Multi-platform IoT firmware supporting up to 32 platforms - OBK 2025 summary

>>21877411
I originally meant about if ret == 0, but i've double-checked and saw that it's freed in client thread.

Are os_malloc and malloc different on any platform, and if so, what's the difference?

>>21877422
In some it may be separate, malloc is default newlib allocator, os_malloc - freertos one.

I finally managed to implement some feature to allow logging to LFS: PR #2076.

Wasn't a easy as I thought on the first look, because writing to lfs is slooow and would crash with an IRQ watchdog if writing more than a few lines.

So now it will use an own thread for flushing the message buffer to lfs.

It allows for "startup logging" with command


Will logg all further starts for the givven number of seconds to lfs (as long as there's free space in lfs).
So if done, disable with


During operation you can use

to write messages to LFS for this number of seconds.
Might be stopped earlier by calling





Feel free to test or propose changes - I only checked with W800 and Linux simulator for now.

@max4elektroda

I remember there was a concern about flash wear so I would suggest you try this on modules with external flash chips just in case.

I would see this as a debugging feature only, not as a regular one.
Idea came back after issue 2064 where the user couldn't see the logs easily and he assumed a startup error when in WiFi client mode. With this feature he could have setup logging to lfs, insert his WiFi credentials and after a start in regular mode switching to safe mode again see the logs.
For all other cases we have TCP logging which can be easily saved, so we could agree it only makes sense for startup, not during operation.

p.kaczmarek2 wrote:

@insmod can you review?
https://github.com/openshwprojects/OpenBK7231T_App/pull/2051/changes
This seems a bit bloated for me, but maybe I'm wrong. It's a #define, after all. So transparent to compiler after initial pass.

I think the actual runtime changes are pretty narrow. Maybe what makes the PR look larger is the supporting test scaffolding in the simulator, which I added first so the allocation-failure paths could be reproduced and validated deterministically. Or do you mean a different kind of bloat?

I've resolved merge conflicts. No worries if you don't think these are good changes and PR should be closed

@max4elektroda make log2lfs byte and disable it by default and I think it's ready to be merged

Added after 10 [hours] 26 [minutes]:

@divadiow are you using windows?

p.kaczmarek2 wrote:

@divadiow are you using windows?

Yes.

Have Ubuntu on a laptop or two

I am experimenting with some simplest windows wrapper for build scripts:
https://github.com/openshwprojects/OpenBK7231T_App/pull/2086
So basically single .bat can build binary. Currently for ESP32 and BK7231, not tested.

I see. OK. but wasn't @DeDaMrAz's Docker creation for the same? maybe this is even easier though?

Well.... I haven't managed to run this Docker thing on my Windows 10 yet. It seems to be making lots of issues for me, both older and newer versions, due to the different reasons.

OK.

Cloned main app and OpenBK7231T submodule only.





didn't troubleshoot further

Looks like Berry prebuild is missing. You have a good point. I didn't realize that.

How is it now, @divadiow ? I also added ESP8266.

started again. same clone and submodule as before.



BK7231N

git submodule update --init sdk/OpenESP8266

am I missing a step?

p.kaczmarek2 wrote:

@max4elektroda make log2lfs byte and disable it by default and I think it's ready to be merged

Actual try at
https://github.com/MaxineMuster/OpenBK7231T_App/actions/runs/24894132784

Maybe @divadiow could give these artifacts a try on RTL platforms?
I added some comments to mainConfig_t adresses.
For both ESPIDF and RTL there was one "implicit" padding for alignment to 4 byte types - I added a short for both.
So just to make sure, a configured device won't lose it's config I'd like some tests.
I already tried on ESP32 and C3, so ESPIDF should be save, but only have one RTL device...

You can use WIFI password2 (for SSID2) which is "beyond" this place for both platforms.
Hint: don't use SSID2 itself since you can't delete it w/o resetting to OpenAP and then device will be in "second SSID-mode" until that - another thing to fix in the future....

Testing logging to lfs (if you want to give it a try) with this command:



so log next startup to LFS for the default 10 seconds:



or log next 3 startups for 15 seconds :



Thanks!

max4elektroda wrote:

Maybe @divadiow could give these artifacts a try on RTL platforms?

I hope to try this out tomorrow

cool. so I did

logStartup2LFS 15 3

and this was the content of the file saved (had to create autoexec.bat first for LFS to be ready)

logStart2LFS_f9d408fb2054 - RTL87X0C


vs uart log out


all 3 wrote a log. it did stop after 3.

the single was fine too. stopwatch shows time logging is accurate within ~1s (human inaccuracy stopping/starting)

Thanks @divadiow!
Good catch. I would expect, if called on a device without lfs ready, it will init lfs on the first call. So we should point out that lfs should be initialized before use.
Did you find any problems with keeping the existing configuration from e.g. release or any other version to this one?

I just configured a load of things with general release- some flags, 6s boot OK, startup command, pins, mqtt and all still OK after OTA to PR