From: Jorgen Loland
Date: May 27 2011 1:18pm
Subject: Re: [Resend] bzr commit into mysql-trunk branch (guilhem.bichot:3315)
 Bug#12595210
List-Archive: http://lists.mysql.com/commits/138336
Message-Id: <4DDFA4A0.4070701@oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 8bit

LGTM

On 05/26/2011 06:24 PM, Guilhem Bichot wrote:
> [This commit e-mail is a repeat.]
>
> #At file:///home/mysql_src/bzrrepos_new/mysql-next-mr-opt-backporting-wl4800/ based on revid:guilhem.bichot@stripped
>
>   3315 Guilhem Bichot	2011-05-26
>        fix for BUG#12595210 - JSON SYNTAX ERROR ASSERT ON WHERE FIELD NOT IN SUBQUERY
>       @ mysql-test/suite/optimizer_trace/t/optimizer_trace_bugs.test
>          test for bug; used to crash. BUG 12595688 was filed along the way.
>       @ sql/item_subselect.cc
>          when doing IN->EXISTS transformation of IN(select),
>          if "select" contains some constant WHERE/HAVING clause we may evaluate it;
>          this may evaluate a subquery (if the WHERE/HAVING has a subquery). This caused
>          a syntax error like this:
>                          {
>                            "transformation": {
>                              "select#": 2,
>                              "from": "IN (SELECT)",
>                              "to": "EXISTS (CORRELATED SELECT)",
>                              "chosen": true** invalid JSON (missing key) ** ,
>                              "?": {
>                                "subselect_execution": {
>                                  "select#": 3,
>          Where:
>          "subselect_execution" and its outer unnamed object come from join->conds->fix_fields()
>          called in Item_in_subselect::row_value_in_to_exists_transformer()
>          (transforming select#2 with IN->EXISTS, evaluating its WHERE clause along the way,
>          which means evaluating the constant select#3).
>          The fix is to open a named object when about to evaluate the constant WHERE,
>          so that subquery execution fits into it.
>          Same for HAVING.
>          Same for transformations of returning-single-column subqueries.
>
>      modified:
>        mysql-test/suite/optimizer_trace/r/optimizer_trace2_no_prot.result
>        mysql-test/suite/optimizer_trace/r/optimizer_trace_bugs.result
>        mysql-test/suite/optimizer_trace/r/optimizer_trace_no_prot.result
>        mysql-test/suite/optimizer_trace/r/optimizer_trace_ps_prot.result
>        mysql-test/suite/optimizer_trace/r/optimizer_trace_range_no_prot.result
>        mysql-test/suite/optimizer_trace/t/optimizer_trace_bugs.test
>        sql/item_subselect.cc


-- 
Jørgen Løland | Senior Software Engineer | +47 73842138
Oracle MySQL
Trondheim, Norway