Misc fixes

[skirpichev]

2e96fb6 Avoid computing _PyHASH_MODULUS in GMPy_MPQ_Hash_Slot()
324e284 Fix compiler warning
8a7b1ff Fix rst markup in docstrings (for emphasis)
12afc77 Use gmpy2.__package__ to get project name in sphinx
b470a3b Set doctest_global_cleanup sphinx directive

[codecov-commenter]

Codecov Report

Merging #452 (874c95b) into master (21ccbf7) will not change coverage.
The diff coverage is n/a.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

@@           Coverage Diff           @@
##           master     #452   +/-   ##
=======================================
  Coverage   84.97%   84.97%           
=======================================
  Files          49       49           
  Lines       11738    11738           
  Branches     2206     2206           
=======================================
  Hits         9974     9974           
  Misses       1764     1764           

Files	Coverage Δ
src/gmpy2_mpz_misc.c	100.00% <ø> (ø)

📣 Codecov offers a browser extension for seamless coverage viewing on GitHub. Try it in Chrome or Firefox today!

[casevh]

I need to test the _PyHASH_MODULUS change on 64-bit Windows. mpz_set_si() accepts a long (32-bit) but _PyHASH_MODULUS is too large. At least that's my first reaction. I will respond later with the test results.

[skirpichev]

On Sat, Nov 04, 2023 at 02:13:46PM -0700, casevh wrote: I need to test this on 64-bit Windows. mpz_set_si() accepts a long (32-bit) but _PyHASH_MODULUS is too large. At least that's my first reaction. I will respond later with the test results.

Yep, we should use mpz_set_ui().

[casevh]

It's worse than that. 64-bit Windows defines long as a 32-bit integer and long long as a 64-bit integer. Most other 64-bit platforms define both long and long long as 64-bit integers. GMP expects si and ui to be long and they don't offer functions that accept long long.

Here is the compile error message with the last change:
\gmpy\src\gmpy2_hash.c(67): warning C4305: 'function': truncation from 'unsigned __int64' to 'unsigned long'

One of the quirks that makes supporting Windows difficult.

Let's just skip the _PyHASH change for now. I found a few other conversion warnings that I need to resolve first.

[skirpichev]

Ok, first comment was removed. It seems we really want _PyHASH_BITS macro.

[skirpichev]

Change in _mpfr_hash() was reverted too, @casevh

[casevh]

Thanks for your patience in dealing with these subtle issues.

[skirpichev]

Hash-related patch was moved to #453