On Aug 12, 2009, at 5:32 PM, Greg Stein wrote:
> On Wed, Aug 12, 2009 at 20:16, Hyrum K.
> Wright<hyrum_at_hyrumwright.org> wrote:
>> ...
>> +++ trunk/subversion/libsvn_client/merge.c Wed Aug 12 11:16:02
>> 2009 (r38702)
>> ...
>> @@ -3675,8 +3677,8 @@ find_gaps_in_merge_source_history(svn_re
>> peg_rev.value.number = young_rev;
>>
>> SVN_ERR
>> (svn_client__get_history_as_mergeinfo(&implicit_src_mergeinfo, url,
>> &peg_rev, young_rev,
>> old_rev,
>> - ra_session, NULL,
>> - merge_b->ctx,
>> scratch_pool));
>> + ra_session, merge_b-
>> >ctx,
>> + result_pool));
>
> Why did the pool change here?
Hmm, not really sure. Actually this entire function only needs a
scratch_pool, since the return values are just revnum_t's. Updated in
r38729.
>
>> ...
>> +++ trunk/subversion/libsvn_client/url.c Wed Aug 12 11:16:02
>> 2009 (r38702)
>> @@ -46,9 +46,16 @@ svn_client_url_from_path(const char **ur
>> apr_pool_t *pool)
>> {
>> svn_opt_revision_t revision;
>> + svn_client_ctx_t *ctx;
>> +
>> + SVN_ERR(svn_client_create_context(&ctx, pool));
>> +
>> + if (!svn_path_is_url(path_or_url))
>> + SVN_ERR(svn_dirent_get_absolute(&path_or_url, path_or_url,
>> pool));
>> +
>> revision.kind = svn_opt_revision_unspecified;
>> return svn_client__derive_location(url, NULL, path_or_url,
>> &revision,
>> - NULL, NULL, NULL, pool);
>> + NULL, ctx, pool, pool);
>> }
>
> Seems a bit "dangerous" to leave this ctx around waiting for the pool
> to be cleared. Couldn't we end up with some dangling handles on DBs?
Agreed, we should probably destroy this manually. But we don't have a
manual destroy function for the client context. Subpool?
-Hyrum
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2383330
Received on 2009-08-13 17:30:45 CEST