.gitignore is ignored by black if more than one source location is specified

[szandara]

Describe the bug

When adding more than one source entry, black ignores the .gitignore entries.

To Reproduce

content of .gitignore

generated/*

Running

black test src

produces

$ black --check tests src
would reformat generated/jtd/__init__.py

Oh no! 💥 💔 💥
1 file would be reformatted, 29 files would be left unchanged.

Running

black src

produces

$ black --check src
All done! ✨ 🍰 ✨

Expected behavior

$ black --check test src
All done! ✨ 🍰 ✨

Environment

• Black's version: 22.8.0
• OS and Python version: MacOS m1

Additional context

works with Black version: 22.6.0

[aaossa]

The bug seems to be caused by this section of get_sources:

black/src/black/__init__.py

Lines 660 to 670 in b60b85b

	elif p.is_dir():
	if exclude is None:
	exclude = re_compile_maybe_verbose(DEFAULT_EXCLUDES)
	gitignore = get_gitignore(root)
	p_gitignore = get_gitignore(p)
	# No need to use p's gitignore if it is identical to root's gitignore
	# (i.e. root and p point to the same directory).
	if gitignore != p_gitignore:
	gitignore += p_gitignore
	else:
	gitignore = None

In the first run (the first SRC) exclude is None, but then the value gets overridden by re_compile_maybe_verbose(DEFAULT_EXCLUDES), so in the second iteration the condition is not met and skips to gitignore = None.

This can be fixed by creating a flag before the for loop, something like is_default_exclude = exclude is None, and check if the value is false afterwards:

if is_default_exclude:
    exclude = re_compile_maybe_verbose(DEFAULT_EXCLUDES)
    ...

[aaossa]

Should I submit a PR? @ichard26

[felix-hilden]

@aaossa Would be greatly appreciated!