[svn-r16136] Several changes:

   Modified H5C2_journal_post_flush() write the super block and flush
the file before truncating the journal.  Failure to do this opened
a window in which the application could crash leaving the HDF5 file
in a state that was un-recoverable.

   The hope is that this will fix the file recovery bug observed on 
RSQ -- but I have not been able to test there.  However, I was able 
to generate a similar bug on Linew, and this fix seems to deal with 
the Linew bug.

   Added a third test to the h5recovery tests.  This is really
a test for the library, but it was easier to use existing test 
code there to construct the new test.  

   The new test runs the same application repeatedly, but setting a 
timer to crash the application at progressively later times.  The object
is to search for windows in which the application leaves the HDF5 file
in an un-recoverable state.

   Also, updated H5recover.c to use HDstrtoll() instead of HDstrtod()
to read some addresses and such from the journal file.

   Tested serial (debug) on Phoenix and Linew, and parallel (debug)
on Jam.

1 files changed, 30 insertions, 5 deletions

diff --git a/src/H5C2journal.c b/src/H5C2journal.c
index b8ec171..ab79a7b 100644
--- a/src/H5C2journal.c
+++ b/src/H5C2journal.c
@@ -184,6 +184,7 @@ H5C2_begin_journaling(H5F_t * f,
 		  FUNC);
 #endif /* JRM */ 
 
+    /* Note that this call flushes the HDF5 file in passing */
     result = H5C2_mark_journaling_in_progress(f, dxpl_id, 
 		                              config_ptr->journal_file_path);
 
@@ -305,7 +306,8 @@ done:
  *                 journal file.
  *
  *              2) Mark the superblock to indicate that we are no longer
- *                 journaling.
+ *                 journaling.  Note that this will flush the HDF5 file 
+ *                 again in passing.
  *
  *              3) Tell the journal file write code to shutdown.  This will
  *                 also cause the journal file to be deleted.
@@ -631,9 +633,11 @@ done:
  *
  * 		3) If the cache_is_clean parameter is true:
  *
- * 		   a) Truncate the journal file
+ * 		   a) Flush the HDF5 file
  *
- * 		   b) Reset cache_ptr->trans_num and 
+ * 		   b) Truncate the journal file
+ *
+ * 		   c) Reset cache_ptr->trans_num and 
  * 		      cache_ptr->last_trans_on_disk to zero.
  * 		
  * Return:      Success:        SUCCEED
@@ -646,7 +650,9 @@ done:
  */
 
 herr_t
-H5C2_journal_post_flush(H5C2_t * cache_ptr,
+H5C2_journal_post_flush(const H5F_t * f,
+                        hid_t dxpl_id,
+                        H5C2_t * cache_ptr,
 		        hbool_t cache_is_clean)
 {
     herr_t result;
@@ -654,6 +660,7 @@ H5C2_journal_post_flush(H5C2_t * cache_ptr,
 
     FUNC_ENTER_NOAPI(H5C2_journal_post_flush, FAIL)
 
+    HDassert( f != NULL );
     HDassert( cache_ptr != NULL );
     HDassert( cache_ptr->magic == H5C2__H5C2_T_MAGIC );
     HDassert( cache_ptr->mdj_enabled );
@@ -672,6 +679,24 @@ H5C2_journal_post_flush(H5C2_t * cache_ptr,
 
     if ( cache_is_clean ) {
 
+        /* Write the superblock to disk */
+
+        result = H5F_super_write(f, dxpl_id);
+
+        if ( result != SUCCEED ) {
+
+            HGOTO_ERROR(H5E_CACHE, H5E_WRITEERROR, FAIL, \
+                        "unable to write superblock to file")
+        }
+
+	result = H5FD_flush(f->shared->lf, dxpl_id, (unsigned)0);
+
+	if ( result > 0 ) {
+
+	    HGOTO_ERROR(H5E_IO, H5E_WRITEERROR, FAIL, \
+		        "low level flush failed")
+	}
+
         result = H5C2_jb__trunc(&(cache_ptr->mdj_jbrb));
 
         if ( result != SUCCEED ) {
@@ -705,7 +730,7 @@ done:
  *
  * 		3) Get the ID of the last transaction on disk.
  *
- * 		4) If the value obtained in 2) above has changed,
+ * 		4) If the value obtained in 3) above has changed,
  * 		   remove all entries whose last transaction has 
  * 		   made it to disk from the journal write in progress
  * 		   list.