Faulty Path/String assumptions

[Ofenhed]

I've noticed multiple bugs regarding data types in utils.rs. Since they are of the same type, I made it a single ticket.

First bug appends two paths with the format! macro instead of using Path::join.

Second bug is similar with a function which seems to try to perform Path::strip_prefix("/").

Third bug converts the executed program and its arguments to String, and gives the Path a different behavior than the arguments (where the Path drops invalid characters and the arguments are simply dropped if they contain invalid characters).

I could also find the format! issue from the first bug in rootfs.rs.

All of the examples above operate on Paths with String as an intermediate data type. String is always UTF-8 compatible, but Path is not. This means that assert_eq!(Path::new(format!("{}{}", a.display(), b.display())), a.join(b)); will fail for some paths. I have not had time to find where these functions are used, so I don't know how big of an issue this is, but it should be fixed before it becomes a serious issue. It should also be verified that CStrings/OsStrings never use Strings as an intermediate data type.

[Furisto]

Thanks for the report. Would you be interested in fixing it?

[Ofenhed]

In the near future I'm afraid I wouldn't have time for that.

[yihuaf]

I can help fix this. I've noticed this before and have a feeling we have to revisit some of this code.