3 * building with CLISP (or explaining why not). This will likely involve
4 a rearrangement of the build system so that it never renames
5 the output from COMPILE-FILE, because CLISP's COMPILE-FILE
6 outputs two (!) files and as far as I can tell LOAD uses both
7 of them. Since I have other motivations for this rearrangement
8 besides CLISPiosyncrasies, I'm reasonably motivated to do it.
9 * urgent EVAL/EVAL-WHEN/%COMPILE/DEFUN/DEFSTRUCT cleanup:
10 ** made inlining DEFUN inside MACROLET work again
11 ** (also, while working on INLINE anyway, it might be easy
12 to flush the old MAYBE-INLINE cruft entirely,
13 including e.g. on the man page)
14 ** fixed bug 137 (more)
15 * faster bootstrapping (both make.sh and slam.sh)
16 ** added mechanisms for automatically finding dead code, and
17 used them to remove dead code
18 ** moved stuff from warm init into cold init where possible
19 (so that slam.sh will run faster and also just because
20 ideally everything would be in cold init)
21 ** profiled and tweaked
22 * fixed (TRACE :REPORT PROFILE ...) interface to profiling
23 * more EVAL/EVAL-WHEN/%COMPILE/DEFUN/DEFSTRUCT cleanup:
24 ** made %COMPILE understand magicality of DEFUN FOO
25 w.r.t. e.g. preexisting inlineness of FOO
26 ** used %COMPILE where COMPILE-TOP-LEVEL used to be used
27 ** removed now-redundant COMPILE-TOP-LEVEL and
28 FUNCTIONAL-KIND=:TOP-LEVEL stuff from the compiler
29 ** (ideally, but perhaps too hard, given what I've discovered
30 about the godawful internals of function debug names):
31 made FUNCTION-NAME logic work on closures, so that
32 various public functions like CL:PACKAGEP which
33 are now implemented as closures (because
34 they're structure slot accessors) won't be so
36 * rewrote long-standing confusing error restarts for redefining
38 * outstanding embarrassments
39 ** cut-and-pasted DEF-BOOLEAN-ATTRIBUTE (maybe easier to fix
40 now that EVAL-WHEN works correctly..)
42 ** :IGNORE-ERRORS-P cruft in stems-and-flags.lisp-expr. (It's
43 reasonable to support this as a crutch when initially
44 bootstrapping from balky xc hosts with their own
45 idiosyncratic ideas of what merits FAILURE-P, but it's
46 embarrassing to have to use it when bootstrapping
48 ** weird double-loading (first in GENESIS, then in warm init)
49 of src/assembly/target/*.lisp stuff, and the associated
50 weirdness of the half-baked state (compiler almost but
51 not quite ready for prime time..) of the system after
53 * fixups now feasible because of pre7 changes
54 ** ANSIfied DECLAIM INLINE stuff (deprecating MAYBE-INLINE)
55 * miscellaneous simple refactoring
57 ** renamed %PRIMITIVE to %VOP
58 * These days ANSI C has inline functions, so..
59 ** redid many cpp macros as inline functions:
60 HeaderValue, Pointerp, CEILING, ALIGNED_SIZE,
61 GET_FREE_POINTER, SET_FREE_POINTER,
62 GET_GC_TRIGGER, SET_GC_TRIGGER, GetBSP, SetBSP,
63 os_trunc_foo(), os_round_up_foo()
