Greg Stein wrote:
> On Thu, Jan 22, 2009 at 22:29, Hyrum K. Wright <hyrum_at_hyrumwright.org> wrote:
>> ...
>> +++ trunk/subversion/libsvn_subr/sqlite.c Thu Jan 22 13:29:38 2009 (r35403)
>> @@ -185,8 +185,11 @@ svn_sqlite__step(svn_boolean_t *got_row,
>> int sqlite_result = sqlite3_step(stmt->s3stmt);
>>
>> if (sqlite_result != SQLITE_DONE && sqlite_result != SQLITE_ROW)
>> - /* Extract the real error value and reset the statement. */
>> - SVN_ERR(svn_sqlite__reset(stmt));
>> + {
>> + SQLITE_ERR(sqlite_result, stmt->db);
>> + /* Reset the statement. */
>> + svn_error_clear(svn_sqlite__reset(stmt));
>> + }
>
> Eh? The SQLITE_ERR() is going to pop it right out of this function. It
> won't ever run __reset().
>
> Seems like the right thing is to get the svn_error_create() called,
> stash that, then do the clear(reset), then return the svn error. A
> little jiggering in the macros at the top of the file should expose
> the error creation logic.
r35414.
-Hyrum
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1044905
Received on 2009-01-23 15:35:05 CET