Lars, I believe that we specifically disallow "_svn" in the repository
for interoperability between different clients that are using both
".svn" and "_svn" administrative directories. If we allowed "_svn"
into the repository, how would the users who have SVN_ASP_DOT_NET_HACK
pull that directory from the repository?
- Dan
On Sun, 11 Nov 2007, Lars Kruse wrote:
> [beware: this message was originally posted on the users list yesterday
> (http://subversion.tigris.org/servlets/ReadMsg?list=users&msgNo=71890)
> Karl Fogel suggested, to put it here, too]
>
>
> Hi,
>
> I am using subversion v.1.4.4 on debian (x86, lenny).
>
> Recently I encountered a strange problem: I was unable to update a working
> directory when giving its complete path (output: 'skipping "."').
> This problem disappeared, as soon as I changed to that directory before
> updating.
>
> I tracked down the problem to the name of the last directory node: the
> problem only occours, when the directory is called "_svn".
>
> It seems like subversion ignores directories named like that due to
> compatibility towards the windows environment.
> (http://subversion.tigris.org/svn_1.3_releasenotes.html - "Enhancements &
> bugfixes" -> "Official support for windows '_svn' directories")
>
> I am running linux and I did not set the mentioned environment variable
> (SVN_ASP_DOT_NET_HACK), thus I do not expect, that directories called "_svn"
> are treated as invalid.
>
> See the commands and comments below to reproduce this behaviour:
> (a working local copy called 'test' is required)
>
> # check the working copy
> $ svn up test
> Revision 5.
> # ok
>
> # check if the working copy is usable after renaming to "_svn"
> $ mv test _svn
> $ svn up _svn/
> Skipped '.'
> # failure!
>
> # change into the directory and do the same
> $ cd _svn/
> $ svn up
> Revision 5.
> # ok
>
> # are subdirectories accessible?
> $ cd ..
> $ svn up _svn/wg-ludwig/
> Revision 5.
> # ok
>
>
> btw: the same applies to directories named ".svn". But in this case it would be
> less likely, that someone names a working copy like that.
> My case of using "_svn" as a name is real. It took me some hours to
> identify the root cause :(
>
> It seems to me, as if svn would be guessing, that the parent directory is
> meant, in case someone accesses directories called ".svn" or "_svn".
> I do not expect a command line client to do this kind of guessing.
>
> In case you really have to rely on this guessing, I would prefer, if there
> would be a test (e.g. is there a subdirectory ".svn" inside of the specified
> directory?), that would make sure, that the specified direcotory was _really_ a
> mistake and not just "could-be-a-mistake".
- application/pgp-signature attachment: stored
Received on Mon Nov 12 18:59:32 2007