<stintel>
you're best to rm build_dir and staging_dir every time you build, otherwise something will creep in :P
<mangix>
rsalvaterra: also enable ccache
<mangix>
speeds up compilation after nuking build_dir
<mangix>
...why are there x32 patches?
<mangix>
who uses x32?
<rsalvaterra>
mangix: I would LOVE to use x32…
<mangix>
why?
<rsalvaterra>
Because 32-bit pointers with a 64-bit address space?
<rsalvaterra>
Smaller i$ footprint?
<lipnitsk>
how to get kernel_oldconfig/kernel_menuconfig to modify target/linux/generic/config-5.{4,10}?
<rsalvaterra>
The x32 ABI was actually suggested by Knuth himself.
<lipnitsk>
or is it only editable by hand?
<mangix>
rsalvaterra: Sounds horrible. But then so is mips16
<rsalvaterra>
mangix: You can't compare MIPS16 with x32… MIPS16 is a hardware feature, x32 is a pure software ABI.
<rsalvaterra>
Aw, WTF…!
<rsalvaterra>
src/lib/crypt_ops/crypto_openssl_mgt.c:256:15: warning: implicit declaration of function 'ENGINE_by_id' [-Wimplicit-function-declaration]
<damex>
Grommish: cavium3 should not have that happening unless something goes completely wrong. what exactly is going wrong with watchdog?
<Grommish>
dunno, but every once in a while I get that register dump from the watchdog when it attempts to reboot, never consistently though
<Grommish>
which requires a full power pull to clear
Tapper has joined #openwrt-devel
<Pepe>
lipnitsk: That's not about complaining. My opinion about refreshing patches is that it should not be mandatory. We are adding more bricks into the wall against contributors. Which is bad. If we are fixing typo, the build should proceed and should not says: "Ehm... patches are wrong.". Even we do tree wide commit where we fixes patches, I think that refreshing patches should be opt-in.
<Pepe>
Even these days there is lack of contributors in PRs. They can not do Signed-off-by, then dont do compile either run test and so on. Sometimes it is stalled or they are getting angry if they should write commit message. :(
<lipnitsk>
alternatively, we could add infrastructure to refresh them more easily (without having to clone openwrt). But yeah, if there is evidence that it impedes contributions you have a point. Nothing wrong with teaching proper commit etiquette though. Being friendly is one thing, accepting unmaintainable or messy changes is another.
<olmari>
Exactly... keeping up maintainability and thus general quality of coding etc should be kept up to high standards... ways to achieve that more easily is then again very welcomed... in mine opinion :)
<Pepe>
But doing refresh can lead to some issues, so its better to get it run tested again even small changes can go badly. Someone does not want to do that don't know why in case of php packages, but anyway. In CI there is no possible to skip patches step and thus force you to do that, which was my point. Others projects have possibility to skip some of those tests, we don't have it. But I agree with you on
<Pepe>
this guys, but as I said there are not many contributors these days and we are making it harder and harder for someone
<lipnitsk>
refresh is just commiting what quilt already does every time it prepares a package for build, we are just commiting the truth into the repo. If it's broken the patch needs to be fixed
<Grommish>
What if you don't use quilt?
<lipnitsk>
there is argument about the superfluous changes, but it's more about getting into the habit of refreshing, or making that a more easy process somehow?
<lipnitsk>
how do you test your changes then? You need to at least compile it with openwrt, right? Or just rely on CI? :)
<Grommish>
I meant for patch creation, sorry. Don't mind me
<Grommish>
I realized after the fact I'm wandering into an argument ;p
<Pepe>
The truth is that a lot package maintainers are rely on CI these days. :(
<lipnitsk>
ya you can create the patch however you want, just run make package/feeds/packages/<your_pkg>/refresh. But if that last step is difficult for a lot of contributors then we do have a problem
<lipnitsk>
Maybe we should at least put the command into the PR template
<Grommish>
Sadly, I'm on of the ones Pepe referenced earlier regarding Commit messages. But, I freely admit I got pissy about the situation, but he was right in the situation
<lipnitsk>
maybe a cool web app to refresh your patches? LOL
<Grommish>
I'd rather have standards than not, because I rely on people with more knowledge making the decisions on what get's committed
<lipnitsk>
I think the main point is achieving a balance that makes maintainers as well as contributors happy
<Pepe>
This. +1
<lipnitsk>
so discussion should continue and the change can be easily turned off if needed.
