Index: /trunk/ccl/ChangeLog =================================================================== --- /trunk/ccl/ChangeLog (revision 5632) +++ /trunk/ccl/ChangeLog (revision 5633) @@ -1,2 +1,436 @@ +2006-12-19 01:24 gb + + * library/parse-ffi.lisp: Just to be on the safe side (new pathname + bugs, or old ?), append -BAK to (NAMESTRING (TRUENAME path)), not + just (NAMESTRING path) when renaming. (MERGE-PATHNAMES call is + producing a pathane with a non-null VERSION component, as it + arguably should have been all along. + +2006-12-18 13:22 gb + + * lib/sequences.lisp: COERCE/MAKE-SEQUENCE try to simplify vector + types into concrete vector types. + +2006-12-18 13:04 gb + + * level-1/l1-pathnames.lisp: Try (halfheartedly) to maintain + PATHNAME-VERSION of physical pathnames. (Totally meaningless, but + failing to do this leads to many test failures.) + +2006-12-18 13:03 gb + + * level-1/l1-numbers.lisp: Don't call out to do sqrt. + +2006-12-18 12:57 gb + + * level-1/l1-files.lisp: Try (halfheartedly) to maintain + PATHNAME-VERSION of physical pathnames. (Totally meaningless, but + failing to do this leads to many test failures.) + +2006-12-18 12:54 gb + + * level-0/X86/x86-float.lisp: Do single/double-float sqrt inline. + +2006-12-16 05:28 gb + + * level-0/PPC/ppc-float.lisp: Recognize that fsqrt/fsqrts + instructions have a minor opcode in frb and only have one source + operand. + +2006-12-16 05:27 gb + + * level-1/ppc-error-signal.lisp: If we get an FP exception during a + SQRT, note that there's only one operand and that it's in frb. + +2006-12-16 05:25 gb + + * lisp-kernel/ppc-exceptions.c: Handle (some) unimplemented + instructions, notably fsqrt/fqrts, which are optional and not + implemented on some PPC machines. + +2006-12-16 05:24 gb + + * lisp-kernel/ppc-asmutils.s: Add get_fpscr(). + +2006-12-14 19:47 gb + + * lisp-kernel/ppc-exceptions.c: Need to handle SIGQUIT on ppc, too. + +2006-12-14 16:43 gb + + * compiler/X86/X8664/x8664-vinsns.lisp: Constraints on some + double-float arithmetic vinsns should say :double-float, not + :single-float. (I suspect cut-and-paste, frankly.) + +2006-12-14 16:25 gb + + * level-1/l1-lisp-threads.lisp: *initial-lisp-thread*, not + initial-thread. + +2006-12-14 16:24 gb + + * level-1/x86-trap-support.lisp: Use frame pointer when signaling + error. + +2006-12-14 16:14 gb + + * library/lispequ.lisp: Add a %PHYSICAL-PATHAME-VERSION slot to + physical pathnames. (Yes, this is kind of silly.) + +2006-12-14 16:01 gb + + * level-1/x86-trap-support.lisp: Recognize the kernel's attempt to + signal errors for (e.g.) failed memory allocation. + +2006-12-14 16:00 gb + + * lisp-kernel/x86-exceptions.c: If a memory allocation request + fails, signal a lisp error (what a concept.) + +2006-12-14 15:51 gb + + * lisp-kernel/x86-exceptions.h: Define DARWIN_USE_PSEUDO_SIGRETURN. + + Might be worth revisiting this. + +2006-12-14 15:50 gb + + * lisp-kernel/x86-exceptions.c: Go back to using the + pseudo_sigreturn mechanism on Darwin, since doing a real sigreturn + from a fake context may not be reliable. + + When handling an interrupt with a partially created temp stack + frame, do the right thing. (That basically means reset the frame + before calling out and recreating it if we return from the + callout.) + + Handle SIGQUIT on the altstack, to provide a way of forcibly + terminating threads that's a bit more GC-safe than doing it via + #_pthread_cancel. + + Note the tcr of the gc thread on gc entry; if an exception happens + in that thread, print a message and sleep for 60 seconds. + + Note that there's still a race condition here: if thread A gets an + exception and thread B gets the mach_exception_lock (to GC) before + the exception thread can handle the message, we deadlock (thread A + would be suspended in the kernel and can't respond to a suspend + signal.) We can tell that this has happened and make a good guess + as to why it happened, but dealing with it will require some + thought. (If we've already issued a suspend signal and eventually + raise the suspended thread's resume semaphore, it'll suspend and + resume once the mach exception thread makes it runnable. Who + cleans up its exception context in that case ? Does it matter ?) + +2006-12-14 15:19 gb + + * lisp-kernel/thread_manager.c: Ensure that newly-created threads + have all signals unmasked. + + If a tcr is dead, ensure that its stack areas are empty. + + Try to run shutdown_thread_tcr as a pthread cleanup action, + hopefully before Mach clobbers the thread's signal mask. + +2006-12-14 15:16 gb + + * lisp-kernel/pmcl-kernel.c: lazarus() runs with a lock on the + TCR_AREA_LOCK. + +2006-12-14 15:15 gb + + * lisp-kernel/memprotect.h: Don't declare do_memo_reset(), which + hasn't existed in years. + +2006-12-14 15:14 gb + + * level-1/version.lisp: Bump. + +2006-12-14 15:14 gb + + * level-1/l1-readloop.lisp: PREPARE-TO-QUIT: wait for the + termination semaphore. + +2006-12-14 15:13 gb + + * level-1/l1-processes.lisp: PROCESS objects are created with a + termination semaphore. + +2006-12-14 15:12 gb + + * level-1/l1-lisp-threads.lisp: THREAD-ENABLE: second arg is + TERMINATION-SEMAPHORE. + + KILL-LISP-THREAD: don't use #_pthread_cancel, which may force the + target thread to transition from running lisp code to running + foreign code without advertising that fact. Send a SIGQUIT instead, + and trust the SIGQUIT handler to manage that transition. Don't try + to kill the initial thread. + +2006-12-09 09:52 gb + + * level-1/l1-streams.lisp: STRING-OUTPUT-STREAMs need to override + their CLEAR-OUTPUT methods. + + MAKE-STRING-INPUT-STREAM: handle displaced strings better (remember + their displaced-index-offsets.) + +2006-12-09 09:49 gb + + * lisp-kernel/thread_manager.c: Disable underflow traps in new + threads on x86-64. (The PPC never got around to enabling them.) + +2006-12-09 09:47 gb + + * lib/macros.lisp: WITH-OUTPUT-TO-STRING: if string provided, + evaluate the ELEMENT-TYPE expression (for side-effects) and ignore + the result of its evaluation. + + Some fixes in CTYPECASE handling, which hopefully didn't break + anything else. + +2006-12-09 09:36 gb + + * compiler/X86/x862.lisp: In aset2/aset3, ensure that the (final, + 1d) index register is set - even if the index is a constant - if we + think that the store needs EGC memoization. + + (Note that the test case that exposed this involved storing + SINGLE-FLOATs, which don't need EGC write-barrier handling on + 64-bit platforms.) + + In aref2, don't worry about reserving unscaled index registers when + targeting the 1d index; the unscaled index registers aren't live + after, could be NULL if the index is a constant, and any overlap + between one of them and the 1d index should be handled in the + vinsn(s). + +2006-12-09 09:30 gb + + * level-1/l1-typesys.lisp: Ensure that the (relatively new) + built-in-classes SIMPLE-FIXNUM-VECTOR and FIXNUM-VECTOR have + class-translations. + +2006-12-09 09:27 gb + + * level-1/l1-aprims.lisp: In list case of ELT: who says that value + returned by NTHCDR is a LIST ? (It's not necessarily a list, so + don't segfault taking its CAR without checking.) + +2006-12-06 18:48 gb + + * level-0/X86/x86-float.lisp: x86-64 version of + SFLOAT-SIGNIFICAND-ZEROS: count zero bits in the shifted float, not + in %imm1 (which is undefined here ...) + +2006-12-06 17:05 gb + + * compiler/nx0.lisp: DEFINE-COMPILER-MACRO: if a SETF name, use the + right symbol for the named block. + + COMPILER-MACRO-FUNCTION, (SETF COMPILER-MACRO-FUNCTION): deal with + (SETF NAME). + +2006-12-05 22:44 gb + + * release-notes.txt: Update for 061205. + +2006-12-05 20:31 gb + + * lisp-kernel/darwinppc64/Makefile: Change LDFLAGS, to keep ld64 + from putting a linkedit segment in the wrong place and making + 32-bit memory unmappable. + +2006-12-05 20:30 gb + + * level-1/version.lisp: Bump. + +2006-12-05 20:29 gb + + * lisp-kernel/pmcl-kernel.c: Use _exit(), just in case. + +2006-12-05 18:53 gb + + * level-1/l1-reader.lisp: *ALTERNATE-LINE-TERMINATOR* set to + #\Return on #+darwin-target, not just #+darwinppc-target. Change + doc string for *ALTERNATE-LINE-TERMINATOR*. + +2006-12-04 15:13 gb + + * level-1/version.lisp: Append a patchlevel to the version string, + at least until new tarballs can be built. + +2006-12-04 15:12 gb + + * level-1/l1-streams.lisp: If stream element type upgrades to + FIXNUM, upgrade it further (to largest supported SIGNED-BYTE + element-type.) + +2006-12-04 13:02 gb + + * level-0/l0-bignum32.lisp: BIGNUM-BIGNUM-GCD (whatever it's + called): don't bother to shift by 0 bits. Do shift the initial U + if non-zero trailing 0 BITS (was testing for non-zero trailing + DIGITS.) + +2006-12-04 12:52 gb + + * level-1/l1-streams.lisp: Choke earlier on unsupported stream + element-types; do some upgrading to support some more element types + (BIT -> (UNSIGNED-BYTE 8)). + + Maybe support arbitrarily large integer types someday soon. Maybe + not. + +2006-12-04 12:48 gb + + * level-1/l1-clos-boot.lisp: Make built-in classes for istruct + types (compiler stuff, mostly). + +2006-12-04 12:46 gb + + * level-0/X86/X8664/x8664-bignum.lisp: Ensure that %ASHL handles + shift count of 32 (and returns 0 in that case.) + +2006-12-04 12:46 gb + + * level-0/l0-bignum64.lisp: In BIGNUM-SHIFT-RIGHT-LOOP-1, use + %ASHL, not %ILSL. (Don't want nonsense in the high 32 bits, just + in case.) + +2006-12-04 12:43 gb + + * level-0/l0-numbers.lisp: Bugs in GCD-2. Long-standing bugs in + GCD-2 ... + +2006-12-04 09:26 gb + + * level-1/l1-streams.lisp: STREAM-CLEAR-INPUT: don't inherit method + for STRING-INPUT-STREAMs. + +2006-12-04 09:24 gb + + * level-0/l0-bignum64.lisp: BIGNUM-SHIFT-RIGHT-1: handle shift by 0 + bits. BIGNUM-BIGNUM-GCD (whatever it's called): don't bother to + shift by 0 bits. + +2006-12-04 09:21 gb + + * level-1/ppc-trap-support.lisp: Reset error-reentry count before + doing %SLOT-UNBOUND-TRAP. + +2006-12-04 09:21 gb + + * level-1/l1-streams.lisp: Default STREAM-POSITION method for + STREAMs, returns NIL. STREAM-POSITION method for + STRING-INPUT-STREAMs. + +2006-12-04 09:19 gb + + * level-1/l1-pathnames.lisp: Derive ccl-directory from pathname + containing heap image. + +2006-12-03 07:41 gb + + * level-1/l1-streams.lisp: FORCE-OUTPUT, FINISH-OUTPUT on + STRING-OUTPUT-STREAMs are no-ops. + + Need an ioblock-based character-read-vector function for + string-input-streams. + +2006-12-03 07:37 gb + + * lib/describe.lisp: Minimal DESCRIBE support for + CCL::BASIC-STREAM. + +2006-12-03 07:36 gb + + * lib/macros.lisp: If :element-type is supplied in + WITH-OUTPUT-TO-STRING, don't quote it. + +2006-12-03 05:11 gb + + * level-1/x86-trap-support.lisp: Define RIP-REGISTER-OFFSET, which + isn't necessarily 16. + +2006-12-03 05:10 gb + + * level-1/x86-error-signal.lisp: Use RIP-REGISTER-OFFSET, which + isn't necessarily 16. + + Use WITH-ERROR-REENTRY-DETECTION n %xerr-disp; dtrt for + %slot-unbound-trap. + +2006-12-03 05:09 gb + + * level-1/l1-streams.lisp: Tweak stream hierarchy a little, to + avoid some NO-NEXT-METHODs. + +2006-12-03 02:42 gb + + * compiler/nx1.lisp: Handle VALUES type specifiers in THE forms a + bit better. + +2006-12-03 02:41 gb + + * compiler/nx0.lisp: Handle VALUES type specifiers in THE forms a + bit. + +2006-12-03 02:28 gb + + * lisp-kernel/x86-spentry64.s: Do vector bounds checking by + clearing the low byte of the header, shifting it right, and + comparing the index to that. (It's not reliable to shift the index + left.) + +2006-12-03 02:26 gb + + * level-1/l1-unicode.lisp: Memory/vector encode functions: don't + write BOM, leave that up to higher-level code. + + Start to write some higher-level code. + +2006-12-03 02:23 gb + + * level-1/l1-sysio.lisp: Fix some indentation. + +2006-12-03 02:21 gb + + * compiler/X86/X8664/x8664-vinsns.lisp: CHECK-MISC-BOUNDS: shift + the header word right (after clearing its low byte) and compare the + index to that. (Shifting the index left - d'oh - obviously fails + for indices of large magnitude.) + +2006-11-29 21:07 gb + + * lib/db-io.lisp: Indentation change. + + Add CDB-ENUMERATE-KEYS. + +2006-11-29 18:37 gb + + * lib/level-2.lisp: %DESTRUCTURE-LAMBDA-LIST: other callers of + VERIFY-LAMBDA-LIST (hopefully) check to see if the alleged list is + LISTP first. (I hope ...). Do so. + +2006-11-28 18:42 gb + + * lisp-kernel/x86-spentry64.s: Wrong register typo in .SPaset2. + +2006-11-28 16:14 gb + + * examples/cocoa-window.lisp: DEFAULT-FONT: retain the font. + +2006-11-28 16:12 gb + + * examples/objc-runtime.lisp: Wrap a (primitive) + WITH-AUTORELEASE-POOL around the code which detaches the dummy + NSThread. + +2006-11-28 16:09 gb + + * examples/cocoa-listener.lisp: Don't claim that a callback + parameter is ignored; use IGNORABLE instead. + 2006-11-21 01:28 gb