1 planned incompatible changes in 0.8.x:
2 * (not done yet, but planned:) When the profiling interface settles
3 down, it might impact TRACE. They both encapsulate functions, and
4 it's not clear yet how e.g. UNPROFILE will interact with TRACE
5 and UNTRACE. (This shouldn't matter, though, unless you are using
6 profiling. If you never profile anything, TRACE should continue to
8 * (not done yet, but planned:) Inlining can now be controlled the
9 ANSI way, without MAYBE-INLINE, since the idiom
10 (DECLAIM (INLINE FOO))
12 (DECLAIM (NOTINLINE FOO))
13 (DEFUN BAR (..) (FOO ..))
14 (DEFUN BLETCH (..) (DECLARE (INLINE FOO)) (FOO ..))
15 now does what ANSI says it should. The CMU-CL-style
16 SB-EXT:MAYBE-INLINE declaration is now deprecated and ignored.
21 ** ports with less than 256Mb of heap (sparc, ppc and mips)
22 now don't fail bit-vector.impure-cload.lisp
23 * faster bootstrapping (both make.sh and slam.sh)
24 ** added mechanisms for automatically finding dead code, and
25 used them to remove dead code
26 ** moved stuff from warm init into cold init where possible
27 (so that slam.sh will run faster and also just because
28 ideally everything would be in cold init)
29 ** profiled and tweaked
30 * fixed (TRACE :REPORT PROFILE ...) interface to profiling
31 * more EVAL/EVAL-WHEN/%COMPILE/DEFUN/DEFSTRUCT cleanup:
32 ** made %COMPILE understand magicality of DEFUN FOO
33 w.r.t. e.g. preexisting inlineness of FOO
34 ** used %COMPILE where COMPILE-TOP-LEVEL used to be used
35 ** removed now-redundant COMPILE-TOP-LEVEL and
36 FUNCTIONAL-KIND=:TOP-LEVEL stuff from the compiler
37 ** (ideally, but perhaps too hard, given what I've discovered
38 about the godawful internals of function debug names):
39 made FUNCTION-NAME logic work on closures, so that
40 various public functions like CL:PACKAGEP which
41 are now implemented as closures (because
42 they're structure slot accessors) won't be so
44 * outstanding embarrassments
45 ** :IGNORE-ERRORS-P cruft in stems-and-flags.lisp-expr. (It's
46 reasonable to support this as a crutch when initially
47 bootstrapping from balky xc hosts with their own
48 idiosyncratic ideas of what merits FAILURE-P, but it's
49 embarrassing to have to use it when bootstrapping
51 * fixups now feasible because of pre7 changes
52 ** ANSIfied DECLAIM INLINE stuff (deprecating MAYBE-INLINE,
53 including e.g. on the man page)
54 ** (maybe) allow INLINE of a recursive function, so that the
56 * miscellaneous simple refactoring
58 ** renamed %PRIMITIVE to %VOP
59 ** A few hundred things named FN and FCN should be
60 named FUN (but maybe not while drichards is
61 working on a Windows port).
62 * These days ANSI C has inline functions, so..
63 ** redid many cpp macros as inline functions:
64 HeaderValue, Pointerp, CEILING, ALIGNED_SIZE,
65 GET_FREE_POINTER, SET_FREE_POINTER,
66 GET_GC_TRIGGER, SET_GC_TRIGGER, GetBSP, SetBSP,
67 os_trunc_foo(), os_round_up_foo()
68 ** removed various avoid-evaluating-C-macro-arg-twice
70 * Either get rid of or at least rework the fdefinition/encapsulation
71 system so that (SYMBOL-FUNCTION 'FOO) is identically equal to
73 * Make the system sources understandable to the system, so that
74 searching for sources doesn't error out quite so often
75 (e.g. in error handlers)
76 ** provided a suitable readtable for reading in the source
77 files when necessary, and a mechanism for activating
78 this readtable rather than the standard one.
79 * Some work on conditions emitted by the system
80 ** eliminated COMPILER-WARN and COMPILER-STYLE-WARN, which
81 were simply limited versions of WARN and STYLE-WARN.
82 ** made STYLE-WARN parallel WARN more closely (by accepting
83 a condition type, which should be a subtype of
84 STYLE-WARNING, and initargs, as well as a format
85 string and format arguments for SIMPLE-STYLE-WARNING.
86 (WARN can also be used to signal STYLE-WARNINGs, but
87 STYLE-WARN helps to document the code)
88 ** eliminated use of INHIBIT-WARNINGS by code emitted by the
89 system from user code.
90 ** caused use of INHIBIT-WARNINGS to signal a STYLE-WARNING.
91 ** eliminated use of INHIBIT-WARNINGS within the system
92 ** deprecated INHIBIT-WARNINGS, causing its use to signal a
94 ** began work on developing a class hierarchy of conditions
96 ** annotated conditions emitted by the system to have
97 references to documentation where applicable, so that
98 users can easily find an explanation for the
99 conditions they're seeing.