<Grommish>
Discussion should always continue.. If nothing else, it'll show why it won't work - at least for now, or why it should be done. Info is good, more is better
<lipnitsk>
Grommish: it's very humbling working with a large FOSS community - I know it's a learning for me because I haven't really been brewing in these circles much
<lipnitsk>
good learnings, but time needs to be invested. And time is what a lot of people don't have..
<Grommish>
I know I started very upset with adrianschmutzler because I thought he was being overly anal-retentive about commits ;p
<Grommish>
Unti I saw what happens when you aren't ;/
<Grommish>
(I also took it as a personal attack, wrongfully because I was of the mindset I tested this already.. ahem.. Not a bright spot for me)
<Grommish>
In any case, I don't understand the CI stuff anyway, except my package breaks it due to space limitations, but I'm all for standards
<lipnitsk>
ultimately, whatever gets more users onto the platform is probably the right way to proceed
<lipnitsk>
btw github is a blessing, at least for the package repo
<Grommish>
Not if the quality drops *shrug* I could run Tomato on my R8000 if I wanted to settle for the 3.x kernel to use the cts.ko mob, but I want something that isn't going to cause me to wonder
<Grommish>
Costs me half my paid for throughput to do go
<lipnitsk>
trying out matrix, let's see.
<Grommish>
Besides, nothing stops anyone who doing their own thing. The target I commited for my device only covers 1/3rd of the device space, I locally maintain the other 2 slot functionality
<Grommish>
doesn't mean it should be part of the repo
ivanich has joined #openwrt-devel
lipnitsk[m] has joined #openwrt-devel
<Grommish>
but I can host it as a fork, I can and have run forks of the packages defined for the custom cpu arch setting
<ynezz>
what about just refreshing the patches in the CI, providing the diff output as downloadable artifact with copy&paste instructions in curl|git am format? Some projects are doing similar for failed code style checks for example
<ynezz>
so even if the CI check fails, author has relatively easily actionable fix in hand
<mangix>
sounds pointless
<mangix>
you can just run make package/x/refresh
<mangix>
much easier than git am
<Grommish>
If the CI check fails, they've got the logs, no?
<mangix>
aparcar[m]: hahaha well played. The reason the treewide patch missed it is because there's a different directory for patches based on BUILD_PATENTED.
<aparcar[m]>
mangix: uhm?
<aparcar[m]>
tell me more
<aparcar[m]>
is make refresh always "safe"?
<Borromini>
hey guys. i set the DTS to extract the MAC from the device's ART partition, a dump shows it's there, but ag71xx tells me 'invalid mac' and uses a random one. what am i doing wrong? https://paste.debian.net/plain/1186293/
<Borromini>
00 22 3f 0d 06 12 is what the MAC is
<mangix>
It is safe yes.
<mangix>
aparcar[m]: those patches are used when BUILD_PATENTED is disabled. I'm guessing lipnitsk has it enabled locally.
<mangix>
fdk-aac is one of those problematic packages
<aparcar[m]>
okay anything I should do?
<mangix>
No.
<mangix>
The PR is fine
<aparcar[m]>
great fftm
feriman has quit [Quit: WeeChat 3.0.1]
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<Borromini>
i gotta say i am appreciating the blessings of git commit --amend
<rsalvaterra>
Borromini: How have you lived until now without it? :)
<Borromini>
rsalvaterra: lots of git resets and wondering why :P
<rsalvaterra>
My greatest epiphany was when I discovered git rebase -i. :)
<Borromini>
what does that do
<Borromini>
interactive rebase?
<rsalvaterra>
Oh, my…!
<rsalvaterra>
Yes, interactive rebase.
<Borromini>
ok. i'll try that one time :P
<rsalvaterra>
Borromini: No. You will understand what it does, then you will do it all the time, especially when iterating patch series. ;)
<rsalvaterra>
It's so incredibly powerful it almost feels like magic.
