Changeset a0b94685 in rtems

Timestamp:

May 14, 1999, 2:43:53 PM (22 years ago)

Author:

Joel Sherrill <joel.sherrill@…>

Branches:

4.10, 4.11, 4.8, 4.9, 5, master

Children:

2bd9383

Parents:

123d4b7d

Message:

Removed check for initialized ID at request of Rosimildo DaSilva?
<rdasilva@…> who encountered random failures in his
port of omniORB2.

Files:

• c/src/exec/posix/src/mutex.c (modified) (2 diffs)
• cpukit/posix/src/mutex.c (modified) (2 diffs)

c/src/exec/posix/src/mutex.c

@@ -235 +235 @@
     return EINVAL;
 
+  /*
+   *  This code should eventually be removed.  
+   *
+   *  Although the POSIX specification says:
+   *
+   *  "Attempting to initialize an already initialized mutex results
+   *  in undefined behavior."
+   *
+   *  Trying to keep the caller from doing the create when *mutex
+   *  is actually a valid ID causes grief.  All it takes is the wrong
+   *  value in an uninitialized variable to make this fail.  As best
+   *  I can tell, RTEMS was the only pthread implementation to choose
+   *  this option for "undefined behavior" and doing so has created
+   *  portability problems.  In particular, Rosimildo DaSilva
+   *  <rdasilva@connecttel.com> saw seemingly random failures in the
+   *  RTEMS port of omniORB2 when this code was enabled.
+   *
+   *  Joel Sherrill <joel@OARcorp.com>     14 May 1999
+   */
+
+  
+#if 0
   /* avoid infinite recursion on call to this routine in _POSIX_Mutex_Get */
 
@@ -251 +273 @@
     }
   }
+#endif
  
   if ( !the_attr->is_initialized ) 

cpukit/posix/src/mutex.c

@@ -235 +235 @@
     return EINVAL;
 
+  /*
+   *  This code should eventually be removed.  
+   *
+   *  Although the POSIX specification says:
+   *
+   *  "Attempting to initialize an already initialized mutex results
+   *  in undefined behavior."
+   *
+   *  Trying to keep the caller from doing the create when *mutex
+   *  is actually a valid ID causes grief.  All it takes is the wrong
+   *  value in an uninitialized variable to make this fail.  As best
+   *  I can tell, RTEMS was the only pthread implementation to choose
+   *  this option for "undefined behavior" and doing so has created
+   *  portability problems.  In particular, Rosimildo DaSilva
+   *  <rdasilva@connecttel.com> saw seemingly random failures in the
+   *  RTEMS port of omniORB2 when this code was enabled.
+   *
+   *  Joel Sherrill <joel@OARcorp.com>     14 May 1999
+   */
+
+  
+#if 0
   /* avoid infinite recursion on call to this routine in _POSIX_Mutex_Get */
 
@@ -251 +273 @@
     }
   }
+#endif
  
   if ( !the_attr->is_initialized )