Re: svn commit: r10133 - in trunk/subversion: include libsvn_subr tests/clients/cmdline tests/libsvn_subr

[Klaus Rennecke <kr...@tigris.org>]

Hmm, I think this will backtrack over the host name in URLs. Interesting 
feature, but intended? And I think absolute_path will be wrong for full 
URLs except file:///

If there is a schema part, skip initial segment excluding the following 
slash? Something like while(*src && *src != '/') { *(dst++) = *(src++); 
} perhaps.

Then the following slash which will be swallowed by the next if{} and 
trigger absolute_path. And avoid backtracking over the host name.

Host names are case-insensitive by the way, that might be another 
transformation if strict equality is a goal. But this may clash with the 
callers that use this function for plain file names.

/Klaus


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

Re: svn commit: r10133 - in trunk/subversion: include libsvn_subr tests/clients/cmdline tests/libsvn_subr

[Josh Pieper <jj...@pobox.com>]

Klaus Rennecke wrote:
> Hmm, I think this will backtrack over the host name in URLs. Interesting 
> feature, but intended? And I think absolute_path will be wrong for full 
> URLs except file:///
> 
> If there is a schema part, skip initial segment excluding the following 
> slash? Something like while(*src && *src != '/') { *(dst++) = *(src++); 
> } perhaps.
> 
> Then the following slash which will be swallowed by the next if{} and 
> trigger absolute_path. And avoid backtracking over the host name.

You are of course correct, I'll work on fixing it and adding some test
cases.

Speaking of review, after I put this change in, svn_path_canonicalize
could probably use several pairs of extra eyes.  I copied most of the
logic from apr_filepath_merge, but that kind of pointer manipulation
is just the kind of code that security holes like to live in.

> Host names are case-insensitive by the way, that might be another 
> transformation if strict equality is a goal. But this may clash with the 
> callers that use this function for plain file names.

I think it's best to not worry this for now.

-Josh

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