<Borromini>
ok
Tapper has quit [Ping timeout: 240 seconds]
<rsalvaterra>
And now to try and build my Omnia image again… hopefully make clean solved my tor/openssl errors…
<Borromini>
;)
<Borromini>
seems eth0 is coming up but no mac, eth1 isn't, but wiphy is detected
<Borromini>
not too bad for my first own dts i'd say :P
feriman has joined #openwrt-devel
<ldir>
interactive rebase with the 'fixup' command - perfect commits first time
gladiac133779 has quit [Quit: k thx bye]
<Borromini>
neat
Acinonyx has joined #openwrt-devel
<PaulFertser>
Borromini: the trick used by old ar71xx code is to use a big offset, so no matter what the flash size is, the extra top bits will be ignored, and you'll end up referencing 0x10000 before the end of the flash. And 0x1000 into it.
<Borromini>
PaulFertser: ok, thanks
gladiac133779 has joined #openwrt-devel
<Borromini>
i'll need some help getting eth1 up but one step at a time :)
Acinonyx_ has quit [Ping timeout: 240 seconds]
<rsalvaterra>
mangix: Hm… even after make clean, tor-basic fails to build. :/
<Pepe>
rsalvaterra: Can you create an issue in packages feed? My colleague is updating it regularly and having some log output would be nice. I will ask him to take a look at it anyway.
<rsalvaterra>
Pepe: Will do. I was registering to create an issue in the Tor project itself. :P
<Pepe>
That's will do as well!
<rsalvaterra>
Done for the packages feed.
<rsalvaterra>
Yeah, but registration is moderated, so it could take a while before it's even approved… :/
guimo has quit [Remote host closed the connection]
Night-Shade has quit [Read error: Connection reset by peer]
<Borromini>
hey guys. trying to port a wnr2000 v1 from ar71xx to ath79. it's an ar9132 device according to what i could find, with an ar8126 switch. ar71xx had a ath79_eth0_data.has_ar8216 setting that i'm not finding an ath79/dts equivalent for?
Strykar has quit [Remote host closed the connection]
Strykar has joined #openwrt-devel
Night-Shade has quit [Ping timeout: 240 seconds]
dedeckeh has quit [Quit: Connection closed]
Night-Shade has joined #openwrt-devel
rmilecki has quit [Read error: Connection reset by peer]
rmilecki has joined #openwrt-devel
feriman has quit [Ping timeout: 240 seconds]
rmilecki has quit [Ping timeout: 246 seconds]
kontaxis has quit [Remote host closed the connection]
pine127 has quit [Remote host closed the connection]
kontaxis has joined #openwrt-devel
pine127 has joined #openwrt-devel
<bkallus>
I have a BCM63xx router to which i'm trying to port openwrt. the stock firmware doesn't have any tools to dump the nand, and i'd like to make a backup. I have it booting from initramfs images over tftp, but openwrt can't see the mtd partitions. How do I tell the boot process where the mtd partitions are?
<bkallus>
i have a root shell on the stock fw, so i am able to read /proc/mtd.
<lipnitsk>
bkallus: off the top of the head - check your devicetree config, nand driver.
<PaulFertser>
bkallus: but is it able to see the NAND at all? Because if you just need to dump all of it probably you do not need partitions. Sometimes the layout is hardcoded, sometimes there's some kind of partition table, depends on the vendor.
<zx2c4>
lipnitsk: im trying to repro that guy's issue...
<lipnitsk>
yeah I'm having no luck. iperf fail is probably a red herring, but maybe not - 5.10 support is still in early stages. There are a lot of things enabled in his config and a lot of variables - which SHA is he building from, at what SHA are his packages, etc..
kontaxis has quit [Remote host closed the connection]
pine127 has quit [Remote host closed the connection]
<lipnitsk>
That said, the WireGuard makefile changes are completely independent of linux 5.10 and a minimal build with luci-app-wireguard for his target works just fine
<zx2c4>
"If that is directed at me, my presumption is that what I am seeing directly related to this PR. if I did not think that I would not be posting here. But if you think that a real config failing just by enabling what this PR targets..."
<zx2c4>
so :shrug:
<lipnitsk>
I'll spend a bit of time on it just to cover our bases and maybe to report some new unrelated issue.
kontaxis has joined #openwrt-devel
<zx2c4>
lipnitsk: so far so good
<zx2c4>
mine built iperf successfully too
<lipnitsk>
good. you took his config and disabled ccache?
