So the current is this:

Do we need another level to test some more edge cases?

I'd rather put bad.txt under subdir (which I was certain it was 😃) to get that extra level.

I believe that would catch all possible errors.

This time I've really changed the file hierarchy to:

        tmpdir
        ├── good.txt
        ├── bad.txt
        ├── ignoredir
        │   └── subdir
        │       └── bad.txt
        └── bad.js

I think we have enough tests to proceed with the changes without regressions.

The current skip matching isn't that bad. The 3 main issues are:

• handle --skip=ignoredir/ with a trailing /
• skip relative paths when codespell is called on a relative path, for example handle --skip=my/path in addition to --skip=./my/path when codespell is called on the current directory
• skip absolute paths when codespell is called on a relative path, for example handle --skip=/absolute/path when codespell is called on a relative path (probably the least important issue)

Do we want another bad.txt in ignoredir too?

The current skip matching isn't that bad.

Yeah I think it was actually just your tofix labelling. 😄

The 3 main issues are:

• handle --skip=ignoredir/ with a trailing /
• skip relative paths when codespell is called on a relative path, for example handle --skip=my/path in addition to --skip=./my/path when codespell is called on the current directory
• skip absolute paths when codespell is called on a relative path, for example handle --skip=/absolute/path when codespell is called on a relative path (probably the least important issue)

Sounds good.

We could add a bad.txt in the 1st level subdirectory. However, we know when we have succeeded ignoring ignnoredir because ignnoredir/subdir/bad.txt will be ignored too. Furthermore, if ignoring a file works both for bad.txt in the top directory and ignoredir/subdir/bad.txt in the 2nd level subdirectory, I think testing in the 1st level subdirectory too is far-fetched. We could do that of course, but then why not test a 3rd level, and a 4th level, and...

Although we can't currently do ignoredir/subdir test can we? To prove it doesn't hit/miss the one within ignoredir?

We probably also want another file in the subdirectories, maybe with another typo, or unique typos per file or something, so we can confirm we aren't just ignoring everything in ignoredir/subdir when we are just trying to block bad.txt within there.

Actually the clever thing might be to have a different power of two number of typos in each different test file, then we know we've got ignored exactly the files we planned to!

Sort of the same with this, if it's actually gone wrong and just matched *, we'd be none the wiser.

Perhaps bad and badder 😆 .