On Wed, May 11, 2011 at 3:01 AM, Greg Stein <gstein_at_gmail.com> wrote:
> On Tue, May 10, 2011 at 16:31, <pburba_at_apache.org> wrote:
>>...
>> +++ subversion/trunk/subversion/libsvn_client/merge.c Tue May 10 20:31:32 2011
>>...
>> @@ -4619,9 +4626,10 @@ record_skips(const char *mergeinfo_path,
>> ### skipped? */
>> ;
>> }
>> + svn_pool_destroy(iterpool);
>> SVN_ERR(update_wc_mergeinfo(NULL, merge_b->target_abspath,
>> mergeinfo_path, merges,
>> - is_rollback, merge_b->ctx, pool));
>> + is_rollback, merge_b->ctx, scratch_pool));
>
> If you move that destroy to after the update call, then you can pass
> iterpool for that func's scratch_pool. Then it'll get destroyed. That
> will help to lower the high-water mark for the caller-provided
> scratch_pool (which may be important, depending on how long the caller
> lets it go before clearing it).
Done r1101957
Paul
>>...
>
> Cheers,
> -g
>
Received on 2011-05-11 18:34:02 CEST