Fix multi_line_output=10 emitting unparsable code around trailing comments

[dylanpulver]

Fixes #2650.

multi_line_output=10 (HANGING_INDENT_WITH_PARENTHESES) can rewrite a valid file into one that no longer parses when an import carries a trailing comment:

$ isort --multi-line 10 --line-length 40 --force-grid-wrap 2 -
from a import (b, c, d  # trailing)

) sits inside the comment.

$ isort --multi-line 10 -
from mypkg.submodule import (alpha, beta, delta, epsilon, eta  # noqa: F401
    gamma, zeta)

the comma after eta is gone. Both raise SyntaxError.

Cause

The sibling grouped modes hoist the comment onto the ( line, so their last line never ends in a comment. This mode keeps it on the import line and then appends punctuation unconditionally: the final return puts ) after the comment, and at a wrap point statement + "," puts the comma after the comment, where comments.add_to_line strips it off with the old comment before re-adding.

Fix

_add_syntax() places ,/) before a trailing comment on the last line, used at both sites.

Verification

• 7 sources x 12 modes x 3 line lengths x 6 option sets (1512 runs): 17 unparsable outputs before, all mode 10; 0 after.
• Suite: 625 passed 1 skipped before, 626 after (the new test); nothing else changed.
• isort --profile hug --check-only, ruff format --check, ruff check, mypy clean.
• Reverting wrap_modes.py fails the new test on '(' was never closed; a partial fix that only relocates ) still fails it on the lost comma.

[Manny7717]

Verified locally against head f09b1ce (base = origin/main @ 131f4ad). Bug is real and the fix is correct.

Bug confirmed: mode 10 (HANGING_INDENT_WITH_PARENTHESES) appended punctuation after a trailing comment. CLI repro on base: from a import b, c, d # trailing\nfrom a import e\n with --multi-line 10 --line-length 40 --force-grid-wrap 2 emits from a import (b, c, d, e # trailing) — the ) lands inside the comment (SyntaxError). Same input on head emits valid from a import (b, c, d, e) # trailing.

Regression coverage proven: transplanted the new test into the base worktree — it fails there with the exact reported SyntaxError: '(' was never closed, and passes on head.

No collateral: full unit suite — head 624 passed / 2 failed / 1 skipped; base+new-test 623 passed / 3 failed. The 2 failures on head (test_settings_path_skip_issue_909, test_skip_paths_issue_938, FileNotFoundError) reproduce identically on base (pre-existing env noise, unrelated). ruff check and ruff format --check clean on both changed files.

Fix review: _add_syntax() correctly handles both append sites (the wrap-point comma and the final closing paren), preserves original spacing before # (code[len(code.rstrip()):] or " "), and is a no-op passthrough when the last line carries no comment. partition("#") is safe here because import statements in this mode cannot contain # before the comment start. The in-test matrix (all WrapModes × 3 line lengths × 3 sources × trailing-comma on/off, every output ast.parsed) is a good guard against this class of bug across sibling modes.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.37%. Comparing base (131f4ad) to head (f09b1ce).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2651   +/-   ##
=======================================
  Coverage   99.37%   99.37%           
=======================================
  Files          41       41           
  Lines        3181     3190    +9     
  Branches      686      688    +2     
=======================================
+ Hits         3161     3170    +9     
  Misses         12       12           
  Partials        8        8           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.