thejoker8814 is now known as Guest2523
thejoker8814 has joined #openwrt-devel
Guest2523 has quit [Ping timeout: 480 seconds]
tSYS has quit [Quit: *squeak*]
tSYS has joined #openwrt-devel
noltari has quit [Quit: Bye ~ Happy Hacking!]
noltari has joined #openwrt-devel
madwoota has joined #openwrt-devel
rz has joined #openwrt-devel
fakuivan has quit [Remote host closed the connection]
fakuivan has joined #openwrt-devel
gch981213 has joined #openwrt-devel
<rmilecki> nbd: pong
<rmilecki> nbd: not yet, sorry
<rmilecki> weekend turned out really busy for me
goliath has joined #openwrt-devel
jkl has quit [Quit: Gone.]
jkl has joined #openwrt-devel
<KGB-2> https://tests.reproducible-builds.org/openwrt/openwrt_mpc85xx.html has been updated. (100.0% images and 100.0% packages reproducible in our current test framework.)
tidalf has joined #openwrt-devel
rz has quit [Ping timeout: 480 seconds]
goetz has joined #openwrt-devel
robimarko has joined #openwrt-devel
cmonroe has quit [Ping timeout: 480 seconds]
<rmilecki> nbd: i just compiled the latest master, tested it, dropped 600-net-disable-GRO-by-default.patch and tested again
<rmilecki> dropping 600-net-disable-GRO-by-default.patch lowers NAT performance
<rmilecki> 2 more runs to go to share results
rsalvaterra has joined #openwrt-devel
<nbd> is that with or without flow offloading?
<rmilecki> very default OpenWrt config, no flow offloading
<rmilecki> eec18118d02f ("mpc85xx: convert WS-AP3710i to simpleImage wrapper"): bcm53xx with 6.1.86 *with* 600-net-disable-GRO-by-default.patch
<rmilecki> [ 3] 0.0-60.0 sec 5.20 GBytes 744 Mbits/sec
<rmilecki> [ 3] 0.0-60.0 sec 5.29 GBytes 757 Mbits/sec
<rmilecki> [ 3] 0.0-60.0 sec 5.25 GBytes 751 Mbits/sec
<rmilecki> eec18118d02f ("mpc85xx: convert WS-AP3710i to simpleImage wrapper"): bcm53xx with 6.1.86 *without* 600-net-disable-GRO-by-default.patch
<rmilecki> [ 3] 0.0-60.0 sec 4.30 GBytes 616 Mbits/sec
<rmilecki> [ 3] 0.0-60.0 sec 4.32 GBytes 618 Mbits/sec
<rmilecki> [ 3] 0.0-60.0 sec 4.32 GBytes 618 Mbits/sec
<rmilecki> it's NAT speed for BCM47094 client running iperf -c
<rmilecki> and iperf -s running on WAN-connected machine
<rmilecki> sorry, i meant BCM47094 connected laptopt running iperf -c
<rmilecki> [iperf -s laptop] ----- wan [BCM47094 router] lan4 ----- [iperf -c laptop]
<nbd> could you please make a perf flamegraph without 600-net-disable-GRO-by-default.patch?
<rmilecki> will try to find time for that
<nbd> thanks
<nbd> my guess is the extra csum check is the main source of slowdown, since unfortuntately bcm53xx does not have rx checksum offload
<nbd> it's really weird that all this time broadcom didn't bother to add such a simple useful feature to the ethernet mac
<robimarko> QCA messed up IPQ807x/IPQ60xx with that as well, but I am sure it has it but its not exposed in the driver
<nbd> rmilecki: what's the name of the device that you tested with?
rsalvaterra has quit []
<rmilecki> Luxul XWR-3150 V1
<rmilecki> bcm47094-luxul-xwr-3150-v1.dts
Mangix has quit [Read error: Connection reset by peer]
goetz has quit [Quit: goetz]
goetz has joined #openwrt-devel
enyc has joined #openwrt-devel
<owrt-images-builds> Build [#207](https://buildbot.openwrt.org/images/#/builders/4/builds/207) of `master_mvebu/cortexa53` failed.
