On 5/7/07, Eric Gillespie <epg@pretzelnet.org> wrote:
> > I have a question. Why do you think that the MD5 checksums should be
> > passed in raw format instead of in cstring format? Your change breaks
> > compatibility with any code which previously used the checksum return
> > value, so we should have a good reason if we want to change it.
>
> The old code was the incompatible one; why should the Python
> binding change things like this on you? To be complete, we'd
> have to wrap the functions that take that raw checksum
> (e.g. svn_wc_process_committed*) to convert the hex string back
> to the raw data; why convert it back and forth? Python
> applications can always use binascii.b2a_hex to get the
> human-readable hex form.
Ah, ok, I get it. You're basically saying that the old bindings were
effectively broken because the result string couldn't be used by any
Subversion functions which accept checksums as input. In this case
it's in my opinion fine to change the typemap.
At release time, we should probably add a note about this to the
CHANGES file in case somebody depends on the old behavior.
If nobody objects to this change, feel free to commit.
Cheers,
David
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon May 7 20:08:17 2007