[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [BUG] svnserve aborts on 'svn log' from 1.4.3 client

From: Daniel Rall <dlr_at_collab.net>
Date: 2007-05-31 20:30:17 CEST

On Thu, 31 May 2007, Malcolm Rowe wrote:

> On Thu, May 31, 2007 at 07:44:01AM -0500, Hyrum K. Wright wrote:
> > >> and it is aborting because **fmt is 'b'. Count is 6, nesting level is 0.
> > >>
> > >> I think the pattern specified in serve.c:log_cmd line 1557 must have
> > >> an optional parameter wrong or need one added, but I haven't figured
> > >> it out yet.
> > >>
> > >
> > > Just guessing, but is it because the pattern includes an optional
> > > boolean? We don't support those in ra_svn -- there's no way to indicate
> > > back to the caller whether the parameter was present or not.
> >
> > Is there another way to pass an optional boolean? Couldn't the caller
> > assume some value, in this case FALSE, if the parameter isn't present?
> >
>
> It could, but the parsing code has no way to communicate that back,
> because it's expecting a svn_boolean_t *. (I mean, I suppose it _could_
> just not touch the value and rely on the caller to set the default
> _first_, but that would be different to how we handle other optional
> types.
>
> I think for other cases we may have transmitted a boolean inside an
> optional tuple, e.g. (?b). That's a bit sucky, though - it's using the
> implementation constraints to drive the interface.

I was surprised to see that naked "?b" in the protocol string. Try
Malcolm's suggestion -- we've had to do that in other places as well.

  • application/pgp-signature attachment: stored
Received on Thu May 31 20:31:24 2007

This is an archived mail posted to the Subversion Dev mailing list.