On Wed, 22 Dec 2004 11:44:11 +0100, Branko ÄŒibej wrote:
>There are basically two issues here. The more common one is when a user
>adds a file on one OS/FS that causes problems on another OS/FS (this
>isn't limited to case collisions). The only way to reliably detect this
>is on the affected client, but of course once its detected, said client
>usually can't do anything about it. The solution here is usually to
>rename the file on the server (using URLs), but we can't do that
>automatically.
It is up to the team that designs clients and servers to build
in warnings and internal logic into all the clients so it is aware of
every other OS combination and warn the user: if you commit this, OS x,
y, z will have the following problems with it. Start with the OSs we
currently support and add more support as new OSs come into the fold.
>Then there's the horrible issue of broken tools changing file case on
>you behind your back (this tends to happen too often on Windows),
>usually after some version of the file has alrealy been committed. SVN
>could detect this change locally, but because of the OS/FS/locale
>dependencies I mentioned above, we must find a way to do this /without/
>using strcasecmp. (Why? Because SVN uses UTF-8 internally, and there's
>no locale-independent case-folding function).
Java has a local-independant case-folding function. Why can't
Subversion's portability layer provide the same sort of thing?
Gili
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Dec 22 22:10:08 2004