The problem is that safe_kill_win fails to detect a dead process. OpenProcess() will
succeed even after the process died, it will first fail after the last handle to process
is closed.
To fix the problem, check process status with GetExitCodeProcess() and consider
process to be dead if the exit code returned by this routine is not STILL_ALIVE.
Bug in Perl
Scrap attempt to do this smartly on AIX, just drop the test and assume it's OK
This commit undoes the previous push and adds a line to ignore on AIX
Based upon patch contributed by Stewart Smith
mysql-test/lib/My/SafeProcess/safe_process.cc:
Fix style -- remove unneeded spaces.
Specify C linkage for the signal handling functions.
Check return value from read()/write().
Suspected reason for the failure is that safe_process.exe already runs in a job that does not allow breakaways.
The fix is to use a fallback - make newly created process the root of the new process group. This allows to kill process together with descendants via GenerateConsoleCtrlEvent (CTRL_BREAK_EVENT, pid)
perl
The problem here was the method how MTR gets its unique thread ids.
Prior to this patch, the method to do it was to maintain a global
table of pid,mtr_unique_id) pairs. The table was backed by a text
file. The table was cleaned up one in a while and dead processes leaking
unique_ids were determined with with kill(0) or with scripting tasklist
on Windows.
This method is flawed specifically on native Windows Perl. fork() is
implemented with starting a new thread, give it a syntetic negative PID
(threadID*(-1)), until this thread creates a new process with exec()
However, neither tasklist nor any other native Windows tool can cope with
negative perl PIDs. This lead to incorrect determination of dead process
and reusing already used mtr_unique_id.
The patch introduces alternative portable method of solving unique-id
problem. When a process needs a unique id in range [min...max], it just
starts to open files named min, min+1,...max in a loop . After file is
opened, we do non-blocking flock(). When flock() succeeds, process has
allocated the ID. When process dies, file is unlocked . Checks for zombies
are not necessary.
Since the change would create a co-existence problems with older version
of MTR, because of different way to calculate IDs, the default ID range
is changed from 250-299 to 300-349.
Another fix that was necessary enable --parallel option was to serialize
spawn() calls on Windows. specifically, IO redirects needed to be protected.
This patch also fixes hanging CRTL-C (as described in Bug #38629) for the
"new" MTR. The fix was already in 6.0 and is now downported.
Error log gets truncated when mysqld is restarted by MTR
mysql-test/include/check-warnings.test:
Since server doesn't have log-error, get it from env. var. set by MTR
mysql-test/lib/My/ConfigFactory.pm:
"Hide" log-error in a comment
mysql-test/mysql-test-run.pl:
"Hide" log-error in my.cnf by comment
add --console to arguments on Windows
Move .err file to var/log
- output callstacks from crash using cdb debugger which is part
of "Debugging Tools for Windows". Output other interesting
information - function parameters, possibly source code fragment
and other goodies of "!analyze" cdb extension.
- Allow the new process to break away from any job that this
process is part of so that it can be assigned to the new JobObject
we just created. This is safe since the new JobObject is created with
the JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE flag, making sure it will be
terminated when the last handle to it is closed(which is owned by
this process).
- Add a "skip-ssl=1" to [mysqltest] section so that
mysqltest will not run with ssl turned on by default
but stil be able to turn it on when requested
- This avoids that check_warnings and check_testcase
connects to the server woth SSL turned on
- Additional patch with improved protection by putting it all inside an "eval"
- Calling 'hostpath' on a truncated socket may also croak.
- Remove the need to create any directory parts of "path" inside the function.
SIGABRT is sent to relevant processes after a timeout
client/mysqltest.cc:
Fixed signal handlers to mysqltest actually dumps core
mysql-test/lib/My/CoreDump.pm:
Added support for dbx
mysql-test/lib/My/SafeProcess.pm:
Added dump_core to force process to dump core
mysql-test/lib/My/SafeProcess/safe_process.cc:
Traps SIGABRT and sends this on to child
mysql-test/mysql-test-run.pl:
When test times out, force core dumps on mysqltest and servers
Problem: when mtr tries to create a directory, and the target
exists but is a file instead of directory, it tries several times
to create the directory again before it fails.
Fix: make it check if the target exists and is a non-directory.
mysql-test/lib/My/File/Path.pm:
mkpath() now stops with appropriate error message if the target
exists but is a non-directory.
mysql-test/lib/My/File/Path.pm:
Extend 'copytree' to take an optional "umask" parameter that will be used while copying the files
mysql-test/mysql-test-run.pl:
Pass umask 0022 to copytree so that the copied files will be created world readable and the mysqld
can LOAD DATA INFILE them
Post-push fixes making it work on pushbuild's valgrind host, and clarifying the output.
mysql-test/lib/My/CoreDump.pm:
- Improved parsing of mtr output so that it works on pushbuild's "valgrind" host.
- Added stack trace for the thread that coredumped, to make output more readable when there are many threads.
- Added explanation of what the output consists of.
- Added early removal of temp file.
