[Bug 674862] Gnucash crashes after creating a new SX using the Mortgage

Wizard and SQL Backend

This addresses the crash reported on Ubuntu with Postgres by Krzysiek.
The stack trace he posted showed the crash was the result of
dereferencing a NULL GDate*, and this change protects against that. It
still doesn't address the more basic problem of why on Win32 and with
mysql and pgsql (but not SQLite3) the mortgage wizard is writing corrupt
dates.

[BP]

git-svn-id: svn+ssh://svn.gnucash.org/repo/gnucash/trunk@22171 57a11ea4-9604-0410-9ed3-97b8803252fd

src/engine/SchedXaction.c

@@ -577,12 +577,23 @@ xaccSchedXactionSetName( SchedXaction *sx, const gchar *newName )
 const GDate*
 xaccSchedXactionGetStartDate(const SchedXaction *sx )
 {
+    g_assert (sx);
     return &sx->start_date;
 }
 
 void
 xaccSchedXactionSetStartDate( SchedXaction *sx, const GDate* newStart )
 {
+    if ( newStart == NULL || !g_date_valid( newStart ))
+    {
+        /* XXX: I reject the bad data - is this the right
+         * thing to do <rgmerk>.
+         * This warning is only human readable - the caller
+         * doesn't know the call failed.  This is bad
+         */
+        g_critical("Invalid Start Date");
+        return;
+    }
     gnc_sx_begin_edit(sx);
     sx->start_date = *newStart;
     qof_instance_set_dirty(&sx->inst);
@@ -592,27 +603,29 @@ xaccSchedXactionSetStartDate( SchedXaction *sx, const GDate* newStart )
 gboolean
 xaccSchedXactionHasEndDate( const SchedXaction *sx )
 {
-    return g_date_valid( &sx->end_date );
+    return sx != NULL && g_date_valid( &sx->end_date );
 }
 
 const GDate*
 xaccSchedXactionGetEndDate(const SchedXaction *sx )
 {
+    g_assert (sx);
     return &sx->end_date;
 }
 
 void
 xaccSchedXactionSetEndDate( SchedXaction *sx, const GDate *newEnd )
 {
-    if ( g_date_valid( newEnd )
-            && g_date_compare( newEnd, &sx->start_date ) < 0 )
+    if (newEnd == NULL
+	|| !g_date_valid( newEnd )
+	|| g_date_compare( newEnd, &sx->start_date ) < 0 )
     {
         /* XXX: I reject the bad data - is this the right
          * thing to do <rgmerk>.
          * This warning is only human readable - the caller
          * doesn't know the call failed.  This is bad
          */
-        g_critical("New end date before start date");
+        g_critical("Bad End Date: Invalid or before Start Date");
         return;
     }