Bug #31745 mysqld crash handler does not work on windows

Description:
While working on BUG#18441 I found that each thread in mysqld will install an empty signal handler. This has the effect that a SIGSEGV in for example 'mysql_update' will more or less be ignored. It will close the connection to the client causing the segfault, but when returning from the signal handler the thread will start to execute 'main' thus starting a new mysqld(in the signal handler thread).

Have rewritten the signal handling code for windows in order to avoid this problem  - see the attached patch.

All threads will have a "handle_win_signal" installed as signal handler for all the signals to catch. That function will:
- in case of SIGINT do a "ordered shutdown" by signaling the shutdown thread which will 'kill_server' 
- in all other case it will call the 'handle_segfault' function (which need to be moved out of the #ifdef that exposes it only on some systems). This will give us the same behavior as on Linux ie. an error log file will be generated indicating which signal was caught, the time it occurred, special settings used and so on.

How to repeat:
Add code that uses a NULL pointer to 'mysql_update'

$ bk diffs -u
===== sql_update.cc 1.222 vs edited =====
--- 1.222/sql/sql_update.cc     2007-07-31 13:42:53 +02:00
+++ edited/sql_update.cc        2007-10-22 12:13:26 +02:00
@@ -134,6 +134,10 @@
   SELECT_LEX    *select_lex= &thd->lex->select_lex;
   bool need_reopen;
   List<Item> all_fields;
+
+  char* p= 0;
+  *p= 1;
+
   DBUG_ENTER("mysql_update");

Then run a SQL query like "UPDATE t1 set a=a" and watch the behaviour.
 
   LINT_INIT(timestamp_query_id);

Suggested fix:
All threads will have a "handle_win_signal" installed as signal handler for all the signals to catch. That function will:
- in case of SIGINT do a "ordered shutdown" by signaling the shutdown thread which will 'kill_server' 
- in all other case it will call the 'handle_segfault' function (which need to be moved out of the #ifdef that exposes it only on some systems). This will give us the same behavior as on Linux ie. an error log file will be generated indicating which signal was caught, the time it occurred, special settings used and so on.

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/36036

ChangeSet@1.2530, 2007-10-22 15:10:06+02:00, msvensson@shellback.(none) +3 -0
  Bug#31745 mysqld crash handler does not work on windows
  - Rework signal handler code to either shutdown using the shutdown
    thread or call 'handle_segfault' to print out crash info and then
    crash the mysqld

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/42471

ChangeSet@1.2568, 2008-02-18 10:51:24+01:00, vvaintroub@wva. +5 -0
  Bug#31745 - crash handler does not work on Windows
  - Replace per-thread signal()'s with  SetUnhandledExceptionFilter(). 
    The only remaining signal() is for SIGABRT (default abort()
    handler in VS2005 is broken, i.e removes user exception filter)
  - remove MessageBox()'es  from error handling code
  - Windows port for print_stacktrace() and write_core()

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/42540

ChangeSet@1.2568, 2008-02-19 12:37:39+01:00, vvaintroub@wva. +5 -0
  Bug#31745 - crash handler does not work on Windows
  - Replace per-thread signal()'s with  SetUnhandledExceptionFilter(). 
    The only remaining signal() is for SIGABRT (default abort()
    handler in VS2005 is broken, i.e removes user exception filter)
  - remove MessageBox()'es  from error handling code
  - Windows port for print_stacktrace() and write_core() 
  - Cleanup, removed some unused functions

approved by email

Pushed into 5.1.24-rc

Pushed into 5.0.58

Pushed into 6.0.5-alpha

Noted in 5.0.58, 5.1.24, 6.0.5 changelogs.

The mysqld crash handler failed on Windows.

Marked Bug#29385 as a duplicate.