Problem: mysqlbinlog does not free memory if an error happens.
Fix: binlog-processing functions do not call exit() anymore. Instead, they
print an error and return an error code. Error codes are propagated all
the way back to main, and all allocated memory is freed on the way.
client/mysqlbinlog.cc:
- New error handling policy: functions processing binlogs don't just
exit() anymore. Instead, they print a message and return an error
status.
- New policy for the global `mysql' and `glob_description_event': these
are not passed as parameters anymore. The global pointer is used
instead.
- More error situations are detected and reported.
- Better error messages: the program never terminates with exit status 1
without explanation any more. Fixed spelling errors. Use consistent
format of messages (a single line beginning with "ERROR: " or
"WARNING: " and ending with "." is printed to stderr.)
- New memory handling: memory is always freed on program termination.
- Better comments: more functions are explained, doxygen is used, and
more precise formulations in some existing comments.
mysql-test/suite/binlog/r/binlog_base64_flag.result:
Result file updated since output format of mysqlbinlog changed while the
test was disabled.
mysql-test/suite/binlog/t/binlog_killed.test:
Mysqlbinlog now works as described when the binlog is open. Hence, the
--force-if-open flag must be passed
mysql-test/suite/binlog/t/binlog_killed_simulate.test:
Mysqlbinlog now works as described when the binlog is open. Hence, the
--force-if-open flag must be passed
mysql-test/suite/binlog/t/disabled.def:
Now that mysqlbinlog cleans up after itself on abnormal termination, we
can enable this test again.
refining tests as they appear to be non-deterministic.
mysql-test/suite/binlog/r/binlog_killed.result:
results changed
mysql-test/suite/binlog/t/binlog_killed.test:
restructuring the test to deliver more deterministic outcome.
order-by for update and delete did not guaratee the order of scanning.
--error are masked with zero in order to catch more info if killing
really will happen to be non-deterministic on some platform (witnessed
something like that on dl145j)
mysql-test/suite/binlog/t/binlog_killed_simulate.test:
important guard added
destination. Removing wrong (local temp) tests.
mysql-test/suite/binlog/t/binlog_killed_simulate-master.opt:
Rename: mysql-test/t/binlog_killed_simulate-master.opt -> mysql-test/suite/binlog/t/binlog_killed_simulate-master.opt
BitKeeper/deleted/.del-binlog_killed_bug27571.test:
Delete: mysql-test/t/binlog_killed_bug27571.test
BitKeeper/deleted/.del-binlog_killed_bug27571-master.opt:
Delete: mysql-test/t/binlog_killed_bug27571-master.opt
mysql-test/suite/binlog/r/binlog_killed.result:
results changed
mysql-test/suite/binlog/t/binlog_killed.test:
binlog offset made 5.1 specific
mysql-test/suite/binlog/t/binlog_killed_simulate.test:
binlog offset + guard to run the test with stmt/mixed format
mysql-test/suite/binlog/r/binlog_killed_simulate.result:
new results file
2007-10-31 11:48:49 +02:00
Renamed from mysql-test/t/binlog_killed_simulate.test (Browse further)