<zx2c4>
i checked out your branch
<zx2c4>
`git-am`'d those patchwork patches onto it
<zx2c4>
copied his config to .config
<zx2c4>
enabled wireguard
<zx2c4>
hit make
<lipnitsk>
okay, I'm gonna try the same but without wireguard first to see if I can get to where he got in the second paragraph, then enable luci-app-wireguard and try again.
<rsalvaterra>
tmn505: ping
jlsalvador has quit [Remote host closed the connection]
jlsalvador has joined #openwrt-devel
<rsalvaterra>
I'm completely baffled by the fact that one file in the 5.10.16 kernel we download differs from the one in gregkh's 5.10.16 tagged commit. This is pre-quilt patching, obviously. Can someone enlighten me, please?
<lipnitsk>
we download it from kernel.org and validate the hash, so you should ask there :)
SwedeMike has quit [Ping timeout: 272 seconds]
<rsalvaterra>
I do make target/linux/{clean,prepare} V=s QUILT=1. The armada-3720-espressobin-v7.dts is different. What gives…? o_O
jlsalvador has quit [Remote host closed the connection]
jlsalvador has joined #openwrt-devel
jlsalvador has quit [Remote host closed the connection]
jlsalvador has joined #openwrt-devel
victhor has joined #openwrt-devel
Risk64 has quit [Quit: Risk64]
<zx2c4>
lipnitsk: i managed to repro!
<bkallus>
PaulFertser, lipnitsk: Thanks. It's not able to see the nand at all yet. I'll get that working first.
<lipnitsk>
zx2c4: good job! how?
<zx2c4>
lipnitsk: https://א.cc/pUfQNQ0J
<lipnitsk>
some weirdness with kpp then? I'm still building the no-wg version, will add luci-app-wireguard and try to get it here too
<lipnitsk>
it sort of makes sense then, since the minimal build probably needs no KPP, but when it is needed things break
<zx2c4>
DEPENDS+=+CRYPTO_KPP:kmod-crypto-kpp
<zx2c4>
i wonder if that's working
SwedeMike has joined #openwrt-devel
<lipnitsk>
well you could try removing 'CRYPTO_KPP:' to force it
<Hauke>
rsalvaterra: it would be very strange if they are different
<rsalvaterra>
I'm refreshing all the patches again, because it seems I can't trust the kernel tree… :/
pine127_ has joined #openwrt-devel
<tmn505>
rsalvaterra: check the files
<tmn505>
directory
<tmn505>
You have to delete some of the dts which are upstream
<rsalvaterra>
*facepalm*
kontaxis has quit [Ping timeout: 268 seconds]
<rsalvaterra>
sigh…
<Hauke>
it overwrites the file with target/linux/mvebu/files/arch/arm64/boot/dts/marvell/armada-3720-espressobin-v7.dts
<Hauke>
rsalvaterra:
pine127 has quit [Ping timeout: 268 seconds]
<tmn505>
that's why I said they have to be split for 5.4 and 5.10
kontaxis has joined #openwrt-devel
<rsalvaterra>
Yeah, I obviously understand the problem now.
<rsalvaterra>
Sorry for the noise… :/
<tmn505>
np
<lipnitsk>
zx2c4: still having package build failures... though not iperf anymore. did you run ./scripts/feeds update -a && ./scripts/feeds install -a ? It's either some package discrepancy or my host system is just too bleeding edge and some issue there (I'm running Arch)
Tapper has quit [Ping timeout: 240 seconds]
<zx2c4>
lipnitsk: yea, i did
Night-Shade has quit [Read error: Connection reset by peer]
<lipnitsk>
getting "/usr/lib/libpython3.9.so: file not recognized: file format not recognized" while building uwsgi now... maybe because it's a symlink on my system?
<lipnitsk>
/usr/lib/libpython3.9.so: symbolic link to libpython3.9.so.1.0
<lipnitsk>
$ file /usr/lib/libpython3.9.so
<lipnitsk>
why is it even trying my system python? It should be using ./staging_dir/target-arm_cortex-a9+vfpv3-d16_musl_eabi/usr/lib/libpython3.9.so