NULL MERGE this to 5.1

Apply the following InnoDB snapshots:
innodb-5.0-ss1319
innodb-5.0-ss1331
innodb-5.0-ss1333
innodb-5.0-ss1341

Fixes:
- Bug #21409: Incorrect result returned when in READ-COMMITTED with query_cache ON
  At low transaction isolation levels we let each consistent read set
  its own snapshot.
- Bug #23666: strange Innodb_row_lock_time_% values in show status; also millisecs wrong
  On Windows ut_usectime returns secs and usecs relative to the UNIX
  epoch (which is Jan, 1 1970).

- Bug #25494: LATEST DEADLOCK INFORMATION is not always cleared
  lock_deadlock_recursive(): When the search depth or length is exceeded,
  rewind lock_latest_err_file and display the two transactions at the
  point of aborting the search.

- Bug #25927: Foreign key with ON DELETE SET NULL on NOT NULL can crash server
  Prevent ALTER TABLE ... MODIFY ... NOT NULL on columns for which
  there is a foreign key constraint ON ... SET NULL.

- Bug #26835: Repeatable corruption of utf8-enabled tables inside InnoDB
  The bug could be reproduced as follows:

  Define a table so that the first column of the clustered index is
  a VARCHAR or a UTF-8 CHAR in a collation where sequences of bytes
  of differing length are considered equivalent.

  Insert and delete a record.  Before the delete-marked record is
  purged, insert another record whose first column is of different
  length but equivalent to the first record.  Under certain conditions,
  the insertion can be incorrectly performed as update-in-place.

  Likewise, an operation that could be done as update-in-place can
  unnecessarily be performed as delete and insert, but that would not
  cause corruption but merely degraded performance.

6 files changed, 117 insertions, 54 deletions

diff --git a/innobase/dict/dict0dict.c b/innobase/dict/dict0dict.c
index df48a8a4b5a..33aebb25071 100644
--- a/innobase/dict/dict0dict.c
+++ b/innobase/dict/dict0dict.c
@@ -2139,9 +2139,12 @@ dict_foreign_find_index(
 	ulint		n_cols,	/* in: number of columns */
 	dict_index_t*	types_idx, /* in: NULL or an index to whose types the
 				   column types must match */
-	ibool		check_charsets)	/* in: whether to check charsets.
+	ibool		check_charsets,	/* in: whether to check charsets.
 					only has an effect if types_idx !=
 					NULL. */
+	ulint		check_null)
+				/* in: nonzero if none of the columns must
+				be declared NOT NULL */
 {
 #ifndef UNIV_HOTBACKUP
 	dict_index_t*	index;
@@ -2154,10 +2157,11 @@ dict_foreign_find_index(
 		if (dict_index_get_n_fields(index) >= n_cols) {
 
 			for (i = 0; i < n_cols; i++) {
-				col_name = dict_index_get_nth_field(index, i)
-							->col->name;
-				if (dict_index_get_nth_field(index, i)
-						->prefix_len != 0) {
+				dict_field_t*	field
+					= dict_index_get_nth_field(index, i);
+
+				col_name = field->col->name;
+				if (field->prefix_len != 0) {
 					/* We do not accept column prefix
 					indexes here */
 					
@@ -2169,6 +2173,13 @@ dict_foreign_find_index(
 				  	break;
 				}
 
+				if (check_null
+				    && (field->col->type.prtype
+					& DATA_NOT_NULL)) {
+
+					return(NULL);
+				}
+
 				if (types_idx && !cmp_types_are_equal(
 				     dict_index_get_nth_type(index, i),
 				     dict_index_get_nth_type(types_idx, i),
@@ -2290,7 +2301,7 @@ dict_foreign_add_to_cache(
 		index = dict_foreign_find_index(ref_table,
 			(const char**) for_in_cache->referenced_col_names,
 			for_in_cache->n_fields,
-			for_in_cache->foreign_index, check_charsets);
+			for_in_cache->foreign_index, check_charsets, FALSE);
 
