Re: MCL 2.0b1p3 bitmap problems

[Bill St. Clair <bill>]

   Date: Thu, 20 Jun 91 17:19:50 EDT
   From: Adnan Hamid <ahamid@clarity.Princeton.EDU>
   To: info-mcl@cambridge.apple.com
   Subject: MCL 2.0b1p3 bitmap problems
   
   A while back a patch was sent out for make-bitmap, with the :storage keys for
   rref/rset explicitly defined to be :pointer.  However the same problem occurs
   in various other situations:
   
   ? (setq bm (make-bitmap 0 0 10 10))
   #<A Mac Zone Pointer Size 34 #x2C425C>
   ? (rref bm :bitmap.bounds)
   > Error: Illegal attempt to get a pointer to a :RECT within a handle
   > While executing: CCL::EXPAND-HANDLE-RECORD-GET
   > Type Command-. to abort.
   See the RestartsI menu item for further choices.
   1 > 
   
   and the results returned from (copy-record bm :bitmap) contains garbage values.
   
   Could the problem instead be in interfaces;QUICKDRAW.lisp where 
   
   (defrecord (BitMap :handle) 
      (baseAddr :pointer)
      (rowBytes :signed-integer)
      (bounds :rect)
      )
   
   should be defined as 
   
   (defrecord (BitMap :pointer) 
      (baseAddr :pointer)
      (rowBytes :signed-integer)
      (bounds :rect)
      )  
   ?
   
As I think I said when I sent out the MAKE-BITMAP patch, the only way
around these problems is to always be explicit about :STORAGE type.
Our Pascal->Lisp translator makes the default :STORAGE type for a
record be :HANDLE if there is a handle type defined that points at it:

   BitMapPtr = ^BitMap;
   BitMapHandle = ^BitMapPtr;
   BitMap = RECORD
       baseAddr: Ptr;
       rowBytes: INTEGER;
       bounds: Rect;
       END;

becomes:

   BitMapPtr = ^BitMap;
   BitMapHandle = ^BitMapPtr;
   BitMap = RECORD
       baseAddr: Ptr;
       rowBytes: INTEGER;
       bounds: Rect;
       END;

but:

   PenState = RECORD
       pnLoc: Point;
       pnSize: Point;
       pnMode: INTEGER;
       pnPat: Pattern;
       END;

becomes:

   (defrecord PenState 
      (pnLoc :point)
      (pnSize :point)
      (pnMode :signed-integer)
      (pnPat :pattern)
      )

The default :STORAGE types are likely to change for at least a few of
the records each time a new set of interface files is released (each
time the OS is updated).  Hence, your code (and ours) needs to take
responsibility for ensuring that you get a pointer when you want a
pointer and a handle when you want a handle.  The following macros
should help (these will be included in the next round of patches for
2.0b1):

(defmacro href (pointer accessor)
  `(rref ,pointer ,accessor :storage :handle))

(defmacro pref (pointer accessor)
  `(rref ,pointer ,accessor :storage :pointer))

(defmacro hset (pointer accessor thing)
  `(rset ,pointer ,accessor ,thing :storage :handle))

(defmacro pset (pointer accessor thing)
  `(rset ,pointer ,accessor ,thing :storage :pointer))

(defmacro make-record-handle (record-type &rest initforms)
  `(make-record 
    ,(if (consp record-type)
       `(,(car record-type) :storage :handle ,@(cdr record-type))
       `(,record-type :storage :handle))
    ,@initforms))

(defmacro make-record-pointer (record-type &rest initforms)
  `(make-record 
    ,(if (consp record-type)
       `(,(car record-type) :storage :pointer ,@(cdr record-type))
       `(,record-type :storage :handle))
    ,@initforms))

   Incedently, (copy-record) fails to work if the optional record-type argument 
   is not provided:
   [...]

Thank you.  Will be fixed.
   
   And for anyone who is doing source patches (from p1&2 etc.), it might be 
   useful to include in the patch release notes instructions to do 
   
   (reindex-interfaces)
   
   after completeing the patches. 

Many people have had problems because we forgot to mention this.