Fatal RPython error: AssertionError in _opimpl_getfield_gc_any_pureornot

[tgbugs]

I encountered the following error on Python 3.10.14 (75b3de9, Apr 28 2024, 20:54:05) [PyPy 7.3.16 with GCC 12.3.1 20230825]. The particular binary was built from the gentoo pypy3 ebuild.

RPython traceback:
  File "rpython_jit_metainterp_4.c", line 22801, in MIFrame_run_one_step
  File "rpython_jit_metainterp_5.c", line 20200, in handler_getfield_gc_i_1
  File "rpython_jit_metainterp_6.c", line 10518, in _opimpl_getfield_gc_any_pureornot___156_i
Fatal RPython error: AssertionError

Presumably the traceback refers to these lines, though of course the underlying cause of the assertion error must be coming from somewhere else.

pypy/rpython/jit/metainterp/pyjitpl.py

Line 1893 in 75b3de9

staticdata.opcode_implementations[op](self, pc)

pypy/rpython/jit/metainterp/pyjitpl.py

Line 857 in 75b3de9

rop.GETFIELD_GC_I, box, fielddescr, 'i')

pypy/rpython/jit/metainterp/pyjitpl.py

Line 932 in 75b3de9

assert resvalue == upd.currfieldbox.getint()

I have only seen this happen once, but I assume that it isn't supposed to happen at all, so I'm creating this issue for reference so that if I encounter it again I can update the information here.

Looking at my logs this appears to happen before
https://github.com/SciCrunch/sparc-curation/blob/1a3e15f974a40507e921c42c920fc005b937381e/sparcur/objects.py#L2017 because a log message generated by the call to Async is missing and would be the next expected message to appear, and it happens after https://github.com/SciCrunch/sparc-curation/blob/1a3e15f974a40507e921c42c920fc005b937381e/sparcur/export/core.py#L368 because the path-metadata.json file was written. The missing log message happens here https://github.com/tgbugs/pyontutils/blob/3fe47bb0609920f76359470a510be63d79843a0a/pyontutils/asyncd.py#L90. I'm going to add additional debug logging to see if I can narrow down the exact place where the issue happens (if I can get it to happen again).

[cfbolz]

Thanks for the report! Yeah, we would need to find a way to reproduce the problem somewhat more reliably in order to progress with this.