Re: svn commit: r29263 - in branches/mergeinfo-capability/subversion: libsvn_ra_local libsvn_ra_svn

On Feb 10, 2008 5:37 PM, <kfogel_at_tigris.org> wrote:
> Author: kfogel
> Date: Sun Feb 10 17:37:52 2008
> New Revision: 29263
>
> Log:
> On the mergeinfo-capability branch:
>
> Take care of ra_svn:
>
> * subversion/libsvn_ra_svn/ra_svn.h
> (svn_ra_svn__session_baton_t): New field repository_supports_mergeinfo.
>
> * subversion/libsvn_ra_svn/client.c
> (open_session): Initialize above new field.
> (ra_svn_has_capability): If necessary, ask the repository whether it
> supports mergeinfo, and store the answer in the new field.

I'm concerned about this implementation. The former implementation of
ra_svn's svn_ra_has_capability could be called at any time (because it
performed no client/server communication). This implementation
actually runs a get-mergeinfo command. This implies that it can't be
done at an arbitrary time during an RA connection. Have we ensured
that our code never asks about the mergeinfo capability of the server
at any time when a get-mergeinfo command would be illegal (for
example, if the server is responding to a command already, or if we're
in an editor mode, etc)? Does this continue to hold true if we
consider arbitrary API use of svn_ra_get_capability?

--dave

-- 
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org