<Ansuel> i expected more package to be broken... that is nice not that bad
rua has quit [reticulum.oftc.net helix.oftc.net]
gladiac has quit [reticulum.oftc.net helix.oftc.net]
enyc has quit [reticulum.oftc.net helix.oftc.net]
Ansuel has quit [reticulum.oftc.net helix.oftc.net]
tmn505 has quit [reticulum.oftc.net helix.oftc.net]
Forst has quit [reticulum.oftc.net helix.oftc.net]
sa7mfo has quit [reticulum.oftc.net helix.oftc.net]
owrt-images-builds has quit [reticulum.oftc.net helix.oftc.net]
slh has quit [reticulum.oftc.net helix.oftc.net]
sandberm has quit [reticulum.oftc.net helix.oftc.net]
ILEoo has quit [reticulum.oftc.net helix.oftc.net]
enyc has joined #openwrt-devel
sandberm has joined #openwrt-devel
rua has joined #openwrt-devel
gladiac has joined #openwrt-devel
Ansuel has joined #openwrt-devel
tmn505 has joined #openwrt-devel
Forst has joined #openwrt-devel
owrt-images-builds has joined #openwrt-devel
sa7mfo has joined #openwrt-devel
ILEoo has joined #openwrt-devel
slh has joined #openwrt-devel
gladiac has quit [Write error: connection closed]
gladiac has joined #openwrt-devel
aparcar has quit [Ping timeout: 481 seconds]
tohojo has quit [Ping timeout: 480 seconds]
schmars[m] has quit [Ping timeout: 480 seconds]
gladiac is now known as Guest2572
Guest2572 has quit [Write error: connection closed]
robimarko has quit [Ping timeout: 480 seconds]
hitech95 has joined #openwrt-devel
robimarko has joined #openwrt-devel
KGB-2 has quit [Remote host closed the connection]
KGB-2 has joined #openwrt-devel
tidalf has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
schmars[m] has joined #openwrt-devel
<schmars[m]> blocktrron this probably didn't make it through before NickServ kicked me: about your mt7915e txq lock patch https://github.com/openwrt/mt76/issues/690 - i have two zyxel nwa55axe with this issue. did you find more info since february, or should i just go an try the patch?
* Slimey tickles rmilecki with a mpc85xx CPU
cmonroe has joined #openwrt-devel
mattsm has quit [Ping timeout: 480 seconds]
tidalf has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
cmonroe has quit [Read error: Connection reset by peer]
cmonroe has joined #openwrt-devel
goetz has quit [Read error: Connection reset by peer]
goetz has joined #openwrt-devel
<rmilecki> Slimey: you probably meant someone else, i've zero knowledge of that SoC/CPU
tidalf is now known as Guest2605
cmonroe has quit [Ping timeout: 480 seconds]
Guest2605 has quit [Ping timeout: 480 seconds]
cmonroe has joined #openwrt-devel
fakuivan has quit [Ping timeout: 480 seconds]
rz has joined #openwrt-devel
<Slimey> lol k
<Slimey> WS-AP3710i is simlar to a bsap-2030 i have
tidalf has joined #openwrt-devel
tidalf has quit []
tidalf has joined #openwrt-devel
goliath has joined #openwrt-devel
torv has quit [Remote host closed the connection]
torv has joined #openwrt-devel
<svanheule> mrnuke: No idea, sorry. I don't suppose you can correlate it to a certain setting?
<mrnuke> svanheule: I found a command in RTL dialect that changes said byte, but that's about it. Not sure what it does.
<svanheule> I was looking at the wiki earlier today, and guessed it might have something to do with 'power-on after classification', but that's just speculation
<svanheule> you haven't tried the different values to see if there's any obvious effect?
gladiac has joined #openwrt-devel
<mrnuke> svanheule: I did. Couldn't correlate behaviour with value of said settings
<mrnuke> Originally, I thought it might be "port classification enable". But then I'd expect a command in the vendor fw to set the "clasification enable" byte.
