Reason for fatal errors is sometimes unclear

[Philip Stoev]

The reason for RQG errors is sometimes unclear. The actual circumstance that causes a fatal error can happen somewhat earlier and thus be buried in other log output.

For example:

# 2011-11-16T09:59:02 Query: SHOW PROCEDURE CODE h failed: 1289 The 'SHOW PROCEDURE|FUNCTION CODE' feature is disabled; you need MySQL built with '--with-debug' to have it working
# 2011-11-16T09:59:02 Loading Validator GenTest::Validator::ErrorMessageCorruption.
# 2011-11-16T09:59:02 Query: SHOW CREATE VIEW u failed: 1347 'test.u' is not VIEW. Further errors of this kind will be suppressed.
# 2011-11-16T09:59:02 Killing child process with pid 7353...
# 2011-11-16T09:59:02 Killing child process with pid 7359...
# 2011-11-16T09:59:02 Killing child process with pid 7357...
# 2011-11-16T09:59:02 Killing child process with pid 7355...
# 2011-11-16T09:59:02 Killing child process with pid 7351...
# 2011-11-16T09:59:02 Killing child process with pid 7347...
# 2011-11-16T09:59:02 Killing child process with pid 7349...
# 2011-11-16T09:59:02 Killing child process with pid 7345...
# 2011-11-16T09:59:02 Killing child process with pid 7341...
# 2011-11-16T09:59:02 Started periodic reporting process...
# 2011-11-16T09:59:03 Killing periodic reporting process with pid 7340...
# 2011-11-16T09:59:03 Kill GenTest::ErrorFilter(7339)
# 2011-11-16T09:59:03 Test completed with failure status STATUS_ENVIRONMENT_FAILURE (110)

Distance between STATUS_ENVIRONMENT_FAILURE and the actual reason "feature is disabled; you need MySQL built with '--with-debug' to have it working" is too big and it is not apparent that "feature is disabled; you need MySQL built with '--with-debug' to have it working" is a fatal error.

The solution is to label fatal errors more clearly at the time they occur and shorten the distance between them and the actual end of the test/log.