mpfr: subnormalize=True breaks mpfr('inf') and mpfr('nan') about gmpy HOT 8 CLOSED

Today in new ways to fail writing a bug report: not actually listing the failure.

The failing test cases raise ValueError: invalid digits

from gmpy.

I believe I've fixed this in the upcoming 2.1.3 release. Can you test the appropriate binary wheel contained in https://github.com/aleaxit/gmpy/suites/8231102535/artifacts/359235281 ?

from gmpy.

Ill be honest I have no Idea how to install a wheel directly. My bugs were fixed by installing gmpy2 master. I couldn't find a 2.1.3 tag or branch to install.

I do have a new bug though that is probably related.

Passing tests:

with gmpy2.local_context(subnormalize=False): 
    assert gmpy2.is_signed(gmpy2.mpfr('-0'))

with gmpy2.local_context(subnormalize=False):
    assert gmpy2.is_signed(gmpy2.mpfr(float('-0')))

with gmpy2.local_context(subnormalize=True): 
    assert gmpy2.is_signed(gmpy2.mpfr(float('-0')))

Failing test:

with gmpy2.local_context(subnormalize=True): 
    assert gmpy2.is_signed(gmpy2.mpfr('-0'))

If it is helpful in the passing test repr(negative_zero) == "mpfr('-0.0')".
Where in the failing test repr(negative_zero) == "mpfr('0.0')".

from gmpy.

@casevh Would you expect the above to be solved in the 2.1.3 release

from gmpy.

from gmpy.

It was fixed in 2.1.3. I did an updated 2.1.4 release that solved multiple issues with building wheels. The normal pip install process should work.

from gmpy.

@casevh, probably this can be closed.

from gmpy.

Thanks for finding and reporting the bug.

Case

from gmpy.