On 03/12/2012 02:46 AM, 罗勇刚(Yonggang Luo)  wrote:
>
>
> 2012/3/12 Jelmer Vernooij <<email address hidden> <mailto:<email address hidden>>>
>
>     Am 12/03/12 02:10, schrieb Yonggang Luo:
>
>         2012/3/12 Yonggang Luo<<email address hidden>
>         <mailto:<email address hidden>>>
>
>             2012/3/12 Jelmer Vernooij<<email address hidden>
>             <mailto:<email address hidden>>>
>
>                     2012/3/11 Jelmer Vernooij<<email address hidden>
>                     <mailto:<email address hidden>>>
>
>                             Please read the test result carefully. And
>                             you will know why do the
>
>                         things in
>
>                             this way.
>                             The test result is under
>
>                             Description of the Change
>
>                         That only explains *what* is failing for you.
>                         It doesn't explain
>
>             *why*
>
>                 it
>
>                         is failing, and why your fix addresses that.
>                         That's the bit I'm
>
>                 interested
>
>                         in.
>
>                     It's failed just because Python's maketemp didn't
>                     return the corrent
>                     tempfile file path.
>                     For example, tempfile.mkdtemp() on windows will
>                     gives us with the file
>
>                 path
>
>                     c:\users\uername\appdata\local\temp\hash_code ,
>                     but the real file path
>                     is C:\Users\username\AppData\Local\Temp\hash_code.
>                     When I use
>                       tempfile.mkdtemp(dir =
>                                     os.getenv("TEMP", None) or
>                     os.getenv("TMP", None))
>                     it's will give us the exact file path with the
>                     windows file system.
>                     TEMP and TMP is recorded by the OS. and even if
>                     TEMP and TMP is not
>
>                 exist,
>
>                     it's sill works, just because dir accept None as
>                     a default input.
>
>                 TEMP and TMP can be set by the user too as far as I
>                 know. What if the
>
>             user
>
>                 has set them to
>                 "C:\users\USeRName\AppData\Local\Temp\hash_code" ?
>
>             That's a
>
>                 perfectly valid path, as path lookups are case
>                 insensitive.
>
>                 If TEMP and TMP do not exist, we get the same
>                 behaviour as currently so
>                 users who have that will still hit that bg - and that
>                 means this fix is
>                 incomplete in that situation.
>
>                     The unittest failed, is just because Subversion
>                     changed the file path
>                     from  c:\users\uername\appdata\local\temp\hash_code
>                     to C:\users\uername\appdata\local\temp\hash_code,
>                     but under Windows
>
>             file
>
>                     system, these two path represents the same path,
>                     so it's should not be FAILED.
>
>                 Then why not simply call os.path.normcase() on both
>                 paths before
>
>             comparing
>
>                 them? That seems like a much cleaner solution to me.
>
>             That's why I ask you better solution, thanks.
>
>         It's not works, still return the path with incorrect case.
>
>     Can you please elaborate? What doesn't work ? Does
>     os.path.normcase not return a normalized case?
>
> It's didn't change the result, just act with os.path.normcase 's 
> documented case.
> The path is return without change except \ are replaced with /.
>
It sounds like you're using os.path.normpath, not os.path.normcase.

Jelmer