beta.1 crashes after ctrl-d after exception in REPL

[dhalbert]

Same result on CPB, so it's not port-specific:

$ repl
Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.

Press any key to enter the REPL. Use CTRL-D to reload.
Adafruit CircuitPython 5.0.0-beta.1 on 2019-12-10; Adafruit Metro M4 Express with samd51j19
>>> [ctrl-d]
soft reboot

Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.



Press any key to enter the REPL. Use CTRL-D to reload.
Adafruit CircuitPython 5.0.0-beta.1 on 2019-12-10; Adafruit Metro M4 Express with samd51j19
>>> asdf
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
NameError: name 'asdf' is not defined
>>> [ctrl-d]

FATAL: read zero bytes from port
term_exitfunc: reset failed for dev UNKNOWN: Input/output error

[jerryneedell]

I cannot reproduce this
on my CPB

Press any key to enter the REPL. Use CTRL-D to reload.
Adafruit CircuitPython 5.0.0-beta.1 on 2019-12-10; Adafruit Circuit Playground Bluefruit with nRF52840
>>> 
>>> 
>>> adasdsa
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
NameError: name 'adasdsa' is not defined
>>> 
soft reboot

Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.



Press any key to enter the REPL. Use CTRL-D to reload.
Adafruit CircuitPython 5.0.0-beta.1 on 2019-12-10; Adafruit Circuit Playground Bluefruit with nRF52840
>>> 

[jepler]

Locally I built with gcc9 and gcc7. I reproduce this with 9, but not with 7.

@dhalbert has already identified one specific place where code changed and made a problem (volatile is now needed to get the same result); this may be some other thing.

Perhaps we should consider reverting to the older compiler at this time. @tannewt thoughts?

[tannewt]

Has anyone reproduced this under GDB? I wouldn't rollback the compiler unless we can't figure it out.

[jerryneedell]

Locally I built with gcc9 and gcc7. I reproduce this with 9, but not with 7.

@dhalbert has already identified one specific place where code changed and made a problem (volatile is now needed to get the same result); this may be some other thing.

Perhaps we should consider reverting to the older compiler at this time. @tannewt thoughts?

interesting -- That explains why my CPB experience is different. It was build with the old compiler.

[dhalbert]

I am willing to try to figure this out. It probably reflects a real bug. Importantly, we should note these fixes in MicroPython for gcc 9 issues:

micropython#4844
micropython#5008

[tannewt]

Ok, I reproduced it with an optimized build on nRF. The stack trace isn't helpful:

Breakpoint 1, 0x000292fa in reset_into_safe_mode ()
(gdb) bt
#0  0x000292fa in reset_into_safe_mode ()
#1  0x0002a3fc in softdevice_assert_handler ()
#2  0x00024a44 in ?? ()

I'm cherry picking the commits from micropython#5008 now.

[tannewt]

The commit from micropython#4844 fixes this.

[kevinjwalters]

Glad to hear this is fixed, I just noticed it on Adafruit CircuitPython 5.0.0-beta.1 on 2019-12-10; Adafruit Circuit Playground Bluefruit with nRF52840