Re: Trouble with non-ASCII characters in file names

Florian Weimer <fw@deneb.enyo.de> writes:

> #!/bin/sh
> set -ex
> svnadmin create repo
> mkdir wc
> mkdir wc/dir1
> echo contents > `printf "wc/dir1/gr\374n"`
> svn import -m "import" wc file://`pwd`/repo/
> rm -rf wc
> svn checkout file://`pwd`/repo/ wc
> cd wc
> svn copy dir1 dir2
> svn commit -m "commit"
>
> Depending on the locale, this either fails during checkout with:
>
> svn: Malformed XML: not well-formed (invalid token) at line 3
> (UTF-8 case)

The sequence gr\374n is not a valid UTF8 sequence. If you import
using a 8859-1 locale the sequence will get converted to UTF8 and a
subsequent checkout will work.

This illustrates a problem I raised last year, when the client expects
UTF8 input it doesn't validate that the input really is UTF8. This
problem affects both URLs and wc paths.

> If LC_CTYPE=ISO-8859-1, it fails at "svn copy" with:
>
> svn: Can't recode string: Invalid argument

That is a bug in copy, I've just fixed it on the trunk (r8361).

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org