On Fri, 2010-12-24, C. Michael Pilato wrote:
> On 12/23/2010 07:27 AM, Julian Foad wrote:
> >>From IRC:
> >
> > <Bert> julianf: We changed error codes in libsvn_wc all over the place.
> > I don't think we see strict error codes as part of the documented
> > behavior, unless it is in the function documentation
>
> I agree with this. Any time a specific error code is intended for use by an
> API function as a specific message to the caller, we document that fact and
> that behavior becomes part of the contract. Undocumented error codes are
> fair game for changing over time.
OK, then that's fine by me too.
Noorul wrote:
> Daniel Shahaf wrote:
> > Why do you have to place the "svn: " prefix here explicitly?
> Normally
> > svn_handle_error2() would do that for you. This is a red flag ("is
> > a wheel being reinvented here?") for me.
>
> We are actually consuming the error here to print it and proceed with
> the other targets.
Noorul, "svn_handle_error2" has a "fatal" flag that controls whether it
terminates the program or not; setting fatal=FALSE would enable you to
continue processing the other targets. But it prints "the error stack"
which is not what you want here - we want just a single error message.
Try using "svn_handle_warning2" instead. That function appears to do
almost exactly what you want here; the only differences I can see is it
inserts "warning: " before the message, which I think is perfect for
this usage, and it doesn't print the extra "\n", which is easily
rectified.
- Julian
