Re: svn commit: r1310085 - in /subversion/branches/ev2-export/subversion: include/svn_ra.h libsvn_client/add.c libsvn_ra/ra_loader.c

[Greg Stein <gs...@gmail.com>]

On Apr 5, 2012 5:46 PM, <hw...@apache.org> wrote:
>
> Author: hwright
> Date: Thu Apr  5 21:45:42 2012
> New Revision: 1310085
>
> URL: http://svn.apache.org/viewvc?rev=1310085&view=rev
> Log:
> On the ev2-export branch:
> (this branch is going way beyond export, perhaps it should be renamed)
>
> Add svn_ra_get_commit_editor4(), which returns an Ev2 commit editor.  We
> don't currently marshall Ev2 actions across the wire, this just wraps the
> Ev1 editor within the RA layer.
>
> * subversion/include/svn_ra.h
>  (svn_ra_get_commit_editor4): New.
>  (svn_ra_get_commit_editor3): Deprecate.
>
> * subversion/libsvn_client/add.c
>  (drive_editor): Take an Ev2 editor, and drive it directly, rather than
>    wrapping an Ev1 editor.
>  (mkdir_urls): Request an Ev2 editor from the RA layer.
>
> * subversion/libsvn_ra/ra_loader.c
>  (svn_ra_get_commit_editor4): New.
>
> Modified:
>    subversion/branches/ev2-export/subversion/include/svn_ra.h
>    subversion/branches/ev2-export/subversion/libsvn_client/add.c
>    subversion/branches/ev2-export/subversion/libsvn_ra/ra_loader.c
>
> Modified: subversion/branches/ev2-export/subversion/include/svn_ra.h
> URL:
http://svn.apache.org/viewvc/subversion/branches/ev2-export/subversion/include/svn_ra.h?rev=1310085&r1=1310084&r2=1310085&view=diff
>
==============================================================================
> --- subversion/branches/ev2-export/subversion/include/svn_ra.h (original)
> +++ subversion/branches/ev2-export/subversion/include/svn_ra.h Thu Apr  5
21:45:42 2012
> @@ -865,8 +865,28 @@ svn_ra_rev_prop(svn_ra_session_t *sessio
>  *
>  * Use @a pool for memory allocation.
>  *
> + * @since New in 1.8.
> + */
> +svn_error_t *
> +svn_ra_get_commit_editor4(svn_ra_session_t *session,
> +                          svn_editor_t **editor,
> +                          apr_hash_t *revprop_table,
> +                          svn_commit_callback2_t callback,
> +                          void *callback_baton,
> +                          apr_hash_t *lock_tokens,
> +                          svn_boolean_t keep_locks,
> +                          svn_cancel_func_t cancel_func,
> +                          void *cancel_baton,
> +                          apr_pool_t *scratch_pool,
> +                          apr_pool_t *result_pool);

Pool order

>...
> +svn_error_t *
> +svn_ra_get_commit_editor4(svn_ra_session_t *session,
> +                          svn_editor_t **editor,
> +                          apr_hash_t *revprop_table,
> +                          svn_commit_callback2_t callback,
> +                          void *callback_baton,
> +                          apr_hash_t *lock_tokens,
> +                          svn_boolean_t keep_locks,
> +                          svn_cancel_func_t cancel_func,
> +                          void *cancel_baton,
> +                          apr_pool_t *scratch_pool,
> +                          apr_pool_t *result_pool)

Same

> +{
> +  const svn_delta_editor_t *deditor;
> +  void *dedit_baton;
> +  /* Allocate this in a pool, since the callback will be called long
after
> +     this function as returned. */
> +  struct ccw_baton *ccwb = apr_palloc(result_pool, sizeof(*ccwb));
> +  struct svn_delta__extra_baton *exb;
> +  svn_delta_unlock_func_t unlock_func;
> +  void *unlock_baton;
> +  svn_boolean_t send_abs_paths;
> +
> +  ccwb->original_callback = callback;
> +  ccwb->original_baton = callback_baton;
> +  ccwb->session = session;
> +
> +  SVN_ERR(session->vtable->get_commit_editor(session, &deditor,
&dedit_baton,
> +                                             revprop_table,
> +                                             callback
> +                                                 ?
commit_callback_wrapper
> +                                                 : NULL,
> +                                             callback ? ccwb : NULL,
> +                                             lock_tokens, keep_locks,
> +                                             result_pool));
> +
> +  /* Create the Ev2 editor from the Ev1 editor provided by the RA layer.
*/
> +  SVN_ERR(svn_delta__editor_from_delta(editor, &exb,
> +                                       &unlock_func, &unlock_baton,
> +                                       deditor, dedit_baton,
&send_abs_paths,

Isn't SEND_ABS_PATHS a delayed-use param? Does that have to go on the heap?

>...

Cheers,
-g