eduardas has quit [Quit: Konversation terminated!]
PaulFertser has quit [Ping timeout: 272 seconds]
eduardas has joined #openwrt-devel
Night-Shade has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Night-Shade has joined #openwrt-devel
PaulFertser has joined #openwrt-devel
aszeszo has quit [Quit: aszeszo]
aszeszo has joined #openwrt-devel
<rsalvaterra>
One question for the toolchain gurus… HOST_CFLAGS in rules.mk defines the compiler flags for the tools which are going to be built and executed on the host system, right?
anonzadas has joined #openwrt-devel
<olmari>
(sideways footnote to recent group mode change and spam: it will obviously not kick away already joined spambots, but our regular way of kicking away will eventually)
<olmari>
gnhaah.. aand to very wrong channel too, by ne.. sorry for noise...
feriman has quit [Quit: WeeChat 2.9]
aszeszo has quit [Quit: aszeszo]
feriman has joined #openwrt-devel
aszeszo has joined #openwrt-devel
Borromini has quit [Ping timeout: 260 seconds]
<f00b4r0>
lynxis: ping?
zx2c4 has quit [Read error: Connection reset by peer]
zx2c4 has joined #openwrt-devel
victhor has joined #openwrt-devel
<ynezz>
rsalvaterra: yes
<rsalvaterra>
ynezz: Thanks! I'm experimenting with -march=native here.
awgh has quit [Ping timeout: 272 seconds]
awgh has joined #openwrt-devel
koniu has quit [Remote host closed the connection]
<karlp>
but, unsurprisingly, the elements don't exist if I'm setting this up in the render() call,
<karlp>
I was copying some ddns overview, but it's doing it on a .poll(), not in the intial render.
repulse has joined #openwrt-devel
<karlp>
added a form.DummyValue, and overwrote it's o.render() to return the elements.. works ok :)
repulse has quit [Client Quit]
repulse has joined #openwrt-devel
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
dedeckeh has quit [Remote host closed the connection]
MichaelOF has joined #openwrt-devel
dangole has joined #openwrt-devel
<rsalvaterra>
dangole: Hi, there! :)
eduardas has quit [Quit: Konversation terminated!]
<dangole>
rsalvaterra: hi!
<rsalvaterra>
dangole: We were talking about inetd the other day and I forgot to ask you… how about adding socket activation to procd instead? O:)
* rsalvaterra
runs
MichaelOF has quit [Quit: Konversation terminated!]
aszeszo_ has quit [Read error: Connection reset by peer]
aszeszo has joined #openwrt-devel
valku has joined #openwrt-devel
aszeszo has quit [Quit: aszeszo]
aszeszo has joined #openwrt-devel
<dangole>
rsalvaterra: oh yeah, that'd be amazing :)
<rsalvaterra>
dangole: I'm not volunteering, though. :P
<dangole>
i like the idea in general, can be very useful for uhttpd, dropbear and all those occasional-use services
<rsalvaterra>
And it would probably be more secure than inetd, and less heavy than xinetd.
<rsalvaterra>
Anyway, it was just something that came to my mind. No idea at all about how complex such thing would be.
aszeszo has quit [Ping timeout: 260 seconds]
<dangole>
rsalvaterra: depends on how well the services are adapted to that... i also kinda like the idea of having something like inetd (which doesn't need to be as capable as the init process itself) doing that job)
aszeszo has joined #openwrt-devel
aszeszo has quit [Read error: Connection reset by peer]
aszeszo has joined #openwrt-devel
<grift>
right, socket activation does not really make sense if the service doesnt quit on idle
<grift>
except for one-shot services i guess like identd, or git-daemon which fill a request and exit
heffer has quit [Quit: heffer]
heffer has joined #openwrt-devel
<rsalvaterra>
Any receptiveness to building the host binaries with -march=native? It worked just fine here, but I'm assuming host binaries are always run on the machine they were built…
<dangole>
rsalvaterra: not an option as a default, but it does make sense to add this in menuconfig i believe. by default we need generic x86_64 code which runs everywhere as SDK and IB ship compiled binaries
<rsalvaterra>
dangole: Right, I could be overlooking some use cases, hence the question. I'll cook a patch accordingly, then.
<karlp>
is this intended to be "OMG optimzie for my microarch haswell superlake" ?
<dangole>
any ppc users here by any chance? tl-wdr4900 anyone?
<rsalvaterra>
karlp: I propose adding a -Omg optimisation level to GCC. :P
<olmari>
dangole: I have tl-wdr4900 or two in production(s) with openwrt
<rsalvaterra>
dangole: I have an iBook G4, yes. It's bricked at the moment, because I'm an idiot (compressed the initramfs with zstd without proper support in the kernel).
<rsalvaterra>
I'd love to get my hands on a WDR4900, though.
