Re: [PATCH v14 03/16] quote.c: remove path_relative, use relative_path instead

From: Jiang Xin <worldhello.net@gmail.com>
Date: Jun 25, 2013, 11:09 UTC
Message-ID: <CANYiYbGvqHpQpp2UD0nBS5pqeimGE7yDMVYr7q9bGA0hSNNNDw@mail.gmail.com>
In-Reply-To: <7vhagn8gtz.fsf@alter.siamese.dyndns.org>

2013/6/25 Junio C Hamano <gitster@pobox.com>:

Show 8 quoted lines

> Jiang Xin <worldhello.net@gmail.com> writes:
>
>> Since there is an enhanced version of relative_path() in path.c,
>> remove duplicate counterpart path_relative() in quote.c.
>
> There is no nice comparison chart before and after like you had in
> patch 02/16?
>

You mean drawing a table to compare output of path_relative and relative_path?

I will rewrite the commit log for patch 03/16 like the following. Need to polish spellings and grammars.

quote.c: substitute path_relative with relative_path

Substitute the function path_relative in quote.c with the function relative_path. Function relative_path can be treated as an enhanced and robust version of path_relative.

Outputs of path_relative and it's replacement (relative_path) are the same for the following cases:

    path      prefix     output of path_relative  output of relative_path
    ========  =========  =======================  =======================
    /a/b/c/   /a/b/      c/                       c/
    /a/b/c    /a/b/      c                        c
    /a/       /a/b/      ../                      ../
    /         /a/b/      ../../                   ../../
    /a/c      /a/b/      ../c                     ../c
    /x/y      /a/b/      ../../x/y                ../../x/y
    a/b/c/    a/b/       c/                       c/
    a/        a/b/       ../                      ../
    x/y       a/b/      ../../x/y                 ../../x/y
    /a/b      (empty)    /a/b                     /a/b
    /a/b      (null)     /a/b                     /a/b
    a/b       (empty)    a/b                      a/b
    a/b       (null)     a/b                      a/b

But if both of the path and the prefix are the same, or the returned relative path should be the current directory, the outputs of both functions are different. Function relative_path returns "./", while function path_relative returns empty string.

    path      prefix     output of path_relative  output of relative_path
    ========  =========  =======================  =======================
    /a/b/     /a/b/      (empty)                  ./
    a/b/      a/b/       (empty)                  ./
    (empty)   (null)     (empty)                  ./
    (empty)   (empty)    (empty)                  ./

But not panic, the callers of path_relative can handle such cases, or never encounter this issue at all. E.g.

 * In function quote_path_relative, if the output of path_relative is
   empty, append "./" to it, like:

       if (!out->len)
           strbuf_addstr(out, "./");

 * Another caller is write_name_quoted_relative, which is only used by
   builtin/ls-files.c. git-ls-files only show files, so path of files
   will never be identical with the prefix of a directory.

The following differences show that path_relative can not handle extra slashes properly.

    path      prefix     output of path_relative  output of relative_path
    ========  =========  =======================  =======================
    /a//b//c/ //a/b//    ../../../../a//b//c/     c/
    a/b//c    a//b       ../b//c                  c

And if prefix has no trailing slash, path_relative can not work properly either. But since prefix always has a trailing slash, so it's not a problem.

    path      prefix     output of path_relative  output of relative_path
    ========  =========  =======================  =======================
    /a/b/c/   /a/b       b/c/                     c/
    /a/b      /a/b       b                        ./
    /a/b/     /a/b       b/                       ./
    /a        /a/b/      ../../a                  ../
    a/b/c/    a/b        b/c/                     c/
    a/b/      a/b        b/                       ./
    a         a/b        ../a                     ../
    x/y       a/b/       ../x/y                   ../../x/y
    a/c       a/b        c                        ../c
    /a/       /a/b       (empty)                  ../
    (empty)   /a/b       ../../                   ./

Show 14 quoted lines

>>  void write_name_quoted_relative(const char *name, size_t len,
>>                               const char *prefix, size_t prefix_len,
>>                               FILE *fp, int terminator)
>>  {
>>       struct strbuf sb = STRBUF_INIT;
>>
>> -     name = path_relative(name, len, &sb, prefix, prefix_len);
>> +     name = relative_path(name, prefix, &sb);
>
> Are we sure nobody calls prefix_len pointing into the middle of
> string, not at the end of "prefix"?  This is unsafe for such a
> caller, and to make sure we catch them, we should remove the
> now-unused prefix_len parameter from this function.
>

Next two commits will remove the unused parameters, and make this series of patches easy to review. But indeed this commit has flaws, next two commits are fixes. Should I squash them back?

-- 
Jiang Xin