[svn-r21912] Issue 7756 - Creating a dataset in a read-only file caused seg fault when the file is closed. I fixed the problem by putting a condition check early in H5O_create of H5O.c. The old code checked it too late, not until a file space is created. I added a test case in tfile.c to check the creation of group, dataset, attribute, and datatype.

Tested on jam - I've tested the same changes in the trunk with h5committest.

1 files changed, 3 insertions, 1 deletions

diff --git a/src/H5O.c b/src/H5O.c
index bc7a913..f113001 100644
--- a/src/H5O.c
+++ b/src/H5O.c
@@ -1121,10 +1121,12 @@ H5O_create(H5F_t *f, hid_t dxpl_id, size_t size_hint, size_t initial_rc,
 
     /* check args */
     HDassert(f);
-    HDassert(H5F_INTENT(f) & H5F_ACC_RDWR);
     HDassert(loc);
     HDassert(TRUE == H5P_isa_class(ocpl_id, H5P_OBJECT_CREATE));
 
+    /* Check for invalid access request */
+    if(0 == (H5F_INTENT(f) & H5F_ACC_RDWR))
+        HGOTO_ERROR(H5E_OHDR, H5E_BADVALUE, FAIL, "no write intent on file")
 
     /* Make certain we allocate at least a reasonable size for the object header */
     size_hint = H5O_ALIGN_F(f, MAX(H5O_MIN_SIZE, size_hint));