This is a replacement for the original patch given by Ingo.  This one comes
from Monty.
The problem is that merge files now use unix style pathnames on all 
platforms.  The merge file open code was not properly converting those
pathnames back to non-unix when necessary.


myisammrg/myrg_open.c:
  use fn_format to convert pathnames if the files in the merge file have a pathname
mysys/my_getwd.c:
  add a test for '/' in the code that determines if a string inlucdes a pathname

This patch was submitted by Ingo and it appears to work correctly.


sql/ha_myisammrg.cc:
  use FN_LIBCHAR instead of / so buff works correctly on Windows
strings/my_vsnprintf.c:
  add support for %c to my_vsnprintf

sql_udf.cc:
  changed dl_name to udf->dl in mysql_create_function


sql/sql_udf.cc:
  changed dl_name to udf->dl in mysql_create_function

into mysql.com:/usr/home/ram/work/mysql-4.1

sql/spatial.cc:
  just float8get doesn't work well with the local variables -                     
  they can be of processor-specific floating type - not the                       
  standard decimal (by hf)

into serg.mylan:/usr/home/serg/Abk/mysql-4.1

sql/unireg.cc:
  Auto merged

into book.sanja.is.com.ua:/Users/bell/mysql/bk/work-4.1

into serg.mylan:/usr/home/serg/Abk/mysql-4.0

when reporting an error, use an appropriate buffer for udf->name

into mysql.com:/home/mydev/mysql-4.1-4100

After review version.
Added a condition for MERGE tables. These do not have unique
indexes. But every key could be a unique key on the underlying
MyISAM table. So get the maximum key length for MERGE tables
instead of the maximum unique key length. This is used for
buffer allocation in write_record().


mysql-test/r/merge.result:
  Bug#10400 - Improperly-defined MERGE table crashes with INSERT ... ON DUPLICATE KEY UPDATE
  The test result.
mysql-test/t/merge.test:
  Bug#10400 - Improperly-defined MERGE table crashes with INSERT ... ON DUPLICATE KEY UPDATE
  The test case does not in all cases show the problem
  without the bugfix. The improper memory allocation
  might get through undetected in many cases.
sql/ha_myisammrg.h:
  Bug#10400 - Improperly-defined MERGE table crashes with INSERT ... ON DUPLICATE KEY UPDATE
  Set a table_flag for the MERGE engine telling that any index
  might be unique without being specified as such.
sql/handler.h:
  Bug#10400 - Improperly-defined MERGE table crashes with INSERT ... ON DUPLICATE KEY UPDATE
  Added a new table_flag telling that any index
  might be unique without being specified as such.
sql/sql_insert.cc:
  Bug#10400 - Improperly-defined MERGE table crashes with INSERT ... ON DUPLICATE KEY UPDATE
  Changed the freeing of the memory to be symmetric to its
  allocation (my_safe_alloc -> my_safe_afree).
  This is not directly related to the bug.
sql/table.cc:
  Bug#10400 - Improperly-defined MERGE table crashes with INSERT ... ON DUPLICATE KEY UPDATE
  Added a condition for MERGE tables. These do not have unique
  indexes. But every key could be a unique key on the underlying
  MyISAM table. So get the maximum key length for MERGE tables
  instead of the maximum unique key length. This is used for
  buffer allocation in write_record().

into mysql.com:/home/jimw/my/mysql-4.1-clean

mysql-test/r/mysqldump.result:
  Update results
mysql-test/t/mysqldump.test:
  Fix test to not include version number in results

client/mysqldump.c:
  Merge bug fix
mysql-test/r/mysqldump.result:
  Update results
mysql-test/t/mysqldump.test:
  Merge bug fix

client/mysqldump.c:
  Optimize away a call to strend() by using return value of my_snprintf()

into mysql.com:/space/pekka/ndb/version/my41

ndb/src/common/util/version.c:
  allow upgrade 4.1.10->4.1.12

into book.sanja.is.com.ua:/Users/bell/mysql/bk/work-4.1

BitKeeper/deleted/.del-grant_cache-master.opt~1b04b4b1116a7a2:
  Delete: mysql-test/t/grant_cache-master.opt
mysql-test/r/grant_cache.result:
  query cache management moved inside test
mysql-test/t/grant_cache.test:
  query cache management moved inside test to privent server reloading during test

  Added test cases for bug #7914.
sql_select.cc:
  Fixed bug #7914: rollup over expresssions such as sum(a)+1.


sql/sql_select.cc:
  Fixed bug #7914: rollup over expresssions such as sum(a)+1.
mysql-test/t/olap.test:
  Added test cases for bug #7914.
mysql-test/r/olap.result:
  Added test cases for bug #7914.

