Red Hat rsync - 'sign' patch

jw schultz jw at
Tue Jun 10 08:21:09 EST 2003

On Mon, Jun 09, 2003 at 01:56:34PM -0400, Hardy Merrill wrote:
> I recently became the new rsync maintainer for Red Hat,
> and I just completed a review of the current patches that
> we (Red Hat) maintain for 'rsync'.  After removing three
> unnecessary patches (either already incorporated into
> rsync-2.5.6, or were outdated and couldn't be applied),
> we are left with one patch - rsync-2.5.6-sign.patch -
> which I have attached.

It would be much easier if you would use patches against
current cvs.  That is what the patch would have to be
applied against.  There is quite a bit of difference, at
least in line-numbers between 2.5.6 and cvs.

> AFAICT, back in the vicinity of January 2002, a patch
> (call this patch 'A') was developed to close an 'rsync
> remote hole' - the patch switched variables in
> 'sum_struct' to size_t.  According to some documentation
> I have read, that patch itself introduced some comparisons
> between signed and unsigned, which "can cause rsync to
> not work properly".

What documentation?  Define improperly.  I would like more
than just vague hearsay assertions.  A test case showing
improper behaviour or specific code analysis showing a type
mismatch that actually could (even if only in theory) impact

> The patch I have included (rsync-2.5.6-sign.patch), which
> is the only one currently that we (Red Hat) plan to
> maintain going forward, is a patch that attempts to fix
> the signed vs. unsigned comparison's introduced by patch
> 'A'.

You have introduced a type that is not used anywhere else in
rsync "ssize_t". That is generally not a good idea.  Rsync
has to build and run on all sorts of systems and introducing
a new type without having a fallback definition in rsync.h
and autoconf.  ssize_t wouldn't happen to be the equivalent
of off_t would it?

	J.W. Schultz            Pegasystems Technologies
	email address:		jw at

		Remember Cernan and Schmitt

More information about the rsync mailing list