The following patch is to fix a data corruption issue that showed up on
one of the test scripts.
I wrote up some more details in the thread:
Re: Data corruption - [EMAIL PROTECTED] - PERL_HV_ARRAY_ALLOC_BYTES?
The summary is that the size of the structure allocated for the HV data
was based on the size of a pointer on the host system, and not the size
of the structure.
So many writes to the structure will be past the allocated amount.
From what I can see, if and when the process access violation that
results from this data corruption occurs is dependent on what logical
names or symbols exist in the local VMS environment.
-John
[EMAIL PROTECTED]
Personal Opinion Only
--- hv.h_blead Sat Jul 30 22:30:05 2005
+++ hv.h Sat Jul 30 22:31:15 2005
@@ -352,13 +352,13 @@
/* Default to allocating the correct size - default to assuming that malloc()
is not broken and is efficient at allocating blocks sized at powers-of-two.
*/
-# define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE*))
+# define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE))
#else
# define MALLOC_OVERHEAD 16
# define PERL_HV_ARRAY_ALLOC_BYTES(size) \
(((size) < 64) \
- ? (size) * sizeof(HE*) \
- : (size) * sizeof(HE*) * 2 - MALLOC_OVERHEAD)
+ ? (size) * sizeof(HE) \
+ : (size) * sizeof(HE) * 2 - MALLOC_OVERHEAD)
#endif
/* Flags for hv_iternext_flags. */
