sbcl.sbcl/TODO

SBCL TODO
=========

     "There's nothing an agnostic can't do as long as he doesn't know
     whether he believes in anything or not."
       -- Monty Python.

     "God grant me serenity to accept the code I cannot change, courage
     to change the code I can, and wisdom to know the difference."
       -- Erik Naggum

     "Accumulation of half-understood design decisions eventually
     chokes a program as a water weed chokes a canal. By refactoring
     you can ensure that your full understanding of how the program
     should be designed is always reflected in the program. As a water
     weed quickly spreads its tendrils, partially understood design
     decisions quickly spread their effects throughout your program. No
     one or two or even ten individual actions will be enough to
     eradicate the problem."
       -- Martin Fowler, in _Refactoring: Improving the Design of Existing
          Code_, p. 360

     "I wish I didn't know now what I didn't know then."
       -- Bob Seger

  This files is maintained as part of the SBCL distribution, and
  describes flying pies and moons on sticks that SBCL developers dream
  about. The items are in no particular order.

  The omission of an item is no guarantee that no-one would like it,
  just like the inclusion of an item is no guarantee that someone is
  actively working on it.

  In addition to this file, there is the BUGS file, and there are also
  hundreds of FIXME notes in the sources. (Things marked as KLUDGE are
  in general things which are ugly or confusing, but that, for
  whatever reason, may stay that way indefinitely.)

THREADING INTERFACE

  SB-THREAD has some problems: recursivity of a mutex should probably
  be a feature of the mutex, and not of the lexical location. Thread
  local variables are needed. Sessions and sharing the terminal need
  to be thought about.

PEEPHOLE OPTIMIZER

  Have you ever read SBCL disassembly?

DEFGLOBAL

  Global lexical variables. Esp. since with threads special variable
  accesses is no speed daemon.

FINISHING EXTERNAL FORMATS

  Byte order marks. Newline conventions. A way to specify an external
  format without needing to duplicate code. Fixing the inefficiencies.

TIMEOUTS

  Asyncronous unwinds suck horribly, but to implement reliable systems
  one simply needs timeouts. These should probably be local, since
  otherwise they effectively become asynchronous unwinds. Basically,
  for any potentially blocking operation there should be a :TIMEOUT
  arguent (or a version of the operation that accepts the argument).

ADVICE/FWRAP

  SBCL has an internal function encapsulation mechanism, and is able to
  install breakpoint to function start/end -- this is used to implement
  the instrumentation based profiler and tracing. It would be good to
  have this as an exported interface, and it would be good if the
  SYMBOL-FUNCTION / FDEFINITION confusion was fixed: currently the
  latter returns the underlying definition, whereas the first returns
  the encapsulation.

GENERIC FUNCTION TRACING

  This sucks currently. It would also be good to be able to trace
  individual methods.

POLICY MADNESS

  The interactions between various optimization policies are far from
  obvious. Someone should figure out how to make this better, and fix
  it. One option would be to have just a few (eg. DEBUG, SMALL,
  FAST-SAFE, FAST-UNSAFE) "dominant" policies, and expose the rest
  as separately declarable optimization toggles.

  MAYBE-INLINE is also nice, but it would be good if someone could
  figure out how to get rid of it while retaining the semantics it
  provides. Inlining recursive functions is also something to think
  about.

  INHIBIT-WARNINGS really needs to go away.

WINDOWS

  Needs love.

DARWIN

  Needs love, particularly threads and exceptions/signals. slam.sh is
  also broken there.

DRAGONFLY
  Multithreading does not work. Possibly because of bug in mmap/munmap.
  Hint: Comment out call to os_invalidate in perform_thread_post_mortem
  and threads will work, but space will not be freed, of course.

MISC CLEANUPS

  These need to be taken a good look at -- it might be that some of them
  are already done, or otherwise no longer relevant.)

   * EVAL/EVAL-WHEN/%COMPILE/DEFUN/DEFSTRUCT cleanups:
	** make %COMPILE understand magicality of DEFUN FOO
		w.r.t. e.g. preexisting inlineness of FOO
	** use %COMPILE where COMPILE-TOP-LEVEL used to be used
	** remove redundant COMPILE-TOP-LEVEL and
		FUNCTIONAL-KIND=:TOP-LEVEL stuff from the compiler
   * outstanding embarrassments
	** :IGNORE-ERRORS-P cruft in stems-and-flags.lisp-expr. (It's
		reasonable to support this as a crutch when initially
		bootstrapping from balky xc hosts with their own
		idiosyncratic ideas of what merits FAILURE-P, but it's
		embarrassing to have to use it when bootstrapping
		under SBCL!),
   * miscellaneous simple refactoring
	* belated renaming:
		** rename %PRIMITIVE to %VOP
		** A few hundred things named FN should be named FUN
	* These days ANSI C has inline functions, so..
		** redo many cpp macros as inline functions:
			HeaderValue, Pointerp, CEILING, ALIGNED_SIZE,
			GET_FREE_POINTER, SET_FREE_POINTER,
			GET_GC_TRIGGER, SET_GC_TRIGGER, GetBSP, SetBSP,
			os_trunc_foo(), os_round_up_foo()
		** remove various avoid-evaluating-C-macro-arg-twice
			cruft
  * Some work on conditions emitted by the system
	** eliminate COMPILER-WARN and COMPILER-STYLE-WARN, which
		were simply limited versions of WARN and STYLE-WARN.
	** eliminate use of INHIBIT-WARNINGS by code emitted by the
		system from user code.
	** cause use of INHIBIT-WARNINGS to signal a STYLE-WARNING.
	** eliminate use of INHIBIT-WARNINGS within the system
	** deprecate INHIBIT-WARNINGS, causing its use to signal a
		full WARNING.
	** begin work on developing a class hierarchy of conditions
		along semantic lines.

PCL INTEGRATION

  AKA CLOS in cold init.

HIGH LEVEL SOCKET INTERFACE

  Something slightly above the level of BSD sockets would be nice.

RPC INTERFACE

  For talking with other processes.

BLOCK COMPILATION

  You know, like CMUCL does.

(MOSTLY) PARALLEL GC

  Several algorithms exist, even one would be nice.

INCREMENTAL GC

  For those who need it.