Register bugs in structure slot type checking (CSR and WHN, 2003-01-19).
* Add dependent optimization qualities: LET-CONVERTION, TYPE-CHECK, VERIFY-ARG-COUNT; * add SB-EXT:DESCRIBE-COMPILER-POLICY;
Fixed some bugs revealed by Paul Dietz' test suite: ** BOA constructor with &AUX argument without a default value does not cause a type error; ** CONSTANTP now returns true for all self-evaluating objects.
added hooks, more or less as sketched by dan_b on sbcl-devel, to support kmr/acl-style alternate toplevel loop
rewrote MERGE-LISTS* to avoid using a global variable for scratch (motivated by Wolfgang Jenkner's bug report sbcl-devel 2003-01-05, and also by the prospect of threads) rewrote stuff calling MERGE-LISTS* to coerce function designators early once and for all (instead of on every call to MERGE-LISTS*) added test cases trivial style/comment/name tweaks
Fix for averrance failure due to DEBUG-RETURN interfering with type system (explanation APD sbcl-devel 2003-01-10) ... disable insertion of CATCH if it was disabled in an outer LAMBDA; ... enable insertion of CATCH when compiling top-level forms, so that the evaluator continues to have DEBUG-RETURN goodness.
Fix bugs identified by Paul Dietz (cmucl-imp 2003-01-03) for printing and reading arrays ... inspired by but slightly different from Gerd Moellmann fixes cmucl-imp 2003-01-04
Implement the RETURN debugger command. ... CATCH block insertion conditional on (> DEBUG (MAX SPEED SPACE)) ... change interactor policy to make this the case ... note as experimental in DEBUG help string
Try harder to find GNU make (instead of first running make-host-1.sh, then bailing out in make-target-1.sh) as per Fufie, rudi, and Krystof on #lisp. (Google says "don't be evil", and while we're not that good, at least we can try not to be too annoying.:-) Make chill.lisp not die on short package names. (inspired by Rudi Schlatte's patch sbcl-devel 2003-01-04)
Fixed bug in embedded calls of SORT (reported and investigated by Wolfgang Jenkner).
Merged in the mti-1202 branch: ... restored constraint propagation from type assertions; ... removed bogus type declarations in WITH-PACKAGE-ITERATOR;
Clean up regressions, easy new bugs, etc. ... fix one or two embarrassing bugs in ctor MAKE-INSTANCE optimization patch: * REINITIALIZE-INSTANCE now checks arguments again * defining methods on MAKE-INSTANCE (or its siblings) now invalidates the optimized methods for subclasses too ... trivial ADJUST-ARRAY patch
trivial comment fix
belatedly checking in version.lisp again, since I keep doing cvs up, select all files, update version.lisp-expr, check in:-(
new BUGS entry re. (I assume) 0.7.10.31 ctor.lisp stuff trivial OAOO tweak on 0.7.10.32 code
Fix bug in CONSTANT-REFERENCE-P: function name may be (SETF symbol).
Installed ctor.lisp MAKE-INSTANCE optimization (from Gerd Moellmann, via CSR sbcl-devel 2002-12-21) ... wrote tests for those bugs which it fixes ... do not delete fast-init.lisp yet; waiting for user feedback before destroying some CVS history ... update COPYING information to reflect the new copyright owner
Fix compiler crash in new COERCE defoptimizer ... CAREFUL-SPECIFIER-TYPE can return NIL, so handle that case too. (Might it not be nice to provide a compiler warning? Or should that be left to the IR1-TRANSFORM?)
Fixed bug 231: SETQ did not correctly check the type of a variable being set (reported by Robert E. Brown)
Fix up a theoretical bug in the COMPILE DERIVE-TYPE-OPTIMIZER ... '(VALUES FUNCTION BOOLEAN BOOLEAN), not 'FUNCTION (why only a "theoretical" bug? Well, at present, SBCL's type engine returns (VALUES FUNCTION BOOLEAN BOOLEAN) for (VALUES-TYPE-INTERSECTION 'FUNCTION '(VALUES (OR FUNCTION SYMBOL) BOOLEAN BOOLEAN)) so it DWIM. However, CMUCL doesn't (it returns the empty type instead), which led to the discovery of this error by Raymond Toy when he tried to port the optimizer.)
Removed several declarations from src/code/[target-]package.
BUGS file cleanup: ... removed bug entries 178 and 199. New bug: pretty printing conflicts with readability.
Fix bug 223: functional binding of a symbol is constant only if the symbol is in CL package.
Fix obscure DEFSTRUCT :CONC-NAME handling bug, revealed by Paul Dietz' ansi-tests ... :CONC-NAME NIL means something different from :CONC-NAME "" Also some belated tests for extra COERCE logic
Fix bug 222 (as per APD sbcl-devel 2002-12-11) ... define the macro in the correct (restricted) lexical context. Fix bug in COERCE [e.g. (COERCE 2 '(SINGLE-FLOAT 3.0 4.0))] ... when converting to a real type, don't be so lenient if the original datum is a rational. Define improved DERIVE-TYPE-OPTIMIZER for COERCE ... for constant RESULT-TYPE arguments, write a branch that understands complex canonicalization ... leave in old branch for ARRAY-ELEMENT-TYPE return types Write DERIVE-TYPE-OPTIMIZER for COMPILE ... (COMPILE NIL <x>) returns an object of type FUNCTION