Strange slowdown on `str()` after handled FLINT exception

[user202729]

sage: R.<x> = Zmod(4)[]
sage: f = x^2 + 3 * x + 1
sage: g = x^2 + x + 1
sage: %time str(f)
CPU times: user 2.34 ms, sys: 0 ns, total: 2.34 ms
Wall time: 2.37 ms
'x^2 + 3*x + 1'
sage: f.gcd(g)
Flint exception (Impossible inverse):
    Cannot invert modulo 2*2
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
File nmod_poly_linkage.pxi:599, in sage.rings.polynomial.polynomial_zmod_flint.celement_gcd()

RuntimeError: Aborted

During handling of the above exception, another exception occurred:

RuntimeError                              Traceback (most recent call last)
Cell In[5], line 1
----> 1 f.gcd(g)

File element.pyx:4743, in sage.structure.element.coerce_binop.new_method()

File polynomial_template.pxi:390, in sage.rings.polynomial.polynomial_zmod_flint.Polynomial_template.gcd()

File nmod_poly_linkage.pxi:605, in sage.rings.polynomial.polynomial_zmod_flint.celement_gcd()

RuntimeError: FLINT gcd calculation failed
sage: %time str(f)
CPU times: user 1.33 s, sys: 9.42 ms, total: 1.33 s
Wall time: 1.34 s
'x^2 + 3*x + 1'

Apparently it becomes fast again after accessing a nonexistent attribute

sage: %time f[0]
CPU times: user 304 ms, sys: 75 μs, total: 304 ms
Wall time: 305 ms
1
sage: %time f.xxx
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
File <timed eval>:1

File element.pyx:495, in sage.structure.element.Element.__getattr__()

File element.pyx:508, in sage.structure.element.Element.getattr_from_category()

File getattr.pyx:363, in sage.cpython.getattr.getattr_from_other_class()

AttributeError: 'sage.rings.polynomial.polynomial_zmod_flint.Polynomial_zmod_flint' object has no at
tribute 'xxx'
sage: %time f[0]
CPU times: user 78.8 ms, sys: 0 ns, total: 78.8 ms
Wall time: 79.2 ms
1

Environment

• OS: Linux
• Sage Version: latest

Checklist

• I have searched the existing issues for a bug report that matches the one I want to file, without success.
• I have read the documentation and troubleshoot guide

[user202729]

By benchmarking result it looks like the issue lies in verify_exc_value which repeatedly calls garbage collection.

With some testing

%%cython
from cysignals.signals cimport cysigs
def get_cysigs(): return <long> cysigs.exc_value
def get_cysigs_refcount(): return cysigs.exc_value.ob_refcnt

the value remains nonzero. Also get_cysigs_refcount() returns 2.

Reading the docstring:

The implementation is based on reference counting: it checks whether the exception has been deleted. This means that it will break if the exception is stored somewhere.

Move to sagemath/cysignals#215 .