swev-id: pytest-dev__pytest-6202 | Preserve '.[' in long report headlines (remove getmodpath sanitization)

[rowan-stein]

Fixes #22

Summary

• Long failure headlines mutated param ids containing the substring ".[" (e.g., "..[" became ".["), breaking readability and some tooling that parses pytest output.

Root cause

• src/_pytest/python.py (PyobjMixin.getmodpath) returned s.replace(".[", "["). This was originally to avoid test_gen.[0] names from yield-based tests (removed in pytest 4.0). It now corrupts legitimate ids.

Change

• Return the assembled modpath string unchanged.
• Add a regression test to ensure headlines preserve param ids containing ..[ sequences.

Reproduction steps

1. Create bug.py:

import pytest

@pytest.mark.parametrize("a", ["..["])
def test_boo(a):
    assert False

2. Run: pytest -vv bug.py
3. Before this fix, failure headline showed test_boo[.[] instead of test_boo[..[].

Observed failure (before fix)

bug.py F                                                                 [100%]

=================================== FAILURES ===================================
_________________________________ test_boo[.[] _________________________________

a = '..['

    @pytest.mark.parametrize("a",["..["])
    def test_boo(a):
>       assert 0
E       assert 0

bug.py:6: AssertionError
============================== 1 failed in 0.06s ===============================

Tests

• New test testing/test_headline_modpath_param_id.py which runs a failing parametrized test with id a..[b] and asserts the -vv output includes test_boo[a..[b]] unchanged.

Notes

• This change does not affect nodeid semantics; it only preserves the headline/domain string. Yield tests are gone, so the previous sanitization is obsolete.
• Per instructions: not triggering CI here.

[noa-lucent]

Looks good—removing the obsolete sanitization restores the expected headline text, and the new regression test captures the failure case.