<mrnuke> None of the commands the vendor fw sends change the "classification_enable", so clearly, this must be something else
aiyion has quit [Ping timeout: 480 seconds]
aiyion has joined #openwrt-devel
tidalf has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
tidalf has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
Mangix has joined #openwrt-devel
Mangix has left #openwrt-devel [https://quassel-irc.org - Chat comfortably. Anywhere.]
Mangix has joined #openwrt-devel
tidalf has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
robimarko has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
<Mangix> Ansuel: ERROR: package/utils/audit [host] failed to build.
<Mangix> missing stdio.h include. ROFL.
tidalf has joined #openwrt-devel
tidalf has quit []
tidalf has joined #openwrt-devel
tidalf has quit []
<Mangix> ugh GCC14 breaks so many packages
tidalf has joined #openwrt-devel
<Ansuel> wtf missing stdio.h o.O
<Ansuel> with mbedtls and gcc14 we will have lots of FUN.
tidalf has quit []
<Mangix> Ansuel: with every gnulibc and gcc update, fewer and fewer headers are implicitly included
<Mangix> it's funny
<Mangix> musl is supposed to be like that. but then glibc says hold my beer.
<Ansuel> well we are changing approach...
<Ansuel> lots of time ago it was like "sure lets include all this handy shit"
<Ansuel> now it's more "are you sure you want to include function x? It may cause CVE XYZ in the next 10 years"
<Ansuel> meanwhile every python project
<Ansuel> "library printf_but_with_colored_output GO BRRRRR"
tidalf has joined #openwrt-devel
tidalf has quit []
<Ansuel> honestly we are just getting all the harm of years of bad decision on these core library (and musl being better somehow since it was more restrictive from the start)
<Ansuel> but still dropping stdio.h IS FUNNY AHAHAH invalidates 99.99999% of every C guide online ahahah
<Mangix> LOL
<Mangix> GCC14 also includes the algorithm header in fewer libraries
<Mangix> /s/libraries/headers
<Mangix> it's kind of funny. C2x behaviors are not errors on GCC14
<Mangix> *now
<Ansuel> btw very tempted to recover my project of using alpine container for openwrt buildroot
<Mangix> it's interesting that -Wimplicit-function-declaration is now an error. this will break so many automake stuff.
<Mangix> Ansuel: funny I actually tried this recently
<Ansuel> how it went?
<Mangix> undefined X in libtinfo.6
<Mangix> in mkhash or something
<Mangix> something in scripts
tidalf has joined #openwrt-devel
<Ansuel> Mangix they use implicit function to check if something is supported?
<Mangix> eh no. They check functions existence by seeing if they compile and link. But with this error, they will not compile because most of the time, the header is not included.
<Mangix> In meson terms, cc.has_function('fseeko') will fail. cc.has_function('fseeko', prefx: '#include <unistd.h>') will succeed
<Ansuel> ok yes i was referring to that... it will be a shitfest...
tidalf has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
tidalf has joined #openwrt-devel
SlimeyX has joined #openwrt-devel
<Mangix> uhhh
<Mangix> I just tried Alpine and I'm not getting that error
<Mangix> Let's see how far it goes
<Mangix> ERROR: tools/cmake failed to build. That didn't take long
<Mangix> fatal error: linux/fs.h: No such file or directory
<Mangix> LOL
<Ansuel> you want to suffer sir
<Ansuel> you need linux-headers
<Ansuel> actually all of this
<Ansuel> gcc g++ ncurses-dev git rsync perl tar bash patch findutils diffutils grep gzip unzip bzip2 wget python3 file which make gawk coreutils linux-headers argp-standalone musl-fts-dev musl-obstack-dev musl-libintl
tidalf has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<Mangix> installed most of thoser
tidalf has joined #openwrt-devel
<Mangix> funny thing about that header is that it's not really needed for cmake.
<Ansuel> wonder if the musl fix for binutils was backported...
