Bug#33982 debug assertion and crash reloading grant tables after sighup or kill

In certain rare cases when a process was interrupted
during a FLUSH PRIVILEGES operation the diagnostic
area would be set to an error state but the function
responsible for the operation would still signal
success. This would lead to a debug assertion error
later on when the server would attempt to reset the
DA before sending the error message.

This patch fixes the issue by assuring that
reload_acl_and_cache() always fails if an error
condition is raised.

The second issue was that a KILL could cause
a console error message which referred to a DA
state without first making sure that such a
state existed.

This patch fixes this issue in two different
palces by first checking DA state before
fetching the error message.

 

1 files changed, 8 insertions, 2 deletions

diff --git a/sql/sql_servers.cc b/sql/sql_servers.cc
index f8a8dea18ff..1655426bb4a 100644
--- a/sql/sql_servers.cc
+++ b/sql/sql_servers.cc
@@ -241,8 +241,14 @@ bool servers_reload(THD *thd)
 
   if (simple_open_n_lock_tables(thd, tables))
   {
-    sql_print_error("Can't open and lock privilege tables: %s",
-		    thd->main_da.message());
+    /*
+      Execution might have been interrupted; only print the error message
+      if an error condition has been raised.
+    */
+    if (thd->main_da.is_error())
+      sql_print_error("Can't open and lock privilege tables: %s",
+                      thd->main_da.message());
+    return_val= FALSE;
     goto end;
   }