<olmari>
(no other PPC's at use, in computer-wise)
<dangole>
olmari: could you be willing to `opkg update ; opkg install procd-ujail procd-seccomp ; reboot` and report back? (assuming they run snapshot)
<olmari>
dangole: That could be arranged in some days notice.. or likely as in "could possibly try tinker on weekeds on that device"
<dangole>
olmari: i'm hestitating to enable this by default (which we should and discussed previously) without having tested on PPC...
<dangole>
olmari: no rush, would be great help!
<rsalvaterra>
karlp: It's just a couple of lines of code, no big deal… and then we can claim to be Gentoo-compliant. :)
<olmari>
ATM it is self-compiled, likely will be afterwards too, testing something/anything can obviously be whatever
<dangole>
olmari: recent snapshot build would be ideal, just include the before-mentioned packages and enjoy ntpd running non-root and dnsmasq being jailed (and umdns potentially seccomp-filtered, but i'm pretty sure that still needs the ppc syscalls added)
aszeszo has quit [Quit: aszeszo]
muhaha has joined #openwrt-devel
<olmari>
(I love buildroot scripts/env, relates highly to doing stuff, not to actual test that I propably forget until time is :D
jas4711 has joined #openwrt-devel
Katana_Steel has quit [Ping timeout: 240 seconds]
<olmari>
dangole: what should I keep looking then? also is serial-console need anticipated? (wouldn't be first time with wdr4900 ;D )
<dangole>
olmari: serial console won't be needed, at least not for the things i've added and would like you to test (no idea about the general state of master builds for that target though...) just observe that dnsmasq comes up as usual, ntpd should run as user 'ntp' once procd-ujail is installed. if you like, also install umdns and trace the needed syscalls, so seccomp could be used
<iojeda>
So I was wondering what are the steps required to provide support for this device. I will be more than happy to work on it but will need some guidance.
<PaulFertser>
Usually the best way (when possible) is to run initramfs images for testing. And once it's ready you can try flashing overwriting the vendor's firmware.
<jow>
dangole: personally I'm a bit wary about socket activation on embedded devices
<jow>
I'd rather have all services resident in memory to avoid random tcp connections suddenly triggering oom by activating daemons while the system is under memory pressure
<rsalvaterra>
jow: Hm… If you're *that* memory-constrained, does it really matter? :/
<iojeda>
Thanks PaulFertser, where can I find in the documentation 1.- how to build that image. 2.- how to run it on the device without flashing
<PaulFertser>
iojeda: I guess there's no documentation for that, but you can follow the guides for devices using the same SoC, plus the forum threads mentioned on that page.
<PaulFertser>
rsalvaterra: it would be better to know about it in advance.
<iojeda>
Ok I will check thanks, a lot of reading to do :P
<rsalvaterra>
I don't know what's the idle overhead of uhttpd/LuCI, since I haven't used them in years. On my C6, dropbear uses about 1.1 MiB.
ivanich has quit [Quit: Konversation terminated!]
koniu has quit [Remote host closed the connection]
ivanich has joined #openwrt-devel
lukedashjr has joined #openwrt-devel
luke-jr has quit [Ping timeout: 260 seconds]
koniu has joined #openwrt-devel
lukedashjr is now known as luke-jr
jas4711 has quit [Remote host closed the connection]
<f00b4r0>
is there an opkg equivalent to "dpkg -S" ? I'm trying to figure out where a file comes from
<rsalvaterra>
They're SIMD-heavy tests, for sure, but I wasn't expecting a 9 % difference from -march=x86-64. Not bad at all. It would be nice to see timed GCC compilations.
Borromini has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
<aparcar[m]>
jow: do you think using src-git-full could solve the opkg size mismatch issue? I suspect that missing commits in the feed repos cause partly arbitrary SOURCE_DATE_EPOCHS, which could cause these minimal variations of size.
<aparcar[m]>
so phase2 could use src-git-full rather than the shallow clone via --depth=1
dedeckeh has quit [Remote host closed the connection]
iojeda has quit [Remote host closed the connection]
<f00b4r0>
jow: FYI I tried restarting slashdirt-02 but it doesn't seem to be able to connect
<jow>
aparcar[m]: quite likely
<jow>
f00b4r0: odd. will investigate later
Darkmatter66 has joined #openwrt-devel
<aparcar[m]>
Should I create a PR replacing git-src with git-src-full or would that break the builders?
Darkmatter66_ has quit [Ping timeout: 265 seconds]
<jow>
I guess that should work
<f00b4r0>
jow: no emergency. The host has a failing disk (raid1); and I'm still waiting for fiber (was supposed to get it this year but covid threw a spanner in those plans).
<f00b4r0>
I'll leave it running though, if you want to debug the issue
Misanthropos has quit [Read error: Connection reset by peer]