Changeset 5961

Timestamp:

Feb 23, 2007, 9:20:04 AM (18 years ago)

Author:

Gary Byers

Message:

* empty log message *

File:

• trunk/ccl/ChangeLog (modified) (1 diff)

trunk/ccl/ChangeLog

@@ +1 @@
+2007-02-23 10:13  gb
+
+        * lib/foreign-types.lisp: Changes to GUESS-ALIGNMENT.  Again, this
+        only makes sense for cases where we have to guess; a bitfield
+        that's 8 bits wide still has an alignment of 1.
+        
+        Factor in the alignment of the containing int when bitfields are
+        involved.
+        
+        When parsing a record type, set the "bits" slot of each field. 
+        (Not too many things look at that slot, but we want the effect of
+        loading a record from the database to match the effect of parsing a
+        definition as exactly as possible, and to be able to use EQUALP to
+        compare the results.)
+
+2007-02-23 10:06  gb
+
+        * lib/db-io.lisp: Some extra fields in
+        objc-message-info/objc-method-info structures.
+        
+        When calculating overall alignment of a structure, treat bitfields
+        as if they were :unsigned-int (since there's a "containing"
+        unsigned int that affects alignment.  N.B: this should be imposed
+        for anything that has an alignment of 1, not just a size of 1.)
+        
+        When loading integer types, try to use the precomputed types rather
+        than consing a new one.  This (a) reduces consing a little (b) is
+        more likely to get the alignment correct.  For (b), we really need
+        to distinguish between bitfields and addressable integers.
+
+2007-02-23 09:30  gb
+
+        * compiler/X86/x862.lisp: Nonsense in x862-set-bit.
+
 2007-02-22 02:33  gb