`MemoryError` when calling `report_prefix`, `reset_last_report` and `reset_start_time`

[Joseph-Edwards]

The runner has the following member functions (among others): report_prefix, reset_last_report and reset_start_time. When calling any of these functions on an instance of a derived class (I've checked with ToddCoxeter and KnuthBendix), you are met with:

File ~/miniforge3/envs/libsemigroups_pybind11_live/lib/python3.14/site-packages/libsemigroups_pybind11/detail/cxx_wrapper.py:169, in wrap_cxx_mem_fn.<locals>.cxx_mem_fn_wrapper(self, *args)
    167 if len(args) == 1 and isinstance(args[0], list):
    168     args = [[to_cxx(x) for x in args[0]]]
--> 169 result = getattr(to_cxx(self), cxx_mem_fn.__name__)(*(to_cxx(x) for x in args))
    170 if result is to_cxx(self):
    171     return self

MemoryError: std::bad_alloc

Additionally, when trying to set the report prefix with thing.report_prefix("foo"), Python segfaults. The same is not true when these functions are called on Runner instances, so I assume this has something to do with a bad conversion.

[james-d-mitchell]

I can't reproduce this in the main branch:

In [1]: from libsemigroups_pybind11 import *

In [2]: p = Presentation("abAB")

In [3]: tc = ToddCoxeter(congruence_kind.onesided, p)

In [4]: tc.report_prefix("ABC")
Out[4]: <_libsemigroups_pybind11.Reporter at 0x105512070>

In [5]: tc.report_prefix()
Out[5]: 'ABC'

In [6]: tc.reset_last_report()
Out[6]: <_libsemigroups_pybind11.Reporter at 0x1055ccf30>

In [7]: tc.reset_start_time()
Out[7]: <_libsemigroups_pybind11.Reporter at 0x105562030>

Although I notice that the returned objects all have different addresses, which seems suspicious.

Although IIRC we had deprecated these already anyway. Can we make the deprecation notice like the recent one for parse_relations? And can you confirm that the issue still occurs after a clean install?

[Joseph-Edwards]

Interesting. I get the failures both on main and the released version through mamba.

Currently last_report is deprecated, but the others are not. Should we deprecate them too?

[james-d-mitchell]

Yup let's deprecate, there's no reason to expose these functions.

[james-d-mitchell]

And I suspect that the issue is related to the fact that we are trying to return by reference but failing to:

In [7]: tc.reset_last_report() is tc.report_prefix("ABC")
Out[7]: False

Also just to mention I cannot reproduce this in v1.4.4 as distributed by mamba either.