Test failure for arb_mat_spd_lll_reduce

[fredrik-johansson]

@j-kieffer

With large enough test multiplier, the test for arb_mat_spd_lll_reduce fails (well, it prints out failures although the program passes due to a missing flint_abort).

> export FLINT_TEST_MULTIPLIER=100
> build/arb_mat/test/main arb_mat_spd_lll_reduce
arb_mat_spd_lll_reduce...
FAIL (identity)
[3.883670689 +/- 0, 1.970703095 +/- 5.5288e-191]
[1.970703095 +/- 5.5288e-191, 4.968811095 +/- 7.2768e-190]
[[1 0]
[-1 1]
]
FAIL (identity)
[1.000000000 +/- 1.5918e-95, -0.01562500000 +/- 2.5603e-97]
[-0.01562500000 +/- 2.5603e-97, 0.9924469068 +/- 4.4828e-95]
[[0 1]
[1 0]
]
FAIL (identity)
[0.9776651859 +/- 1.3362e-79, 0.4943847656 +/- 3.0063e-79]
[0.4943847656 +/- 3.0063e-79, 1.242202812 +/- 8.0753e-79]
[[1 0]
[-1 1]
]
FAIL (identity)
[0.9689941406 +/- 2.1880e-162, -0.4902724027 +/- 4.9296e-162]
[-0.4902724027 +/- 4.9296e-162, 4.247570008 +/- 4.4282e-162]
[[1 0]
[1 1]
]
FAIL (identity)
[3.875976592 +/- 1.1747e-153, 1.968749773 +/- 3.0751e-154]
[1.968749773 +/- 3.0751e-154, 4.999999762 +/- 6.8990e-154]
[[1 0]
[-1 1]
]
FAIL (identity)
[0.9699406550 +/- 0, -0.4906244986 +/- 0]
[-0.4906244986 +/- 0, 1.248172295 +/- 3.2946e-83]
[[1 0]
[1 1]
]
FAIL (identity)
[3.875976680 +/- 2.6970e-194, 1.968750030 +/- 6.8494e-195]
[1.968750030 +/- 6.8494e-195, 5.000000000 +/- 2.1918e-193]
[[1 0]
[-1 1]
]
FAIL (identity)
[0.9689946100 +/- 4.2055e-79, 0.4921876192 +/- 6.7978e-79]
[0.4921876192 +/- 6.7978e-79, 12.50000334 +/- 6.0761e-77]
[[1 0]
[-1 1]
]
FAIL (identity)
[0.9689941480 +/- 1.8614e-124, -0.3829687899 +/- 4.0906e-124, -0.4921875019 +/- 4.1955e-124]
[-0.3829687899 +/- 4.0906e-124, 1.151358080 +/- 6.3452e-123, 0.3195238290 +/- 1.8490e-123]
[-0.4921875019 +/- 4.1955e-124, 0.3195238290 +/- 1.8490e-123, 4.265624999 +/- 2.6095e-122]
[[1 0 0]
[0 1 0]
[1 0 1]
]
arb_mat_spd_lll_reduce                            0.91   (PASS)

[j-kieffer]

Thanks for the bug report and sorry for the delay! Should be fixed by #2404