One size fits all - madness that way lies.

Loved this article, been trying to get QEMU working with Espressif chips (on and off) for a while.

In my experience with LLMs and coding we have a 50/25/25 Rule:

50% of the time - Great code that works
25% of the time - OK, I'll just add a few corrections / tweaks
25% of the time - The code is rubbish

Still, I can read and correct faster than I can type 200 lines of new code.

Read that a while ago, interesting, and sometimes disturbing, book.

Interesting that you have to use addr2line. The last time I was using heap tracing I used the command "idf.py monitor" in a terminal to run my application. The script then decoded all of the tracebacks from the standalone heap trace output automatically.

The heap tracking tools in IDF are fairly good. The standalone mode works well for me but is limited to the amount of memory on the board.

Logging to host has no memory constraints but I found it did not reconcile the calls to free with those to malloc.