Re: regression tests failing over ra_dav

On Wed, Oct 23, 2002 at 09:30:58PM -0500, cmpilato@collab.net wrote:
> Philip Martin <philip@codematters.co.uk> writes:
>
> > svn_ra_dav__get_file calls filter_props, which calls
> > set_special_wc_prop, which uses add_prop_to_hash to put
> > svn_stringbuf_t pointers into the hash. Looks like rev 3459 is the
> > culprit.
>
> Culprit? I'm offended. My change merely revealed an existing bug
> that we were lucky to get away with because the data and len members
> of svn_string_t's and svn_stringbuf_t's fell in the right places. The
> interfaces for ra->get_file and get_dir state that the hash has
> string_t's in it ... add_prop_to_hash() was clearly violating that.
>
> Either way, I'm on the fix to the *real* culprit now...

Right-o.

Note that if you had made len/blocksize the first field of svn_stringbuf_t,
then you'd *really* be evil. But you put the pool as the first param, so
when somebody goes to deref the "data" pointer, they are still successful
since that is a valid pointer. But if they deref'd a small length value?
Blam!

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org