Changes in support of libhdf5 1.14.4

[ajelenak]

The current changes make all test pass with the libhdf5 release branch.

[ajelenak]

I assume the float16 datatype should be tested somewhere?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.58%. Comparing base (060a0d0) to head (ded3c29).
Report is 13 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2406      +/-   ##
==========================================
+ Coverage   89.53%   89.58%   +0.04%     
==========================================
  Files          17       17              
  Lines        2380     2391      +11     
==========================================
+ Hits         2131     2142      +11     
  Misses        249      249              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[aragilar]

The Mac failure was a codecov issue, I've told it to rerun.

@ajelenak We should test it, I'm not sure if there are any current tests though.

[takluyver]

Thanks, this all looks OK to me. There are at least some tests with a float16 data type, e.g.

h5py/h5py/tests/test_dataset.py

Lines 1502 to 1513 in 0f071a4

	@ut.skipUnless(hasattr(np, 'float16'), "NumPy float16 support required")
	def test_mini(self):
	""" Mini-floats round trip """
	self._exectest(np.dtype('float16'))
	# TODO: move these tests to test_h5t
	def test_mini_mapping(self):
	""" Test mapping for float16 """
	if hasattr(np, 'float16'):
	self.assertEqual(h5t.IEEE_F16LE.dtype, np.dtype('<f2'))
	else:
	self.assertEqual(h5t.IEEE_F16LE.dtype, np.dtype('<f4'))

Is 1.14.4 close enough that we can assume this will still work on the release version, or is there scope for last minute surprises?

[ajelenak]

1.14.4 should be released in two days, on 11 April. I just tested this branch against the latest release code and all still works.

[takluyver]

Great, thank-you