array tests: handle different hexdigests from zlib-ng (#1678)

[AdamWill]

As explained in the issue, zlib-ng produces different hex digests from original zlib. This adjusts the tests slightly to allow for this.

TODO:

• GitHub Actions have all passed
• Test coverage is 100% (Codecov passes)

Removed TODO items are irrelevant as this only changes tests. This is more or less the same as #1971 , but for the main (v2) branch rather than v3 branch.

[AdamWill]

If I run black on the changed file here locally it does indeed want to change a bunch of stuff, but none of it is stuff this PR touches - the issues already exist.

[d-v-b]

Thanks for the fix. For posterity, the ideal way to handle two different versions of zlib would be to condition the test case on the detected zlib version, but our test design makes that very tedious and not worth the effort. I think what you have done here is good!

[d-v-b]

we are seeing test failures due to numpy 2.0 I think

[AdamWill]

yeah, it looks like numpy got some custom types. I guess you might need to do stuff like:

diff --git a/src/zarr/v2/core.py b/src/zarr/v2/core.py
index c1223dac..04da6749 100644
--- a/src/zarr/v2/core.py
+++ b/src/zarr/v2/core.py
@@ -759,7 +759,7 @@ class Array:
 
         Retrieve a single item::
 
-            >>> z.get_basic_selection(5)
+            >>> int(z.get_basic_selection(5))
             5
 
         Retrieve a region via slicing::

or something along those lines.

[QuLogic]

I guess the NumPy stuff was fixed by #2073, so this might just need a rebase.

[AdamWill]

Rebased.