# fmt skip internal error fix (space between # and f)

[andrerod22]

Description

Bug report: #2680
Test that replicates bug (see fmtskip7.py in tests/data):

a, b = 1, 2
c =   6 #fmt: skip
d = 5

Resolve: It's seen that of all the formatting options, the inline option, or # fmt skip causes an internal error for black, when the comment should be reformatted like # fmt off. A proposal for this issue was made in the fmtskip_nospace branch which aims to hard code logic handling this bug within the make_comment function of comments.py.

Note: This does NOT handle arbitrary spacing for the # fmt: skip which should be handled in another pull request.

Issues with current pull request: Building on different platforms fails due to the unit test fmtskip7.py failing when running pre-config. An investigation is further needed to resolve this issue.

Checklist

• Add a CHANGELOG entry if necessary?
• Add / update tests if necessary? See fmtskip7.py in tests/data
• Add new / update outdated documentation?

[felix-hilden]

Thanks for the PR! For housekeeping, the list appears as checkboxes if you leave an empty space for "no" and put in an "x" for "yes".

Some points: did you run tests before submitting? Now your test is failing, at least in the CI. So I'm guessing it does not work locally either. Also for the record, I'd like to take care of the multi-space case while we're at it. But I'll be fine if other maintainers just want to push this in.

[andrerod22]

Looking back at the unit test that failed, the error is within _assert_format_equal
Specifically, the expected output is the following:

And the actual output differs here

Notice how the NAME variable in the expression statement is different. It seems as though, Black is expecting just the variable I had written on the same line and completely ignores the #fmt: skip comment on the same line. This is strange behavior and I will bring it up in the discord later.

[JelleZijlstra]

Any updates here? I agree with @felix-hilden that it'd be good to handle other spacing edge cases at the same time.

[github-actions]

diff-shades reports zero changes comparing this PR (0ed191b) to main (0abe85e).

---

What is this? | Workflow run | diff-shades documentation

[JelleZijlstra]

It appears #2970 already fixed this crash in a more general way, so I'm going to close this PR. Thanks for your contribution!