[Denny-j-mangalathu]

A multiline JavaScript template literal evaluates to the same string whether its source file uses LF, CRLF, or CR. Babel currently preserves physical CRLF and CR in the extracted message ID, so a catalog generated from a Windows checkout can miss the runtime lookup.

Normalize physical line endings in template literals before decoding escapes. Explicit carriage-return escapes still produce carriage returns. Token positions are calculated from the original source.

See the note under ECMAScript template literal semantics.

The CRLF and CR regression cases fail on upstream master. With the fix, 88 extraction, lexer, and extraction frontend tests pass. Frontend tests used locale data from the published Babel package.

Prepared with Codex assistance.

[chrikrah]

@Denny-j-mangalathu this can be merged from my side; CRLF and CR fail on master.

non-blocking: #1385 edits the same unquote_string line; I'll re-test whichever lands second.

# git archive 307a6ac (head), base a18709c, Python 3.13, pytest 9, babel.__file__ from the export
$ python -m pytest -q tests/messages/test_js_extract.py tests/messages/test_jslexer.py
26 passed
# jslexer.py from a18709c, tests kept
$ python -m pytest -q tests/messages/test_js_extract.py
FAILED test_extract_normalizes_template_line_endings[\r\n]
FAILED test_extract_normalizes_template_line_endings[\r]
2 failed, 18 passed
# CLDR imported with scripts/download_import_cldr.py
$ python -m pytest -q tests --ignore=tests/messages/test_setuptools_frontend.py
6836 passed, 1219 skipped, 1 xfailed
# not run: test_setuptools_frontend.py (needs the package installed)
# git 2.43.0, heads 307a6ac (#1450) and 521f0f9 (#1385)
$ git merge-tree --write-tree 307a6ac 521f0f9
CONFLICT (content): Merge conflict in babel/messages/jslexer.py
exit 1