Proposal: revisit folding skipped tests in summary

[nicoddemus]

Currently we "fold" skipped tests in the summary:

===================================== short test summary info =====================================
SKIPPED [1] .tmp\test_summary_1.py:20: not available on this platform
SKIPPED [1] .tmp\test_summary_1.py:35: not available on this platform
XFAIL .tmp\test_summary_1.py::test_mem_crash
  reason: quarantine due to crashes
XPASS .tmp\test_summary_1.py::test_mem_crash_new
FAILED .tmp\test_summary_1.py::TestSimple::test_name - NameError: name 'name' is not defined
FAILED .tmp\test_summary_1.py::test_boolean_ops - assert False is True
PASSED .tmp\test_summary_1.py::TestSimple::test_passes
PASSED .tmp\test_summary_1.py::TestSimple::test_passes1
PASSED .tmp\test_summary_1.py::test_passes_non_critical
================== 2 failed, 3 passed, 2 skipped, 1 xfailed, 1 xpassed in 0.09s ===================

They are the only entries which don't show the node id, showing the file name and line number instead, and also try to fold/accumulate the summaries which come from the same skip marker/reason.

I tracked it down to prokaktus@3795b08, which introduced the -r option.

I wonder if this inconsistency is still worth having? Personally I would prefer the SKIPPED message to look the same as XFAIL/XPASS, and show up as once for each skipped test, instead of having them folded.

I notice there's a comment in the current code too regarding markers:

pytest/src/_pytest/terminal.py

Lines 1319 to 1321 in 28e8c85

	# Folding reports with global pytestmark variable.
	# This is a workaround, because for now we cannot identify the scope of a skip marker
	# TODO: Revisit after marks scope would be fixed.

Perhaps @RonnyPfannschmidt can remark on that comment.

[RonnyPfannschmidt]

The folding was introduced for sensible reporting even when there are hundreds of skips for reasons

AFAIK those numbers tend to be hit with skips often enough to warrant it, i haven't observed similar usage for xfail /xpass which is used in very different use cases

Im explicitly in favour of the difference here

[eli-schwartz]

Hmm, hundreds of skips due to parameterization I assume?

I think I would personally prefer the test class::function name on its own instead of a line number, as I'd assume that has to be just as foldable as the line number.