all patchbots fail on integer.pyx

[Frédéric Chapoton]

Hello,

it seems that all patchbots currently break on a doctest in integral.pyx, see below. The cause is not identified so far. Could be random-seed usage, changes in gmp, or anything else. Any idea ?

Frédéric

File "src/sage/rings/integer.pyx", line 6434, in sage.rings.integer.Integer._shift_helper

Failed example:

    try:

        print('Possible error output from gmp')

        1 << (2^60)

    except (MemoryError, OverflowError, RuntimeError):

        pass

    else:

        print("Failed to raise exception")

Expected:

    Possible error output from gmp...

Got:

    gmp: overflow in mpz type

    Possible error output from gmp

[Lorenz Panny]

This looks like a buffering issue: The test was apparently written
under the expectation that the Python print() would get flushed to
stdout before the GMP error gets printed in the library call, but
this is untrue if Python caches the output until after calling GMP.
Hence we're seeing the output lines in the wrong order.

I suspect passing flush=True to the print() call should fix it.

[Matthias Koeppe]

I have opened https://trac.sagemath.org/ticket/32737 for this