Re: svn commit: r10143 - in trunk/subversion: libsvn_subr tests/libsvn_subr

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

Greg Hudson wrote:
> On Mon, 2004-07-05 at 09:06, jpieper@tigris.org wrote:
> > -  while (path->len > 0 && path->data[path->len - 1] != '/')
> > +  while (path->len > 0 && path->data[path->len] != '/')
> >      --path->len;
> > +
> > +  if (path->len == 0 && path->data[0] == '/')
> > +    path->len++;
> 
> The valid indices of path->data range from 0 to path->len - 1.  Your new
> code starts accessing path->data[path->len] right off the bat, which is
> invalid, and then accesses path->data[0] at the end if path->len is 0,
> which is also invalid if path->len was 0 to start with.

I though svn_stringbuf_t maintained the invariant that
path->data[path->len] == '\0'?

-Josh

-- 
To be is to be related.
		-- C.J. Keyser.

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

Re: svn commit: r10143 - in trunk/subversion: libsvn_subr tests/libsvn_subr

[Greg Hudson <gh...@MIT.EDU>]

On Mon, 2004-07-05 at 11:50, Josh Pieper wrote:
> > The valid indices of path->data range from 0 to path->len - 1.  Your new
> > code starts accessing path->data[path->len] right off the bat, which is
> > invalid, and then accesses path->data[0] at the end if path->len is 0,
> > which is also invalid if path->len was 0 to start with.
> 
> I though svn_stringbuf_t maintained the invariant that
> path->data[path->len] == '\0'?

Oh, of course you're right.

But the purpose of that invariant is to make str->data work as a C
string.  Using it for other purposes is confusing (as I just
demonstrated by becoming confused), so I like Klaus's approach better.


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