assertion diff: trailing character skip fails when first characters differ (off-by-one in `left[-i]` when `i=0`)

[EternalRights]

Was looking at assertion diff output and noticed the trailing skip optimization wasn't kicking in when the compared strings differed at the start. Traced it to _diff_text in compare_text.py line 62:

if len(left) == len(right):
    for i in range(len(left)):
        if left[-i] != right[-i]:
            break
    if i > 42:
        i -= 10  # Provide some context
        yield (
            f"Skipping {i} identical trailing "
            "characters in diff, use -v to show"
        )
        left = left[:-i]
        right = right[:-i]

-0 == 0 in Python, so left[-0] is actually left[0] — the first character, not the last. When the first chars differ, the loop breaks at i=0, i > 42 is False, and the trailing identical characters never get skipped. Dozens of them, just sitting there in the diff output.

Reproducer:

from _pytest.assertion.compare_text import _diff_text
from _pytest.assertion.highlight import dummy_highlighter

# Case 1: first char same — skip works
result = list(_diff_text("a" + "x" + "z" * 50, "a" + "y" + "z" * 50, dummy_highlighter, 0))
# → contains "Skipping 41 identical trailing characters in diff, use -v to show"

# Case 2: first char differs — skip silently fails
result = list(_diff_text("x" + "z" * 50, "y" + "z" * 50, dummy_highlighter, 0))
# → no "Skipping" message, full 50-char diff output

Fix: loop from i=1 instead of i=0:

for i in range(1, len(left) + 1):
    if left[-i] != right[-i]:
        break

The existing test test_text_skipping_trailing doesn't catch this because both test strings start with 'a', so the accidental left[-0] comparison happens to pass and the loop continues to the actual trailing characters.

This code has been around since 2013 (originally in assertion/util.py, moved to compare_text.py in #14520).

[Aditya-Tripuraneni]

Hello, I’d like to work on this issue if it’s still available. The proposed fix looks straightforward, and I’d like to check for any edge cases and verify the regression coverage before sending a PR.