On Tue, Jun 26, 2012 at 07:47:16AM +0100, Daniel Shahaf wrote:
> Stefan Sperling wrote on Mon, Jun 25, 2012 at 23:27:37 +0200:
> > On Mon, Jun 25, 2012 at 10:23:38PM +0100, Daniel Shahaf wrote:
> > > From the peanut gallery: as long as svn1.7 errors out, but doesn't
> > > corrupt any of the data or metadata, perhaps that's something we're
> > > willing to live with?
> >
> > This approach might be suboptimal for GUI apps that use status to
> > refresh icons etc.
>
> I guess we didn't use a special error code for the "New value in enum"
> case... (so GUIs can recognize it and treat it as ETHISSVNISTOOOLD)
They could catch the error, but existing GUIs based on 1.7 code won't
catch it right now because they don't need to. Requiring the error to
be caught is still a breaking change for existing code.
And I don't like the idea of pushing responsibility for backwards
compat onto the shoulders of users of our APIs.
Received on 2012-06-26 10:55:46 CEST