upsream mtd parser
<Hauke>
rmilecki: thanks, I haven't noticed this is already upstream
feriman has joined #openwrt-devel
hbug__ has quit [Ping timeout: 240 seconds]
kakaka has quit [Ping timeout: 240 seconds]
hbug__ has joined #openwrt-devel
nslu2-log has quit [Quit: Caught sigterm, terminating...]
urjaman has quit [Read error: Connection reset by peer]
urjaman has joined #openwrt-devel
nslu2-log has joined #openwrt-devel
<Hauke>
rmilecki: thanks, one step further, now it complains about: "ubi0 error: validate_ec_hdr: bad VID header offset 2048, expected 512" is this related to 500-UBI-Detect-EOF-mark-and-erase-all-remaining-blocks.patch ?
<rmilecki>
yes, it may be some previous UBI data that wasn't ereased
feriman has quit [Quit: WeeChat 3.0]
<rmilecki>
I think OPenWrt also supports some deadc0de mark
<Hauke>
ok
<rmilecki>
I added 500-UBI-Detect-EOF-mark-and-erase-all-remaining-blocks.patch cause I didn't know about generic solution