 		if (index == NULL) {
 			dict_foreign_error_report(ef, for_in_cache,
@@ -2317,13 +2328,17 @@ dict_foreign_add_to_cache(
 		index = dict_foreign_find_index(for_table,
 			(const char**) for_in_cache->foreign_col_names,
 			for_in_cache->n_fields,
-			for_in_cache->referenced_index, check_charsets);
+			for_in_cache->referenced_index, check_charsets,
+			for_in_cache->type
+			& (DICT_FOREIGN_ON_DELETE_SET_NULL
+			   | DICT_FOREIGN_ON_UPDATE_SET_NULL));
 
 		if (index == NULL) {
 			dict_foreign_error_report(ef, for_in_cache,
 "there is no index in the table which would contain\n"
 "the columns as the first columns, or the data types in the\n"
-"table do not match to the ones in the referenced table.");
+"table do not match to the ones in the referenced table\n"
+"or one of the ON ... SET NULL columns is declared NOT NULL.");
 
 			if (for_in_cache == foreign) {
 				if (added_to_referenced_list) {
@@ -3125,7 +3140,8 @@ col_loop1:
 	/* Try to find an index which contains the columns
 	as the first fields and in the right order */
 
-	index = dict_foreign_find_index(table, column_names, i, NULL, TRUE);
+	index = dict_foreign_find_index(table, column_names, i,
+					NULL, TRUE, FALSE);
 
 	if (!index) {
 		mutex_enter(&dict_foreign_err_mutex);
@@ -3390,7 +3406,7 @@ try_find_index:
 
 	if (referenced_table) {
 		index = dict_foreign_find_index(referenced_table,
-			column_names, i, foreign->foreign_index, TRUE);
+			column_names, i, foreign->foreign_index, TRUE, FALSE);
 		if (!index) {
 			dict_foreign_free(foreign);
 			mutex_enter(&dict_foreign_err_mutex);
diff --git a/innobase/include/rem0rec.ic b/innobase/include/rem0rec.ic
index 9c24f385f4f..1abbb503bab 100644
--- a/innobase/include/rem0rec.ic
+++ b/innobase/include/rem0rec.ic
@@ -982,6 +982,9 @@ rec_offs_nth_size(
 {
 	ut_ad(rec_offs_validate(NULL, NULL, offsets));
 	ut_ad(n < rec_offs_n_fields(offsets));
+	if (!n) {
+		return(rec_offs_base(offsets)[1 + n] & REC_OFFS_MASK);
+	}
 	return((rec_offs_base(offsets)[1 + n] - rec_offs_base(offsets)[n])
 			& REC_OFFS_MASK);
 }
diff --git a/innobase/include/sync0sync.ic b/innobase/include/sync0sync.ic
index a32a82d6e8b..e5c6f56d8ba 100644
--- a/innobase/include/sync0sync.ic
+++ b/innobase/include/sync0sync.ic
@@ -6,6 +6,16 @@ Mutex, the basic synchronization primitive
 Created 9/5/1995 Heikki Tuuri
 *******************************************************/
 
+#if defined(not_defined) && defined(__GNUC__) && defined(UNIV_INTEL_X86)
+/* %z0: Use the size of operand %0 which in our case is *m to determine
+instruction size, it should end up as xchgl. "1" in the input constraint,
+says that "in" has to go in the same place as "out".*/
+#define TAS(m, in, out) \
+	asm volatile ("xchg%z0 %2, %0" \
+	: "=g" (*(m)), "=r" (out) \
+	: "1" (in))	/* Note: "1" here refers to "=r" (out) */
+#endif
+
 /**********************************************************************
 Sets the waiters field in a mutex. */
 
@@ -85,20 +95,10 @@ mutex_test_and_set(
 
 	return(res);
 #elif defined(not_defined) && defined(__GNUC__) && defined(UNIV_INTEL_X86)
-	ulint*	lw;
 	ulint	res;
 
-	lw = &(mutex->lock_word);
-
-	/* In assembly we use the so-called AT & T syntax where
-	the order of operands is inverted compared to the ordinary Intel
-	syntax. The 'l' after the mnemonics denotes a 32-bit operation.
-	The line after the code tells which values come out of the asm
-	code, and the second line tells the input to the asm code. */
+	TAS(&mutex->lock_word, 1, res);
 