<Mangix> I was thinking of writing a patch that matched it against __has_include
tidalf has quit []
<Ansuel> also this meme is needed
<Ansuel> and then i stopped
<Ansuel> can't remember why
<Mangix> IIRC elfutils iconv shennanigans
<Mangix> or intl.
tidalf has joined #openwrt-devel
<Ansuel> do you think those totally doesn't support musl ?
<Mangix> hmm?
tidalf has quit []
<Mangix> Ansuel: tools built.
<Mangix> Let's see if toolchain builds
tidalf has joined #openwrt-devel
<Ansuel> even 7zip ?
<Ansuel> o.O
<Mangix> I don't see it
<Mangix> is it conditionally built?
<Ansuel> yep
<Mangix> Maybe I should enable CONFIG_BUILD_ALL_HOST_TOOLS
<Ansuel> yes if you want to suffer
<Ansuel> that is what i'm testing in ci
<Mangix> yeah I'm looking at it. Not looking forward to llvm
<Ansuel> anyway if you can continue with your normal compile
<Ansuel> also can't remember does musl provide better perf ?
<Mangix> In specific circumstances
<Mangix> IIRC for x86 and aarch64, there are a lot of ASM optimizations in glibc
tidalf has quit []
tidalf has joined #openwrt-devel
tidalf has quit []
tidalf has joined #openwrt-devel
<Ansuel> wait 7zip now compile o.O
<Ansuel> make[2] -C tools/llvm-bpf clean-build
<Ansuel> make[2] -C tools/llvm-bpf compile
<Ansuel> OH NO
<Mangix> hahaha
<Mangix> ERROR: package/feeds/luci/luci-base [host] failed to build.
<Mangix> LOL
tidalf has quit []
<Ansuel> so toolchain compiled correctly?
<Mangix> yesa
<Ansuel> nice
<Ansuel> BUT WTF WHY LUCI FAILED?
<Ansuel> curious of the log
<Ansuel> pls send
<Mangix> /usr/lib/gcc/x86_64-alpine-linux-musl/12.2.1/../../../../x86_64-alpine-linux-musl/bin/ld: po2lmo.o: relocation R_X86_64_32 against `.rodata' can not be used when making a PIE object; recompile with -fPIE
<Mangix> /usr/lib/gcc/x86_64-alpine-linux-musl/12.2.1/../../../../x86_64-alpine-linux-musl/bin/ld: failed to set dynamic section sizes: bad value
<Ansuel> easy to fix i guess
<Ansuel> why the thing worked with before tho ???
<Mangix> no idea
<Ansuel> or it's just alpine having PIE enabled by default
tidalf has joined #openwrt-devel
<Mangix> I imagine some library is not built with PIC
<Ansuel> well if toolchain compile correctly target stuff should not be affected by alpine
<Ansuel> ERROR: tools/llvm-bpf failed to build.
<Ansuel> -.-
<Ansuel> REALLY -.-
<Mangix> right, but HOSTPKG :)
<Ansuel> llvm not supporting musl...
<Mangix> Ansuel: didn't you hear. LLVM has its own libc.
tidalf has quit []
<Ansuel> -.-''''
tidalf has joined #openwrt-devel
hitech95 has quit [Read error: Connection reset by peer]
<Ansuel> sad cause i think all the tools compile correctly except llvm
<Ansuel> using alpine will permit us to catch all the case where one package use host package...
<Mangix> fun. adding fPIE or fPIC does not work
<Mangix> agh
<Ansuel> are you sure the thing is correctly called?
<Mangix> my mind is telling me to rewrite this makefile in meson
<Ansuel> mhhh maybe it's missing on linking stage?
<Mangix> hmm? CC is used to link, no?
<Ansuel> or maybe the library that make use of that is not compiled with fpie
<Mangix> pie is for executables.
<Mangix> I think I'll rewrite in meson at some point.
