<Mangix>
alright, cc.find_function() is not failing with implicit function declaration
<Mangix>
weird
<Ansuel>
mhhhh interesting
<Ansuel>
maybe there is a specific handling for that in the implicit function warning?
<Mangix>
I think it's enabled with -Wall
<Mangix>
which these tests don't add
<Mangix>
or maybe building with O0 silences these
<Mangix>
anyway, whatever the reason, most failures I've seen are in the code itself, not in the build system.
rsalvaterra has joined #openwrt-devel
<Mangix>
alacodecontext.c:2700:33: error: assignment to 'gchar **' {aka 'char **'} from incompatible pointer type 'const gchar * const*' {aka 'const char * const*'}
<Mangix>
eww
<Mangix>
unfortunately softethervpn looks like a disaster to fix.
<Ansuel>
3000 warning?
<Mangix>
softethernetvpn uses CRLF for its files, which makes creation of patches a nightmare. In addition, OpenWrt uses libressl, which softethervpn is not compatible with
<Mangix>
basically, the patch files need to be mixed LF and CRLF
<Mangix>
openssl/conf.h is broken or something like that
tidalf has joined #openwrt-devel
tidalf has quit []
<rsalvaterra>
Mangix: iproute2 pull at 6.9.0, if you'd like to give it a go. This time I build tested all possible iproute2 utilities, just to make sure. :)
<jow>
Mangix: might make sense to reintroduce it as PKG_FIXUP
guidosarducci has joined #openwrt-devel
<Namidairo>
\x: BE5000 is Airoha EN7563 + MT7992. looks like the soc is like the other airoha in that it only runs in 32-bit mode. don't hope for quick support unless they copy pasted entire mediatek cpu blocks again. (who am I kidding they probably did)
slh has joined #openwrt-devel
mrnuke has quit [Read error: Connection reset by peer]
* f00b4r0
notices a duplicate "mode" key for each interface.config in the output of ubus call network.wireless status
robimarko has joined #openwrt-devel
mattsm has quit [Ping timeout: 480 seconds]
Mangix has quit [Read error: Connection reset by peer]
<mrkiko>
With airoha I am feeling something that doesn't happen often to me - I am amazed by the 32-bit limitation :D
<mrkiko>
I am not a "64-bit or nothing" fan at all... but still it feels little bit sad
<robimarko>
Its a totaly arbitrary limit they imposed in the boot FW
<robimarko>
Other than to cause product segmentation it makes 0 sense
<Ansuel>
my best bet is that they have some driver what makes use of 32bit address and they don't want to rework the thing to support 64bit
<mrkiko>
I always tought that while proprietary wi-fi drivers may not be written as well as opn-source ones, they tend to work with no issues. Reading the official thread + Openwrt thread about the gl-mt6000, I should reconsider...
<mrkiko>
they sometime may have problems as well
<robimarko>
The only difference is that they have more people working on them
<Namidairo>
mediatek/airoha/econet all used interchangeably anyway (they own it all), and it ships with hnat and warp modules with mediatek written all over it
<\x>
so these airohas are a53 at the very least, so can we run armv8l on it?
<Ansuel>
mrkiko they work with no issues as they have control of all the chain... most of the time open source driver are a lot of reverse engineer, guessing and searching.... having access to all kind of datasheet, contacts with the hw teams would result in a much different driver...
<\x>
so like 32bit with newer armv8 instructions and possibly crypto?
<Namidairo>
the kconfig targets cortex-a7 so unless they did something horrible to the toolchain i want to say no
<Ansuel>
no idea for other producer but everything from QCOM can be debugged to the level of reproducing the very same state of the machine in the crashing instant. (they dump EVERYTHING and can apply back EVERYTHING... even deeper to L2 cache... ipq soc have a very reduced state but mobile soc with debug enable dumpes GB of data on crash with the correct registers applied)
<mrkiko>
Ansuel: and that's awesome to an extent. That said, I am under the impression they focus a lot on low-level stuff and do very good stuff overall, leaving other parts in a less optimal state (I am under the potentially wrong impression that the wi-fi firmware is pretty fragile)
<mrkiko>
Ansuel: would like to know more about soc vendors design choices and decisions; I am not interested at all in debates like "this is better than that", just to learn more about how things come to be the way they are
<Ansuel>
it's nothing about focus or all... it's just money money money and give a shit product that almost work... the proprietary driver is full of errata and workaround most of the time
* russell--
just tried building for ubiquiti erx and it appears the kernel is too big: WARNING: Image file /home/openwrt/src/lede/build_dir/target-mipsel_24kc_musl/linux-ramips_mt7621/ubnt_edgerouter-x-kernel.bin is too big: 3165615 > 3145728
Mangix has joined #openwrt-devel
<Mangix>
jow: well, it's mostly a workflow issue. I normally git init .;git add .;git commit -m m , make changes, and use git diff to make patches. As git mangles crlf by default, I have to use regular diff
<Mangix>
it might be possible to use git with some config option, not sure.