17 Jun 2024 |
James Munns | jesus oof owie | 14:36:42 |
James Munns | diondokter was this the fix for all the spooky stuff you've been chasing recently? | 14:37:01 |
dirbaio | btw nordic's docs say the shmem stuff must be in the lower 128k of ram | 14:37:12 |
dirbaio | but the addrs there are in the upper 128k | 14:37:33 |
diondokter | In reply to @jamesmunns:beeper.com diondokter was this the fix for all the spooky stuff you've been chasing recently? Yep | 14:38:30 |
James Munns | In reply to @diondokter:matrix.org Yep jeeeeeeeeeez | 14:38:40 |
James Munns | Is that lifetime documented in their API? | 14:40:16 |
diondokter | The zephyr code Jonathan looked at when he started this library had this being a stack variable too.
Then they updated the library, but no doc changes | 14:40:25 |
diondokter | In reply to @jamesmunns:beeper.com Is that lifetime documented in their API? Nooopppeee | 14:40:30 |
dirbaio | https://developer.nordicsemi.com/nRF_Connect_SDK/doc/2.1.0-rc2/nrfxlib/nrf_modem/doc/ug_nrf_modem_porting_os.html
The Modem library needs a region of RAM within the first lower 128KB to share with the modem.
| 14:40:35 |
dirbaio | ah, sorry; 0x2001_0000 is in the lower 128kb 🤦♂️ | 14:41:20 |
James Munns | honestly, this would make a REALLY good blog post diondokter: I see a lot of "borrow checker wouldn't help for embedded" noise, this is like literally an IDEAL case where Rust would have made it clear from the function signature | 14:41:20 |
diondokter | In reply to @jamesmunns:beeper.com honestly, this would make a REALLY good blog post diondokter: I see a lot of "borrow checker wouldn't help for embedded" noise, this is like literally an IDEAL case where Rust would have made it clear from the function signature Definitely! That's a good idea. Debugged this with wassasin and I think we spent a total of 20 mandays on it or so | 14:42:49 |
James Munns | yeah, Erik pinged me and I spent an hour on it, it would have taken me a lot to find that too | 14:43:41 |
diondokter | So we found out that a channel write in the middle of nowhere affected stuff today. Makes sense since that config pointer pointed to somewhere in the task pool | 14:44:37 |
diondokter | Glad we're done haha | 14:44:47 |
James Munns | Yeah, get that marketing value out of dev time spent, at least lol | 14:45:21 |
diondokter | Alright, released | 14:47:45 |
thejpster | just wanted to say "yikes" on https://github.com/nrf-rs/nrfxlib/pull/29.
Also the CHANGELOG URLS need changing. | 14:55:51 |
fuse117 | nrf-modem only works with the nrf9160, correct? i can't work with the nrf9161? | 20:12:02 |
dirbaio | nobody has tried it yet | 20:12:58 |
fuse117 | sorry, the nrf9151. | 20:13:09 |
dirbaio | it could work, the C lib is the same for all modems | 20:13:29 |
fuse117 | i tried it as is with the nrf9151, but it doesn' work. i haven't dug into it yet. | 20:14:11 |
fuse117 | the 9160 uses mfw version 1.* while the 9151 (and the 9161 and 9131) use mfw version 2.*. i am not familiar with these codebases, but i would assume some breaking changes across major versions | 20:16:43 |
diondokter | It should work, especially for the nrf9161 (though I've never tried it).
The 9151 has more changes. It should still work, but it would not surprise me if it needs a newer nrfxlib. The Rust wrapper is now at 2.4.1 | 20:17:41 |
diondokter | @fuse117
| 20:18:01 |
18 Jun 2024 |
wassasin | Debugging it was crazy because the celltower would lock us out regularly, and any change in the code could lead to the problem not triggering because it would change the way and timing the task pool is used | 08:53:15 |
wassasin | And that the error occurs in the closed source library does not help either :') | 08:56:06 |
fuse117 | where does nordic archive their secure partition manager hex files? | 21:21:05 |