On Wed, Sep 5, 2012 at 11:11 AM, Julian Foad <[email protected]> wrote:
> Stefan Sperling wrote:
>
>> On Tue, Sep 04, 2012 at 03:01:22PM -0400, Paul Burba wrote:
>>> I just noticed that this causes a segfault when updating with
>>> the quiet option:
>>
>> Thanks for digging into this and explaining the problem, Paul.
>>
>> I cannot think of a better way to fix this than recording in
>> opt_state whether the notifier is being used. That way we
>> have a way of telling whether it is safe to ask the notifier
>> for a list of conflicted paths.
>>
>> What you do think?
>
> Just a drive-by comment: isn't "ctx->notify_func2 != NULL" a good "notifier
> is in use" flag?
I think you mean ctx->notify_func? Because
libsvn_client/ctx.c:svn_client_create_context unconditionally sets the
notify_func2 member to libsvn_client/ctx.c:call_notify_func, which
calls...ctx->notify_func. So the latter would work, but it seems a
bit dubious since svn_client_ctx_t->notify_func is deprecated:
typedef struct svn_client_ctx_t
{
/** main authentication baton. */
svn_auth_baton_t *auth_baton;
/** notification callback function.
* This will be called by notify_func2() by default.
* @deprecated Provided for backward compatibility with the 1.1 API.
* Use @c notify_func2 instead. */
svn_wc_notify_func_t notify_func;
I think Stefan's suggested patch is the clearer solution.
--
Paul T. Burba
CollabNet, Inc. -- www.collab.net -- Enterprise Cloud Development
Skype: ptburba
> (And I removed 'commits@' from the CC.)
>
> - Julian
