Re: svn commit: r39700 - trunk/subversion/libsvn_wc

On Wed, Sep 30, 2009 at 07:46, Bert Huijben <rhuijben_at_sharpsvn.net> wrote:
>...
> +++ trunk/subversion/libsvn_wc/adm_ops.c        Wed Sep 30 04:46:58 2009        (r39700)
> @@ -886,11 +886,24 @@ mark_tree_deleted(svn_wc__db_t *db,
>                                     iterpool));
>         }
>
> -      tmp_entry.schedule = svn_wc_schedule_delete;
> -      SVN_ERR(svn_wc__entry_modify2(db, child_abspath, svn_node_unknown,
> -                            TRUE, &tmp_entry,
> -                            SVN_WC__ENTRY_MODIFY_SCHEDULE,
> -                            iterpool));
> +      /* If this node has no function after the delete, remove it
> +         directly. Otherwise svn_wc__entry_modify2 would do this for us,
> +         but using the entries api would leave the db handle open */
> +      /* ### BH: This check matches the only case in fold_scheduling()
> +                 that removes the entry via delete scheduling */
> +      if (entry->schedule == svn_wc_schedule_add && !entry->deleted)
> +        {
> +          SVN_ERR(svn_wc__entry_remove(db, child_abspath, pool));
> +          SVN_ERR(svn_wc__db_temp_forget_directory(db, dir_abspath, pool));

Woah... I don't think you really want to pull this into the loop. On
the next iteration, it will simply re-open the PDH.

>...
> @@ -921,11 +934,6 @@ mark_tree_deleted(svn_wc__db_t *db,
>                                     SVN_WC__ENTRY_MODIFY_KEEP_LOCAL,
>                                     iterpool));
>     }
> -  else
> -    {
> -      return svn_error_return(
> -        svn_wc__db_temp_forget_directory(db, dir_abspath, pool));
> -    }

I think it should stay "out here".

Cheers,
-g

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2404747
Received on 2009-10-08 02:03:32 CEST