-	asm volatile("movl $1, %%eax; xchgl (%%ecx), %%eax" :
-	              "=eax" (res), "=m" (*lw) :
-	              "ecx" (lw));
 	return(res);
 #else
 	ibool	ret;
@@ -137,20 +137,9 @@ mutex_reset_lock_word(
         __asm   MOV     ECX, lw
         __asm   XCHG    EDX, DWORD PTR [ECX]                    
 #elif defined(not_defined) && defined(__GNUC__) && defined(UNIV_INTEL_X86)
-	ulint*	lw;
-
-	lw = &(mutex->lock_word);
-
-	/* In assembly we use the so-called AT & T syntax where
-	the order of operands is inverted compared to the ordinary Intel
-	syntax. The 'l' after the mnemonics denotes a 32-bit operation. */
+	ulint	res;
 
-	asm volatile("movl $0, %%eax; xchgl (%%ecx), %%eax" :
-	              "=m" (*lw) :
-	              "ecx" (lw) :
-		      "eax");	/* gcc does not seem to understand
-				that our asm code resets eax: tell it
-				explicitly that after the third ':' */
+	TAS(&mutex->lock_word, 0, res);
 #else
 	mutex->lock_word = 0;
 
diff --git a/innobase/lock/lock0lock.c b/innobase/lock/lock0lock.c
index 06475c8ef7e..77dfca5fdf4 100644
--- a/innobase/lock/lock0lock.c
+++ b/innobase/lock/lock0lock.c
@@ -3259,12 +3259,6 @@ lock_deadlock_recursive(
 
 	*cost = *cost + 1;
 
-	if ((depth > LOCK_MAX_DEPTH_IN_DEADLOCK_CHECK)
-		|| (*cost > LOCK_MAX_N_STEPS_IN_DEADLOCK_CHECK)) {
-
-		return(LOCK_VICTIM_IS_START);
-	}
-
 	lock = wait_lock;
 
 	if (lock_get_type(wait_lock) == LOCK_REC) {
@@ -3296,11 +3290,18 @@ lock_deadlock_recursive(
 
 		if (lock_has_to_wait(wait_lock, lock)) {
 
+			ibool	too_far
+				= depth > LOCK_MAX_DEPTH_IN_DEADLOCK_CHECK
+				|| *cost > LOCK_MAX_N_STEPS_IN_DEADLOCK_CHECK;
+
 			lock_trx = lock->trx;
 
-			if (lock_trx == start) {
+			if (lock_trx == start || too_far) {
+
 				/* We came back to the recursion starting
-				point: a deadlock detected */
+				point: a deadlock detected; or we have
+				searched the waits-for graph too long */
+
 				FILE*	ef = lock_latest_err_file;
 				
