Bug #43090 ARCHIVE engine related warnings on Windows debug build
Submitted: 22 Feb 2009 16:05 Modified: 15 Apr 2009 1:40
Reporter: Ingo Strüwing Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Server: Archive storage engine Severity:S3 (Non-critical)
Version:6.0 OS:Microsoft Windows
Assigned to: Ingo Strüwing CPU Architecture:Any
Triage: Triaged: D4 (Minor) / R1 (None/Negligible) / E1 (None/Negligible)

[22 Feb 2009 16:05] Ingo Strüwing
Description:
main.partition_archive                   [ fail ]  Found warnings/errors in server log file!
        Test ended at 2009-02-18 09:45:31
file_name	line
C:/cygwin/home/istruewing/bzrroot/mysql-6.0-bug38133-6/mysql-test/var/1/mysqld.1/mysqld.err	Error: Freeing pointer out of range at line 399, '.\ha_archive.cc'
C:/cygwin/home/istruewing/bzrroot/mysql-6.0-bug38133-6/mysql-test/var/1/mysqld.1/mysqld.err	Error: Freeing pointer out of range at line 399, '.\ha_archive.cc'
C:/cygwin/home/istruewing/bzrroot/mysql-6.0-bug38133-6/mysql-test/var/1/mysqld.1/mysqld.err	Error: Freeing pointer out of range at line 399, '.\ha_archive.cc'
... (many of these follow)
^ Found warnings!!
ok

main.archive                             [ fail ]  Found warnings/errors in server log file!
        Test ended at 2009-02-18 09:59:42
file_name	line

C:/cygwin/home/istruewing/bzrroot/mysql-6.0-bug38133-6/mysql-test/var/2/mysqld.1/mysqld.err	Error: Freeing pointer out of range at line 399, '.\ha_archive.cc'
... (many of these follow)

main.archive_aio_posix                   [ fail ]  Found warnings/errors in server log file!
        Test ended at 2009-02-18 09:59:44
file_name	line

C:/cygwin/home/istruewing/bzrroot/mysql-6.0-bug38133-6/mysql-test/var/1/mysqld.1/mysqld.err	Error: Freeing unallocated data at line 399, '.\ha_archive.cc'
... (many of these follow)

main.archive_bitfield                    [ fail ]  Found warnings/errors in server log file!
        Test ended at 2009-02-18 10:00:10
file_name	line

C:/cygwin/home/istruewing/bzrroot/mysql-6.0-bug38133-6/mysql-test/var/2/mysqld.1/mysqld.err	Error: Freeing unallocated data at line 399, '.\ha_archive.cc'
... (many of these follow)

main.archive_gis                         [ fail ]  Found warnings/errors in server log file!
        Test ended at 2009-02-18 10:00:12
file_name	line

C:/cygwin/home/istruewing/bzrroot/mysql-6.0-bug38133-6/mysql-test/var/1/mysqld.1/mysqld.err	Error: Freeing unallocated data at line 399, '.\ha_archive.cc'
... (many of these follow)

main.mysqldump-max                       [ fail ]  Found warnings/errors in server log file!
        Test ended at 2009-02-18 10:43:23
file_name	line

C:/cygwin/home/istruewing/bzrroot/mysql-6.0-bug38133-6/mysql-test/var/1/mysqld.1/mysqld.err	Error: Freeing unallocated data at line 399, '.\ha_archive.cc'
^ Found warnings!!
ok

The offending line in ha_archive.cc is:
    my_free((uchar*) share, MYF(0));
That's in free_share().
The share is allocated in get_share() at this line:
    if (!my_multi_malloc(MYF(MY_WME | MY_ZEROFILL),
                          &share, sizeof(*share),
                          &tmp_name, length+1,
                          NullS)) 

When running with --debug, the trace shows that the saher is allocated with my_malloc(), while it is freed with _myfree(), which is part of SAFEMALLOC. No wonder that it complains. It didn't record the allocation.