<jow> po2lmo.o is used both by an executable and a library
<Ansuel> error from llvm
<Ansuel> ../llvm/lib/Support/raw_ostream.cpp:808:11: error: '::lseek64' has not been declared; did you mean 'lseek'?
<Ansuel> | ^~~~~~~
<Ansuel> 808 | pos = ::lseek64(FD, off, SEEK_SET);
<Ansuel> | lseek
<Mangix> Ansuel: missing _LARGEFILE64SOURCE or w/e it was
<Ansuel> LFS64 functions are deprecated
<Ansuel> we had that on openwrt too
<Mangix> Ansuel: maybe everything is building for me as I'm on an older Alpine
<Ansuel> using alpine:latest
<Mangix> Can't figure out how to update it
<Mangix> I'm using distrobox
<Mangix> Ansuel: make menuconfig fails with Error loading shared library libtinfo.so.6: No such file or directory (needed by /home/mangix/devstuff/openwrt/scripts/config/mconf)
gch981213 has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
<Ansuel> mangix maybe ncurses-libs ?
<Mangix> installed
<Mangix> Alpine does not have libtinfo, at all
<Mangix> I haven't looked into why mconf links to it
<Mangix> jow: can't figure out how to fix this in the meantime
<jow> Mangix: ?
<Mangix> fPIE error with luci-base
<jow> couldn't figure out where -fPIE even comes from
<jow> cursory look in rules.mk and include/ did yield nothing
<jow> Mangix: the error you pasted complains about PIE, not PIC
<jow> seems -fPIE get's passed through LDFLAGS
<Ansuel> anyway time for bed will have some fun tomorrow with llvm
<jow> Mangix: so po2lme gets linked with -fPIE (due to $(LDFLAGS)) but all objects are compiled without -fPIE
<Mangix> fPIE is for executables, no?
<jow> *shrug*
<Mangix> I assume libraries need fPIC
<jow> seems when you link an executable with -fPIE then all objects linked into it need to be built with -fPIE too
<jow> the error message you pasted clearly states it
<jow> po2lmo.o: relocation ... can not be used when making a PIE object; recompile with -fPIE
<jow> this is because po2lmo (not the object but the executable `po2lmo`) somehow gets linked with -fPIE
<jow> something passes -fPIE through LDFLAGS
<Mangix> env
<Mangix> hmmm. how do I check LDFLAGS?
<Mangix> jow: added fPIE everywhere now. Still not working.
<jow> the clean target in that makefile is incomplete
<jow> it does not delete po2lmo.o in the toplevel dir (only lib/*.o)
<jow> that might explain your issue
<Mangix> I'm running rm -rf build_dir/hostpkg/luci-base between runs
<Mangix> wait a minute
<Mangix> why is there a ./feeds/luci/modules/luci-base/src/po2lmo.o ?
<Mangix> Untracked files: modules/luci-base/src/plural_formula.c
<Mangix> modules/luci-base/src/plural_formula.h ???
<jow> wrong build dir, apparently
<Mangix> it's building it the feeds dir not build_dir
<jow> it's due to the way Host/Compile is declared
<jow> needs an additional prepare step to copy src/ into bild_dir
<Mangix> hrm
<Mangix> anyway, fixed now
<Ansuel> will it work :o
<Mangix> now ERROR: package/feeds/packages/libtirpc [host] failed to build. :)
<Ansuel> :D
<Mangix> oh that's actually my fault
<Mangix> fatal error: sys/queue.h: No such file or directory :(
<Mangix> I'm somewhat curious if it makes sense to add to prereq-build. It's not needed for base, but is needed for packages
<Ansuel> well fix for that is to ship a custom one...
<Ansuel> we patch musl to provide it
<Mangix> correct. I've been thinking of removing it.
<Mangix> Problerm: too many packages break.
<Ansuel> sad :(
<Mangix> ERROR: package/feeds/packages/kadnode failed to build. <-- mbedtls
<Mangix> ERROR: package/feeds/packages/rust [host] failed to build. <-- ROFL
<Mangix> well, RIP experiment
<Mangix> although for basic builds, Alpine should work now