 				rewind(ef);
@@ -3342,9 +3343,20 @@ lock_deadlock_recursive(
 				}
 #ifdef UNIV_DEBUG
 				if (lock_print_waits) {
-					fputs("Deadlock detected\n", stderr);
+					fputs("Deadlock detected"
+					      " or too long search\n",
+					      stderr);
 				}
 #endif /* UNIV_DEBUG */
+				if (too_far) {
+
+					fputs("TOO DEEP OR LONG SEARCH"
+					      " IN THE LOCK TABLE"
+					      " WAITS-FOR GRAPH\n", ef);
+
+					return(LOCK_VICTIM_IS_START);
+				}
+
 				if (ut_dulint_cmp(wait_lock->trx->undo_no,
 							start->undo_no) >= 0) {
 					/* Our recursion starting point
diff --git a/innobase/srv/srv0srv.c b/innobase/srv/srv0srv.c
index fe9e08d65be..96c0f05111b 100644
--- a/innobase/srv/srv0srv.c
+++ b/innobase/srv/srv0srv.c
@@ -1822,14 +1822,14 @@ srv_export_innodb_status(void)
         export_vars.innodb_pages_written= buf_pool->n_pages_written;
         export_vars.innodb_row_lock_waits= srv_n_lock_wait_count;
         export_vars.innodb_row_lock_current_waits= srv_n_lock_wait_current_count;
-        export_vars.innodb_row_lock_time= srv_n_lock_wait_time / 10000;
+        export_vars.innodb_row_lock_time= srv_n_lock_wait_time / 1000;
 	if (srv_n_lock_wait_count > 0) {
 		export_vars.innodb_row_lock_time_avg = (ulint)
-			(srv_n_lock_wait_time / 10000 / srv_n_lock_wait_count);
+			(srv_n_lock_wait_time / 1000 / srv_n_lock_wait_count);
 	} else {
 		export_vars.innodb_row_lock_time_avg = 0;
 	}
-        export_vars.innodb_row_lock_time_max= srv_n_lock_max_wait_time / 10000;
+        export_vars.innodb_row_lock_time_max= srv_n_lock_max_wait_time / 1000;
         export_vars.innodb_rows_read= srv_n_rows_read;
         export_vars.innodb_rows_inserted= srv_n_rows_inserted;
         export_vars.innodb_rows_updated= srv_n_rows_updated;
diff --git a/innobase/ut/ut0ut.c b/innobase/ut/ut0ut.c
index 1be5939303a..feb03269d91 100644
--- a/innobase/ut/ut0ut.c
+++ b/innobase/ut/ut0ut.c
@@ -20,6 +20,55 @@ Created 5/11/1994 Heikki Tuuri
 
 ibool	ut_always_false	= FALSE;
 
+#ifdef __WIN__
+/*********************************************************************
+NOTE: The Windows epoch starts from 1601/01/01 whereas the Unix
+epoch starts from 1970/1/1. For selection of constant see:
+http://support.microsoft.com/kb/167296/ */
+#define WIN_TO_UNIX_DELTA_USEC  ((ib_longlong) 11644473600000000ULL)
+
+
+/*********************************************************************
+This is the Windows version of gettimeofday(2).*/
+static
+int
+ut_gettimeofday(
+/*============*/
+				/* out: 0 if all OK else -1 */
+	struct timeval*	tv,	/* out: Values are relative to Unix epoch */
+	void*		tz)	/* in: not used */
+{
+	FILETIME	ft;
+	ib_longlong	tm;
+
+	if (!tv) {
+		errno = EINVAL;
+		return(-1);
+	}
+
+	GetSystemTimeAsFileTime(&ft);
+
+	tm = (ib_longlong) ft.dwHighDateTime << 32;
+	tm |= ft.dwLowDateTime;
+
+	ut_a(tm >= 0);	/* If tm wraps over to negative, the quotient / 10
+			does not work */
+
+	tm /= 10;	/* Convert from 100 nsec periods to usec */
+
+	/* If we don't convert to the Unix epoch the value for
+	struct timeval::tv_sec will overflow.*/
+	tm -= WIN_TO_UNIX_DELTA_USEC;
+
+	tv->tv_sec  = (long) (tm / 1000000L);
+	tv->tv_usec = (long) (tm % 1000000L);
+
+	return(0);
+}
+#else
+#define	ut_gettimeofday		gettimeofday
+#endif
+
 /*********************************************************************
 Get the quote character to be used in SQL identifiers.
 This definition must match the one in sql/ha_innodb.cc! */
@@ -82,17 +131,11 @@ ut_usectime(
 	ulint*	sec,	/* out: seconds since the Epoch */
 	ulint*	ms)	/* out: microseconds since the Epoch+*sec */
 {
-#ifdef __WIN__
-	SYSTEMTIME st;
-	GetLocalTime(&st);
-	*sec = (ulint) st.wSecond;
-	*ms  = (ulint) st.wMilliseconds;
-#else
 	struct timeval	tv;
-	gettimeofday(&tv,NULL);
+
+	ut_gettimeofday(&tv, NULL);
 	*sec = (ulint) tv.tv_sec;
 	*ms  = (ulint) tv.tv_usec;
-#endif
 }
 
 /**************************************************************