In the top-level CMakeLists.txt, we have this:

# Do not use SAFEMALLOC for Windows builds, as Debug CRT has the same functionality
# Neither SAFE_MUTEX works on Windows and it has been explicitely undefined in 
# my_pthread.h
IF(NOT WIN32)
  SET(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -DSAFEMALLOC -DSAFE_MUTEX")
  SET(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} -DSAFEMALLOC -DSAFE_MUTEX")
ENDIF(NOT WIN32)

In storage/archive/CMakeLists.txt, we explicitly define SAFEMALLOC and SAFE_MUTEX. This is not a big problem, as long as all memory allocated by ARCHIVE is also freed by it and vice versa. my_malloc() and my_free() are macros that evaluate to _mymalloc() and _myfree() if SAFEMALLOC is defined. Other engines do also explicitly turn these on, whithout problems.

However, ha_archive.cc uses my_multi_malloc(), which is not redefined by SAFEMALLOC. It calls my_malloc() internally to do the final allocation. But since mysys is not compiled with SAFEMALLOC, it calls the real my_malloc().

How to repeat:
Make a debug build on Windows.
Run the mentioned tests.

Suggested fix:
=== modified file 'storage/archive/CMakeLists.txt'
--- storage/archive/CMakeLists.txt      2008-01-30 02:58:57 +0000
+++ storage/archive/CMakeLists.txt      2009-02-22 16:01:25 +0000
@@ -13,8 +13,14 @@
 # along with this program; if not, write to the Free Software
 # Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 
-SET(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -DSAFEMALLOC -DSAFE_MUTEX")
-SET(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} -DSAFEMALLOC -DSAFE_MUTEX")
+#
+# NOTE: The use of my_multi_malloc() in ha_archive.cc excludes the
+# definition of SAFEMALLOC and SAFE_MUTEX here, as long as it is not
+# defined in the top-level CMakeList.txt. So do not define it here
+# explicitly, but inherit from the top-level file.
+#
+#SET(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -DSAFEMALLOC -DSAFE_MUTEX")
+#SET(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} -DSAFEMALLOC -DSAFE_MUTEX")
 
 INCLUDE_DIRECTORIES(${CMAKE_SOURCE_DIR}/include ${CMAKE_SOURCE_DIR}/zlib
                     ${CMAKE_SOURCE_DIR}/sql
[22 Feb 2009 16:19] Bugs System
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/67113

2765 Ingo Struewing	2009-02-22
      Bug#43090 - ARCHIVE engine related warnings on Windows debug build
      
      A windows debug build created warnings in the error log.
      In a windows debug build, we do not define SAFEMALLOC and SAFE_MUTEX.
      But the ARCHIVE engine explicitly defined them in CMakeLists.txt.
      Ha_archive.cc uses my_multi_malloc(). This is part of mysys, which,
      like most parts of the server, are compiled without SAFEMALLOC.
      So my_multi_malloc() calls my_malloc() directly and is not
      redirected to the SAFEMALLOC function _mymalloc(). But since
      ha_archive.cc frees the allocated memory with my_free(), which
      is redefined to _myfree(), SAFEMALLOC complains about a free, for
      which it didn't see an allocation.
      
      Fixed by not explicitly defining SAFEMALLOC and SAFE_MUTEX in
      ARCHIVE's CMakeLists.txt.
     @ storage/archive/CMakeLists.txt
        Bug#43090 - ARCHIVE engine related warnings on Windows debug build
        Disabled explicit definition of SAFEMALLOC and SAFE_MUTEX.
[25 Feb 2009 13:10] Ingo Strüwing
Queued to mysql-6.0-backup.
[26 Mar 2009 12:34] Bugs System
Pushed into 6.0.11-alpha (revid:alik@sun.com-20090326121822-pt84kzxxayzho4mn) (version source revid:rafal.somla@sun.com-20090302164601-znhm4tadplfi2iqu) (merge vers: 6.0.11-alpha) (pib:6)
[15 Apr 2009 1:40] Paul Dubois
Cosmetic change to silence warnings. No changelog entry needed.