Philip Martin wrote:
>Branko ÄŒibej <brane@xbc.nu> writes:
>
>
>
>>>svn_cl__propedit in propedit-cmd.c calls svn_path_split_nts and
>>>assumes that it gets a sensible dirpath. Most (but not all) other
>>>places use something like
>>>
>>> svn_path_split_nts (path, dirpath, basename, pool);
>>> if (svn_path_is_empty_nts (dirpath))
>>> dirpath = ".";
>>>
>>>Is this a bug in svn_cl__propedit or svn_path_split_nts?
>>>
>>>
>>>
>>It's a bug in svn_path_split. It should be returning useful,
>>canonicalized paths. I think we have an issue abou this, and I sort of
>>feel the real culprit is APR, not SVN.
>>
>>
>
>In that case we ought to fix svn_path_split and remove the all the
>svn_path_is_empty_nts tests. I don't think it's an APR problem
>though, it's the absence of separators such as
>
> svn_path_split_nts( "foo" ) gives "" and "foo"
>
No, it's IMHO an APR bug that it doesn't recognize "" as the current
directory. So I must reverse myself -- the problem is not in svn_path_split.
>This behaviour is enforced by the path-split regression tests. Within
>the subversion code there are some places where "" works as the
>current directory, and there are some where it doesn't work.
>
"" should always work, and we shouldn't have to do (platform-specific?)
checks for that in our code, those should be in APR. (All of which
implies that our canonicalization should just strip off any leading
".[/]", thereby changeing "." to "".)
--
Brane Čibej <brane_at_xbc.nu> http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Aug 12 01:09:06 2002