On 2020-07-14 09:48:18 +0200, Matthias Klose wrote:
> There is no 2.1.0 beta4, just a beta1, so I don't know what was packaged in
> February 2020. However the tests now fail with mpfr 4.1.0, seems to be
> consistent across all architectures:
>
> **********************************************************************
> File "test/test_gmpy2_format.txt", line 157, in test_gmpy2_format.txt
> Failed example:
> c.__format__('e')
> Differences (ndiff with -expected +actual):
> - '3.3333333333333331e-01+5e+00j'
> + '3.3333333333333331e-01+5.0000000000000000e+00j'
> ? +++++++++++++++++
FYI, the old MPFR behavior was regarded as a bug:
https://gforge.inria.fr/tracker/index.php?func=detail&aid=21816&group_id=136&atid=619
There were 2 reasonable interpretations of the description in the
MPFR manual that did not leave the output partly unspecified, and
for each of them, some outputs were incorrect. The one that has
been chosen is the one that is closer to ISO C's %e and it does
not change the numerical output value (the only difference is
trailing zeros).
--
Vincent Lefèvre <[email protected]> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
