On 8/18/06, Molle Bestefich <molle.bestefich@gmail.com> wrote:
> Erik Huelsmann wrote:
> > Molle Bestefich wrote:
> > > If that translates to "you think Subversion error messages are fine
> > > the way they are", then you're totally out of sync with the real
> > > world. And real users. No offence :-).
> >
> > If you think that, you didn't read the rest of the message: I think
> > we're totally unclear in our error messages.
>
> Ok. My apologies.
>
> > But I think there's a different reason than most people seem to think:
> > I think we provide errors out-of-context and we should be providing
> > the context instead of making our errors more generic:
>
> I agree that loss of information (from making errors more generic, say) is bad.
> Not sure how your SVN_ERR_W proposal would work.
The SVN_ERR_W macro is exacly for what you say would be probably the
best of both worlds: nice (experienced-user) detail and a 'trail' of
context: it wraps the error received from a call and adds context to
it, by adding a new error 'layer' to the stack.
> > What would it help you if subversion would tell you
> > "Update failed: protocol error"?
[snip]
> I happen to think that the most efficient way to make useful error
> messages is to bubble the message up and add useful information at
> each layer.
>
> svn: update failed
> client: protocol error
> webdav: unexpected response to initial MKCOL
> neon: HTTP authentication failure
Right. That's what SVN_ERR_W is about.
bye,
Erik.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Aug 18 21:49:42 2006