101 =======================================================================
104 [ note: much of the below refers to preparation for merging SB-PCL:FOO
105 and CL:FOO. However, it turned out to be surprisingly
106 straightforward to do this notional end goal without doing many of
107 the preparatory operations. That doesn't mean that plenty of the
108 goals below aren't worthwhile, but the motivation is somewhat
111 * refactored in preparation for moving CLOS into cold init and merging
112 SB-PCL:FOO with CL:FOO (for FOO=CLASS, FOO=CLASS-OF, etc.)
113 ** systematized support for MOP (more regression tests, maybe)
114 to try to make sure things don't get mislaid in the
115 upcoming CLOS restructuring
116 ** extracted type system (and maybe CLASSOIDs) from SB-KERNEL
117 into new SB-TYPE package
118 ** reimplemented GENERIC-FUNCTION as a primitive object (or
119 maybe made SB-MOP:FUNCALLABLE-STANDARD-OBJECT the
120 primitive object, and then let GENERIC-FUNCTIONs
121 inherit from that) instead of structures with
122 :ALTERNATE-METACLASS and funcallableness. Now
123 FUNCALLABLE-INSTANCE can go away. (And now the new
124 funcallable primitive objects need to go into
125 collections like *FUN-HEADER-WIDETAGS* where
126 FUNCALLABLE-INSTANCE objects used to be.)
127 ** reimplemented CONDITIONs as primitive objects instead of
128 structures with :ALTERNATE-METACLASS. Now (between
129 this and the change to GENERIC-FUNCTIONs)
130 DEFSTRUCT :ALTERNATE-METACLASS can go away.
131 ** (maybe) Now INSTANCE_POINTER_LOWTAG can become just
132 STRUCTURE_POINTER_LOWTAG, and the concept of
133 SB-KERNEL:INSTANCE (including INSTANCEP,
134 (SPECIFIER-TYPE 'INSTANCE), etc.) can go away.
135 * moved CLOS into cold init, in order to allow CLOS to be used in the
136 implementation of the core system (e.g. the type system and the
137 compiler) and in order to support merger of CL:CLASS with
139 * (maybe) eliminated warm init altogether in favor of cold init
140 * (maybe, especially if warm init can be eliminated) rationalized
141 the build process, fixing miscellaneous pre-0.5.0 stuff that's
142 transparently not the right thing
143 ** removed separate build directories, now just building in
144 place with .sbclcoldfasl extensions
145 * (maybe) more refactoring in preparation for merging SB-PCL:FOO
146 into CL:FOO: reimplemented type system OO dispatch
147 (!DEFINE-TYPE-METHOD, etc.) in terms of CLOS OO dispatch
148 * added some automatic tests for basic binary compatibility, in hopes
149 that it might be practical to maintain binary compatibility
150 between minor maintenance releases on the stable branch (but no
151 promises, sorry, since I've never tried to do this before, and
152 have no idea how much of a pain this'll be)
153 ========================================================================
154 for 1.0 (fixes of lower priority which I'd nonetheless be embarrassed
155 to leave unfixed in 1.0):
156 * all too many BUGS entries and FIXMEs
157 =======================================================================
158 other priorities, no particular time:
160 * bug fixes, especially really annoying bugs (ANSI or not) and any
161 ANSI bugs (i.e. not just bugs in extras like the debugger or
162 "declarations are assertions", but violations of the standard)
163 * better communication with the outside world (scratching WHN's
164 personal itch): I don't want socket-level stuff so much as I
165 want RPC-level or higher (CORBA?) interfaces and (possibly
166 through RPC or CORBA) GUI support
167 * Especially when ldb is not compiled in, the default "assertion failed"
168 behaviour in many parts of the runtime is unfriendly. It may
169 be appropriate to look at some of these and see if they can be
170 handled in some less abrupt way than aborting
171 =======================================================================
172 important but out of scope (for WHN, anyway: Patches from other people
173 are still welcome!) until after 1.0:
175 * sadly deteriorated support for ANSI-style block compilation
176 (static linking of DEFUNs within a single file or
177 WITH-COMPILATION-UNIT)
178 * various GC issues (exuberant cut-and-paste coding,
179 possibly dangerously over-conservative handling
180 of neighbors of function objects, general GC efficiency)
181 * package issues other than SB!TYPE, SB!MOP, and dead exported
183 * Any systematic effort to fix compiler consistency checks is
184 out of scope. (However, it still might be possible to
185 determine that some or all of them are hopelessly stale
187 =======================================================================
188 other known issues with no particular target date:
190 bugs listed on the man page
192 hundreds of FIXME notes in the sources from WHN
194 various other unfinished business from CMU CL and before, marked with
195 "XX" or "XXX" or "###" or "***" or "???" or "pfw" or "@@@@" or "zzzzz"
196 or probably also other codes that I haven't noticed or have forgotten.
198 (Things marked as KLUDGE are in general things which are ugly or
199 confusing, but that, for whatever reason, may stay that way
201 =======================================================================
202 "There's nothing an agnostic can't do as long as he doesn't know
203 whether he believes in anything or not."
206 "God grant me serenity to accept the code I cannot change, courage to
207 change the code I can, and wisdom to know the difference."
210 "Accumulation of half-understood design decisions eventually chokes a
211 program as a water weed chokes a canal. By refactoring you can ensure
212 that your full understanding of how the program should be designed is
213 always reflected in the program. As a water weed quickly spreads its
214 tendrils, partially understood design decisions quickly spread their
215 effects throughout your program. No one or two or even ten individual
216 actions will be enough to eradicate the problem."
217 -- Martin Fowler, in _Refactoring: Improving the Design of Existing
220 "I wish I didn't know now what I didn't know then."