<f00b4r0>
enyc: it appears it's too late for that. I wouldn't have bypassed jow anyway. It's too bad because there were some important fixes in ucode that won't be in 23.05.3 afaict.
<enyc>
f00b4r0: what does that afect in usage?
<enyc>
f00b4r0: yes; i've discovered 23.05.3 building towards release or whatever
<f00b4r0>
netlink usage and well as error reporting
<f00b4r0>
(that I know of)
<enyc>
f00b4r0: in what ways? and is this longstanding issue or a recent regression?
<f00b4r0>
sorry busy right now. Changelog has the details
Mangix has quit [Read error: Connection reset by peer]
<mrkiko>
Ansuel: on what devices did you test this? Asking to understand if it would be useful for you to test this on a FB7530 or if you did so already
minimal has joined #openwrt-devel
<mrkiko>
Ansuel: didn't know you fixed VDSL :D re-compiling now
rua has quit [Quit: Leaving.]
<mrkiko>
Ansuel: maybe I can paste the failure on the PR
<mrkiko>
I guess it's a bug if the download script checks the hash of files it produced in a previous step by packing git checkouts...
<mrkiko>
even because any time the hash is different so clearly something is going wrong
<mrkiko>
PaulFertser: thanks a lot. Yeah, I will need quilt. But first of all I'll need to get this to compile :D
<mrkiko>
PaulFertser: no, I'll need quilt. Nothing hurts like when the builsystem plays against you :D
<PaulFertser>
mrkiko: probably when it's the compiler? Or assembler?
<mrkiko>
PaulFertser: well yeah
<mrnuke>
QCA-SSDK is magic! It doesn't compile at firt. You fix a bunch of issues and it compiles. You git stash, start from scratch, and compiles again!
<mrnuke>
What a @#$%^&* nightmare
<robimarko>
SSDK is pure chaos
<robimarko>
Its magic that it even compiles
<mrnuke>
It's #ifdef heaven. Very tasty spaghetti code!
<mrnuke>
I got qca-ssdk built for ipq95xx. Now trying qca-nss-dp. It's trying to include files that are not even in the source tree!
<mrnuke>
Then edma-v2/edma.h uses some defines tha are defined in ... soc_ops. IN a header that nobody #includes
<mrnuke>
Even an intern using cuneiform would do betther than that if you just took away the preprocessor
<mrnuke>
robimarko: BTW, your prior work and buildroot repo have been extremely helpful thusfar
<robimarko>
mrnuke: oh nice to see somebody use that, it really helps with upstreaming stuff and running latest kernels
<mrnuke>
I like that I can use LINUX_OVERRIDE_SRCDIR and essentially use a linux clone with full git functionality
Borromini has joined #openwrt-devel
<mrnuke>
robimarko: any iudea what this "ppe_drv" is that I see referenced in edma-v2 (qca-nss-dp) ?
<robimarko>
mrnuke: Its the package processor engine
<robimarko>
There is a separate driver for it in NSS-DRV as far as I remember
<robimarko>
Actually, its a separate repo these days
<mrnuke>
Fun! I have to pull that into the buildroot!
<mrnuke>
Thanks you!
Borromini has quit [Ping timeout: 480 seconds]
Borromini has joined #openwrt-devel
<mrkiko>
PaulFertser: even tough I do "quilt new ...", then "quilt add", quilt seems to not see my changes to the specified file. It worked fine on a file under src before, but not now where I am editing a file in src/common/something
<PaulFertser>
mrkiko: you should use "quilt edit" for changing the file :/
<mrkiko>
PaulFertser: so the wiki isn't fully correct, since "edit" is decribed as ashortcut, while it is not only a shortcut
<PaulFertser>
mrkiko: ah yes, if you do quilt add prior to changing you can change it by other means.
<mrkiko>
PaulFertser: but apparently I am doing something wrong somewhereelse and quilt won't see that change. mhmmm
<mrkiko>
well, in the end it worked so it's something I am overlooking. But overall nice
Borromini has quit [Ping timeout: 480 seconds]
<mrkiko>
it was me, I was using the wrong pakage. Should admit I dnd't choose an easy target to start learning :D
Borromini has joined #openwrt-devel
Borromini has quit []
fakuivan has quit [Remote host closed the connection]
<mrkiko>
so - we have a problem of implicit enum conversions and at the moment the scenario looks like the following: a set of macros returns values that are implicitly casted to different return types, depending on the function we are, sometimes dsl_boolean_t, sometimes (most of the times) dsl_error_t
<mrkiko>
I am trying to explicitly cast all the call-sites, but I don't now if this is the right approach; other ideas coming to mind are disable the related compiler warning somehow (not going to fly probably) or maybe other C stuff I don't knowof
<oliv3r>
Mrnuke; is this new?!?! I need this :) LINUX_OVERRIDE_SRCDIR
<oliv3r>
Hmm, looks like a built root thing.
<oliv3r>
Though I now think I may have used something like this already
Piraty has joined #openwrt-devel
<Piraty>
aparcar: re https://github.com/openwrt/openwrt/pull/14918 , careful with the date version scheme, some upstreams were found to later use real versions strings later, so it's good advised to 0.0.0.20231231 or similar
<Piraty>
can i read about the rationale behind 1. the move off of openwrt's opkg 2. to move to apk (over, let's say yocto's fork of opkg) ?
<mrnuke>
oliv3r: yes, it's a buildroot thing. not sure if openwrt has something similar
<mrnuke>
This thing really gets me in QCA Makefiles: "Makefile:ccflags-y += -Wall -Werror"
<mrnuke>
Why would they turn on -Werror on code that's already very questionable?