Hi Monty, Unfortunately I have no idea what this problem is, except that I have never seen it in the insert test before. Has the insert test been modified recently? I guess it is more likely that the select is delivering a different number of rows than the INSERT failing. On Nov 25, 2010, at 7:21 PM, Michael Widenius wrote:

Hi!

I just merged MySQL 5.1.53 into MariaDB and as part of that, did run the pbxt test suite.

I encountered this test failure:

mysql-test-run --suite=pbxt

URRENT_TEST: pbxt.insert --- /home/my/maria-5.1-release/mysql-test/suite/pbxt/r/ insert.result 2010-11-25 12:02:26.083226000 +0200 +++ /home/my/maria-5.1-release/mysql-test/suite/pbxt/r/ insert.reject 2010-11-25 19:33:48.281738504 +0200 @@ -298,7 +298,7 @@ insert into t2 select t1.* from t1, t2 t, t3 where t1.id1 = t.id2 and t.id2 = t3.id3; select count(*) from t2; count(*) -25500 +25495 drop table t1,t2,t3; create table t1 (a int, b int); insert into t1 (a,b) values (a,b);

This only happened once for this test. (I did get similar failures with one other test for that run, but not for new runs).

I could not repeat it by running the test or suite again.

I also tested to run the test with valgrind, but not problems found.

I noticed this also happens in buildbot from time to time:

See http://buildbot.askmonty.org/buildbot/builders/gentoo-x86-dbg/builds/396/ste... , search after pbxt.insert

Paul, have you ever encountered something like this? Any clues where to start looking? (The tree I am working with is 5.1-release)

Regards, Monty