MySQL Bugs: #47791: Several test cases fail on Solaris with error Thread stack overrun

Bug #47791	Several test cases fail on Solaris with error Thread stack overrun
Submitted:	2 Oct 2009 12:23	Modified:	17 Aug 2011 14:31
Reporter:	Olav Sandstå	Email Updates:
Status:	Duplicate	Impact on me:	None
Category:	MySQL Server: General	Severity:	S3 (Non-critical)
Version:	5.1-bugteam, Betony (M2), Celosia (M3),5.6.99,mysql-trunk	OS:	Any (Solaris 10, rhel5 ia64)
Assigned to:	Marc ALFF	CPU Architecture:	Any
Tags:	experimental, pb2, pushbuild, test failure

Description:
The test main.signal_demo3 fails when running it on Solaris. The output contains:

main.signal_demo3                        [ fail ]
        Test ended at 2009-10-02 14:18:53

CURRENT_TEST: main.signal_demo3
mysqltest: At line 122: query 'call proc_1()' failed with wrong errno 1436: 'Thread stack overrun:  135952 bytes used of a 262144 byte stack, and 128000 bytes needed.  Use 'mysqld -O thread_stack=#' to specify a bigger stack.', instead of 1782...

How to repeat:
On Solaris 10 do the following:

1. Use a 6.0 source three (I used mysql-6.0-codebase-bugfixing)
2. ./BUILD/autorun.sh
3. Configure with:

    ./configure CC=/opt/studio12/SUNWspro/bin/cc CXX=/opt/studio12/SUNWspro/bin/CC CFLAGS="-Xa -m64 -mt -g" CXXFLAGS="-m64 -mt -g" --prefix=/home/os136802/install/mysql6 --with-client-ldflags=-static --with-plugins=myisam,innobase --enable-dtrace --with-debug

4. make
5. cd mysql-test
6. ./mysql-test-run.pl signal_demo3

Does not fail on Linux.

Thank you for the report.

Verified as described.

This test was probably passing for quite some time. We need to find out what change causes it fail fist.

Bug verification team: Have you verified whether this bug exists in releases prior to 6.0?

Thanks,
Manyi

Making the test case experimental on Solaris
staring from Celosia (next-mr).

List of test cases:
  - main.signal_demo3
  - main.sp
  - rpl.rpl_row_sp011
  - sys_vars.max_sp_recursion_depth_func

It's now also failing in Betony/Celosia.

Adding rpl.rpl_row_sp011 to this list.

Adding main.subselect_sj2 & main.subselect_sj2_jcl6 to this list.

Inspection of the main.sp case (in 5.1-bugteam) indicates that stack
frame size for mysql_execute_command() is more than twice the size in
non-optimized build compared to optimized build on sparc32 with
sunstudio12u1. (9.4k compared to 4.3k.)

In main.sp, stored procedure recursion puts multiple invocations of
mysql_execute_command() on the stack, and with enough levels one will
hit the ceiling imposed by check_stack_overrun(), in this case called
by sp_head::execute().

The reason for these differences is likely that the compiler in
non-optimized mode makes no effort to optimize/reuse variable space in
the stack frame for sub-scopes (there are a few large objects
allocated on the stack in sub-scopes in mysql_execute_command()).

Quick reproduction (snipped from main.sp):

   delimiter |
   create procedure bug10100p(prm int, inout res int)
   begin
     set res = res * prm;
     if prm > 1 then
       call bug10100p(prm - 1, res);  
     end if;
   end|
   create procedure bug10100t(prm int)
   begin
     declare res int;
     set res = 1;
     call bug10100p(prm, res);
     select res;
   end|
   delimiter ;

   set session max_sp_recursion_depth=4;
   call bug10100t(5);

and inspect the behavior of the stack pointer in a deugger.

Pushed into mysql-trunk 5.6.99-m5 (revid:alexander.nozdrin@oracle.com-20101113155825-czmva9kg4n31anmu) (version source revid:vasil.dimov@oracle.com-20100629074804-359l9m9gniauxr94) (merge vers: 5.6.99-m4) (pib:21)

Pushed into mysql-next-mr (revid:alexander.nozdrin@oracle.com-20101113160336-atmtmfb3mzm4pz4i) (version source revid:vasil.dimov@oracle.com-20100629074804-359l9m9gniauxr94) (pib:21)

In the release build of 5.6.1-m5, this symptom shows up in test "subquery_sj_none" (specific to Solaris 10 Sparc in debug mode).

Correction:
... (specific to Solaris 10 in debug mode).

It happens on both Sparc-64 and x86_64, the test passes on x86 (32 bit).

Fails every day in the nightly release run of the trunk http://pb2.norway.sun.com/?template=mysql_show_test_failure&test_failure_id=&test_output_i...

Also seen in sys_vars.sp_recusrion_depth

http://pb2.norway.sun.com/?action=archive_download&archive_id=2866255&pretty=please

sys_vars.max_sp_recursion_depth_func     w13 [ exp-fail ]
        Test ended at 2011-01-21 00:55:00

CURRENT_TEST: sys_vars.max_sp_recursion_depth_func
mysqltest: At line 64: query 'CALL sp_addRecords(0,8)' failed: 1436: Thread stack overrun:  138960 bytes used of a 262144 byte stack, and 128000 bytes needed.  Use 'mysqld --thread_stack=#' to specify a bigger stack.

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/129827

3563 Anitha Gopi	2011-01-28
      Bug#47791 : Moved main.subquery_sj_none to experimental group

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/129990

3576 Anitha Gopi	2011-01-31 [merge]
      Bug#47791 : Moved main.subquery_sj_none to experimental group

Pushed into mysql-trunk 5.6.2 (revid:anitha.gopi@sun.com-20110131042308-0l2ggxhv1pgqpmet) (version source revid:anitha.gopi@sun.com-20110131042308-0l2ggxhv1pgqpmet) (merge vers: 5.6.2) (pib:24)

Moving back to verified state. Patch pushed was for just moving a test to experimental state

Duplicate of BUG#45445.