  Added missing ';'.


sql/sql_yacc.yy:
  Added missing ';'.

into serg.mylan:/usr/home/serg/Abk/mysql-4.1


mysql-test/t/range.test:
  Auto merged

mysql-test/mysql-test-run.sh:
  always use --no-defaults
mysql-test/t/range.test:
  range.test does not requires innodb

  updated error message for ndb


ndb/src/ndbapi/ndberror.c:
  updated error message for ndb
BitKeeper/etc/logging_ok:
  Logging to logging@openlogging.org accepted

into mysql.com:/data0/mysqldev/my/mysql-4.1-build


configure.in:
  Auto merged
mysql-test/t/range.test:
  Auto merged
sql/item_strfunc.h:
  Auto merged

into book.sanja.is.com.ua:/Users/bell/mysql/bk/work-4.1

into mysql.com:/home/psergey/mysql-4.1-bug10095

client/mysql.cc:
  Remove usage of c_ptr_safe() as this causes a linkage problem when compiling MySQL without inline functions
sql/sql_base.cc:
  Don't use c_ptr_safe() on this string as this causes a realloc and the String object (allocated by sql_yacc.yy) is never freed

into mysql.com:/home/psergey/mysql-4.1-bug10221

mysql-test/r/subselect.result:
  test fro bug 10020
mysql-test/t/subselect.test:
  bug 10020 test suite
sql/item.cc:
  any constant allowed here BUG#10020)
BitKeeper/etc/logging_ok:
  Logging to logging@openlogging.org accepted

into mysql.com:/Users/kent/mysql/bk/mysql-4.1

  To solve bug#8455, call mysql_affected_rows() earlier, before query to find warnings


client/mysqltest.c:
  To solve bug#8455, call mysql_affected_rows() earlier, before query to find warnings

into mysql.com:/home/psergey/mysql-4.0-bug10095

The problem: base_list::remove didn't modify base_list::last when removing 
the last list element.
The fix: If we remove the last element, find the element before it (by walking
from the beginning of the list) and set base_list::last accordingly.

The list gets corrupted in both 4.0 and 4.1. There are no visible problems in 
current 4.1 because current 4.1 doesn't call where_cond->fix_fields() after 
constant propagation step.


mysql-test/r/select.result:
  Testcase for BUG#10095
mysql-test/t/select.test:
  Testcase for BUG#10095
sql/sql_list.h:
  Fix for BUG#10095: {wrong query results because of incorrect constant propagation} 
  The problem: base_list::remove didn't modify base_list::last when removing 
  the last list element.
  The fix: If we remove the last element, find the element before it (by walking
  from the beginning of the list) and set base_list::last accordingly.

  To solve bug#6193, also reset require when using 'eval' for running query
mysql-test-run.pl:
  In cleanup, remove possible soft link to var directory


mysql-test/mysql-test-run.pl:
  In cleanup, remove possible soft link to var directory
client/mysqltest.c:
  To solve bug#6193, also reset require when using 'eval' for running query

  Added option --vardir=<directory> for feature bug#10096


mysql-test/mysql-test-run.pl:
  Added option --vardir=<directory> for feature bug#10096
mysql-test/lib/mtr_process.pl:
  Added option --vardir=<directory> for feature bug#10096
mysql-test/lib/mtr_report.pl:
  Added option --vardir=<directory> for feature bug#10096

  Not to conflict with Windows return code 2 for "not found", use 62 for "skip"
init_db.sql:
  Initial db updates


mysql-test/lib/init_db.sql:
  Initial db updates
mysql-test/mysql-test-run.pl:
  Not to conflict with Windows return code 2 for "not found", use 62 for "skip"