64 ** removed various avoid-evaluating-C-macro-arg-twice
66 * added mechanisms for automatically finding dead symbols is
67 package-data.lisp-expr (i.e. those symbols not bound,
68 fbound, defined as types, or whatever), and used them
69 to remove dead symbols
70 * made system handle stack overflow safely unless SAFETY is dominated
72 * Either get rid of or at least rework the fdefinition/encapsulation
73 system so that (SYMBOL-FUNCTION 'FOO) is identically equal to
75 =======================================================================
78 * refactored in preparation for moving CLOS into cold init and merging
79 SB-PCL:FOO with CL:FOO (for FOO=CLASS, FOO=CLASS-OF, etc.)
80 ** systematized support for MOP (new regression tests, maybe
81 new SB-MOP package..) to try to make sure things don't
82 get mislaid in the upcoming CLOS restructuring
83 ** extracted type system from SB-KERNEL into new SB-TYPE
85 ** reimplemented GENERIC-FUNCTION as a primitive object (or
86 maybe made SB-MOP:FUNCALLABLE-STANDARD-OBJECT the
87 primitive object, and then let GENERIC-FUNCTIONs
88 inherit from that) instead of structures with
89 :ALTERNATE-METACLASS and funcallableness. Now
90 FUNCALLABLE-INSTANCE can go away. (And now the new
91 funcallable primitive objects need to go into
92 collections like *FUN-HEADER-WIDETAGS* where
93 FUNCALLABLE-INSTANCE objects used to be.)
94 ** reimplemented CONDITIONs as primitive objects instead of
95 structures with :ALTERNATE-METACLASS. Now (between
96 this and the change to GENERIC-FUNCTIONs)
97 DEFSTRUCT :ALTERNATE-METACLASS can go away.
98 ** (maybe) Now INSTANCE_POINTER_LOWTAG can become just
99 STRUCTURE_POINTER_LOWTAG, and the concept of
100 SB-KERNEL:INSTANCE (including INSTANCEP,
101 (SPECIFIER-TYPE 'INSTANCE), etc.) can go away.
102 * moved CLOS into cold init, in order to allow CLOS to be used in the
103 implementation of the core system (e.g. the type system and the
104 compiler) and in order to support merger of CL:CLASS with
106 * (maybe) eliminated warm init altogether in favor of cold init
107 * (maybe, especially if warm init can be eliminated) rationalized
108 the build process, fixing miscellaneous pre-0.5.0 stuff that's
109 transparently not the right thing
110 ** removed separate build directories, now just building in
111 place with .sbclcoldfasl extensions
112 * (maybe) more refactoring in preparation for merging SB-PCL:FOO
113 into CL:FOO: reimplemented type system OO dispatch
114 (!DEFINE-TYPE-METHOD, etc.) in terms of CLOS OO dispatch
115 * merged SB-PCL:FOO into CL:FOO (and similarly CLASS-OF, etc.)
116 * added some automatic tests for basic binary compatibility, in hopes
117 that it might be practical to maintain binary compatibility
118 between minor maintenance releases on the stable branch (but no
119 promises, sorry, since I've never tried to do this before, and
120 have no idea how much of a pain this'll be)
121 ========================================================================
122 for 1.0 (fixes of lower priority which I'd nonetheless be embarrassed
123 to leave unfixed in 1.0):
124 * all too many BUGS entries and FIXMEs
125 =======================================================================
126 other priorities, no particular time:
128 * bug fixes, especially really annoying bugs (ANSI or not) and any
129 ANSI bugs (i.e. not just bugs in extras like the debugger or
130 "declarations are assertions", but violations of the standard)
131 * better communication with the outside world (scratching WHN's
132 personal itch): I don't want socket-level stuff so much as I
133 want RPC-level or higher (CORBA?) interfaces and (possibly
134 through RPC or CORBA) GUI support
135 =======================================================================
136 important but out of scope (for WHN, anyway: Patches from other people
137 are still welcome!) until after 1.0:
139 * sadly deteriorated support for ANSI-style block compilation
140 (static linking of DEFUNs within a single file or
141 WITH-COMPILATION-UNIT)
142 * various GC issues (exuberant cut-and-paste coding,
143 possibly dangerously over-conservative handling
144 of neighbors of function objects, general GC efficiency)
145 * package issues other than SB!TYPE, SB!MOP, and dead exported
147 * Any systematic effort to fix compiler consistency checks is
148 out of scope. (However, it still might be possible to
149 determine that some or all of them are hopelessly stale
151 =======================================================================
152 other known issues with no particular target date:
154 bugs listed on the man page
156 hundreds of FIXME notes in the sources from WHN
158 various other unfinished business from CMU CL and before, marked with
159 "XX" or "XXX" or "###" or "***" or "???" or "pfw" or "@@@@" or "zzzzz"
160 or probably also other codes that I haven't noticed or have forgotten.
162 (Things marked as KLUDGE are in general things which are ugly or
163 confusing, but that, for whatever reason, may stay that way
165 =======================================================================
166 "There's nothing an agnostic can't do as long as he doesn't know
167 whether he believes in anything or not."
170 "God grant me serenity to accept the code I cannot change, courage to
171 change the code I can, and wisdom to know the difference."
174 "Accumulation of half-understood design decisions eventually chokes a
175 program as a water weed chokes a canal. By refactoring you can ensure
176 that your full understanding of how the program should be designed is
177 always reflected in the program. As a water weed quickly spreads its
178 tendrils, partially understood design decisions quickly spread their
179 effects throughout your program. No one or two or even ten individual
180 actions will be enough to eradicate the problem."
181 -- Martin Fowler, in _Refactoring: Improving the Design of Existing
184 "I wish I didn't know now what I didn't know then."