ESP32-C3 Serial instability

[bill88t]

This has been a long standing issue that just nobody has bothered with.
It applies to all CP versions.

Under several circuimstances, CircuitPython on ESP32-C3 decides to lock up stdin or go out of sync.

Some ways to reproduce it:

• Press any key before before circuitpython has finished init.
• Press too many keys at once.
• Work with serial for more than 3 minutes.
• Use ljinux. It very heavily relies on ANSI escapes for graphics.

Ctrl+C usually makes it through regardless, but doesn't clear stdin (this could be used as a workaround).

Too many Ctrl+C at once send garbage:

[board@ljinux| ~]> [[[[[5;20R[[^C
> ;ins5;ins5ins5ins5;20Rins5ins5ins5ins5ins5ins5ins5ins5ins5ins5ins5ins5ins5;20R

In some rare cases stdin goes out of sync instead of locking up.
What this means is that I type abcdef, abc is echo'ed back and then if I press ghi, def is echo'ed.

[RetiredWizard]

@bill88t Was this resolved by #7973?

[bill88t]

There is still a oddly specific Cx desync issue which I have a workaround in place for.
With that one workaround it's perfectly stable.
Without it, my terminal line is not printing no matter how hard I .flush().

[gitcnd]

no matter how hard I .flush().

what exactly, are you flushing?

sys.stdout.flush() # AttributeError: 'FileIO' object has no attribute 'flush'

Same problem has been driving me crazy on both my ESP32-C3 (Qt Py) and my ESP32-S2

[bill88t]

what exactly, are you flushing?

Oh I am using the jcurses library and virtual consoles on Cx and regular esp32 for interoperability between boards.
Kinda forgot about that. Yea flushing has no effect there.
Oh and btw the workaround to the bug is sending ansi escapes. b"\x1b[s\x1b[1B\x1b[1C\x1b[u" works really well.