Handle FLINT exceptions gracefully

[rwst]

Sage crashes with

sage: (x^2222222222+1).polynomial(QQ)
...
#2  0x00007f46536d606d in sigdie()
#3  0x00007f46536d575a in sage_signal_handler()
#4  0x00007f46596f03e0 in __restore_rt()
#5  0x00007f46590e70a0 in raise()
#6  0x00007f465865c3a0 in abort()
#7  0x00007f4645da4b40 in flint_memory_error()
#8  0x00007f4645daa400 in flint_calloc()
#9  0x00007f4645e00150 in fmpq_poly_realloc()
#10 0x00007f4645e058a0 in fmpq_poly_fit_length()
#11 0x00007f4645dff850 in fmpq_poly_set_coeff_si()
#12 0x00007f463c1b6f30 in __pyx_pf_4sage_5rings_10polynomial_25polynomial_rational_flint_25Polynomial_rational_flint_58__pow__() at /home/ralf/sage/src/build/cythonized/sage/rings/polynomial/polynomial_rational_flint.cpp:10287
  10282     *             if self._is_gen:
  10283     *                 fmpq_poly_set_coeff_si(res.__poly, n, 1)             # <<<<<<<<<<<<<<
  10284     *             else:
  10285     *                 sig_on()
  10286     */
> 10287          fmpq_poly_set_coeff_si(__pyx_v_res->__pyx___poly, __pyx_v_n, 1);
  10288          goto __pyx_L16;
  10289        }
  10290        /*else*/ {
  10291    
#13 0x00007f465933e270 in ternary_op()
#14 0x00007f4659340520 in PyNumber_Power()
#15 0x00007f46593e62b0 in arithmetic() at /home/ralf/sage/local/lib/python2.7/site-packages/sage/symbolic/expression_conversions.py:1110
  1105            if not any(repr(v) in self.varnames for v in ex.variables()):
  1106                return self.base_ring(ex)
  1107            elif operator == _operator.pow:
  1108                from sage.rings.all import Integer
  1109                base, exp = ex.operands()
> 1110                return self(base)**Integer(exp)
  1111            else:

Depends on #17668

Component: c_lib

Keywords: flint_memory_error

Author: Jeroen Demeyer

Branch/Commit: 4835d4b

Reviewer: Ralf Stephan

Issue created by migration from https://trac.sagemath.org/ticket/17629

[rwst]

Attachment: sage_crash_mG0GBp.log

crash report

[rwst]

comment:3

Has nothing to do with fraction, crashes with only x^2222222222+1 too. If this is memory, can we work around with sparse polynomials, at least give the user the option?

[rwst]

Changed keywords from conversion to conversion, sparse, dense

[jdemeyer]

comment:4

Sparse polynomials should be a different ticket.

[jdemeyer]

Changed keywords from conversion, sparse, dense to flint_memory_error

[rwst]

comment:5

Replying to @jdemeyer:

Sparse polynomials should be a different ticket.

See #17782 (EDIT)

[jdemeyer]

Author: Jeroen Demeyer

[jdemeyer]

Dependencies: #17668

[jdemeyer]

Branch: u/jdemeyer/ticket/17629

[jdemeyer]

Commit: 4835d4b

[jdemeyer]

New commits:

47217bc	Replace PY_NEW and PY_NEW_SAME_TYPE by Cython code
4835d4b	Handle FLINT exceptions gracefully

[rwst]

Reviewer: Ralf Stephan

[rwst]

comment:9

Except for cosmetics this practically replaces sig_on with sig_str(...) in FLINT code and adds an exception for 0^(-1). It also passes make ptestlong. So, depending on the review for #17668 this is fine.

[vbraun]

Changed branch from u/jdemeyer/ticket/17629 to 4835d4b