Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Commits on Aug 3, 2002
  1. @telent

    0.7.6.gc-cleanup-branch.7

    telent authored
    	Remove mailing list address from "please report this as a bug"
    	message in gc-common.c.  Minor edit to manual page
Commits on Aug 2, 2002
  1. @telent

    0.7.6.gc-cleanup-branch.6

    telent authored
    	Found some missing #ifdef GENCGC -> LISP_FEATURE_GENCGC replacements:
    	purify now works again in gencgc
  2. @telent

    0.7.6.gc-cleanup-branch.5:

    telent authored
    	Oodles of gcc warning cleanups for machines with different-sized
    	lispobj and void * (e.g. alpha)
    	cheneygc runs again
Commits on Jul 31, 2002
  1. @telent

    0.7.6.gc-cleanup-branch.4

    telent authored
    	Mostly-mechanical preprocessor symbol cleanup
    	GENCGC->LISP_FEATURE_GENCGC
  2. @telent

    0.7.6.gc-cleanup-branch.3:

    telent authored
    	rename gc.c to cheneygc.c
    	twiddle Config.* to cope
    	no changes were made to actual file contents: this is just a rename
  3. @telent

    0.7.6.gc-cleanup-branch.2

    telent authored
    	compilation fixes for cheneygc:
    
    	... some objects lost their `inline' and/or `static'
    	 qualifiers, so we can see tham from both gc-common.o and gc.o
    
    	... some variable names fixed up
    
    	... gc_general_alloc becomes the supported
             allocation-during-gc interface, so the flags for
             boxed/unboxed, fast/slow are now needed in gc-internal.h to
             be shared by both collectors.  (All we do in cheneygc is
             ignore them, but that's not the point)
    
    	... copy_foo_object defns created as wrappers for copy_object.
             these will go away again later in favour of ignorable flags
             to copy_object
  4. @telent

    new

    telent authored
  5. @telent

    0.7.6.gc-cleanup-branch.1:

    telent authored
    	Note that this is all on a branch.  It is not expected to be stable
    	or even necessarily to build on all ports
    
    	First iteration towards Once And Only Once in garbage collection,
    
    	(1) collects together most of the scavenging/translating/sizing
    	routines into the new gc-common.c file
    
    	(2) delete lots of the duplicated code in gencgc gc_*_alloc()
    	routines
    
    	The old Cheney stop & copy gc (henceforth known as CHENEYGC)
    	probably does not build in this checkin.  GENCGC builds and
    	works at least superficially (passes tests etc)
Commits on Jul 30, 2002
  1. @telent

    0.7.6.11:

    telent authored
    	update supported architecture list in manual page.  No exciting
    	changes
Commits on Jul 29, 2002
  1. 0.7.6.10:

    William Harold Newman authored
    	suppressed DEFTRANSFORM CONCATENATE, since we still haven't
    		fixed bug 188 (so now CONCATENATE will presumably run
    		more slowly, but at least it won't have absurdly long
    		compile times)
  2. 0.7.6.9:

    William Harold Newman authored
    	documentation tweaks
Commits on Jul 26, 2002
  1. @csrhodes

    0.7.6.8:

    csrhodes authored
    	Final deletion of %DETECT-STACK-EXHAUSTION, as per APD sbcl-devel
    		2002-07-26
    	... delete export of %D-S-E
    	... increase fasl file version number
    	Minor BUGS cleanup and cross-referencing
Commits on Jul 25, 2002
  1. @csrhodes

    0.7.6.7:

    csrhodes authored
    	Fix up the SPARC fixups, *sigh*
    	... not only did my previous fix not run, it didn't compile, as
    		SIMPLE_FUN_CODE_OFFSET isn't defined in C, only in assembler.
    	... rewrite code address determination in two stages, and save
    		the fun object itself to go into reg_CODE on the SPARC
    	... [ keen observers will note that this is all terribly confusing. ]
  2. @csrhodes

    0.7.6.6:

    csrhodes authored
    	Fix two typos for SB-SHOW
Commits on Jul 24, 2002
  1. @telent

    0.7.6.something

    telent authored
    	More stack checking fixes
    	... when frobbing signal context on x86, set arg count (ecx) to 0,
    	    should fix problem observed on OpenBSD
    	... define a special control-stack-exhausted condition so that
                ignore-errors won't ignore stack exhaustion
    	... fix the test to match the condition
  2. @csrhodes

    0.7.6.4:

    csrhodes authored
    	SPARC fixups
    	... make DB's wonderful stack exhaustion work on the SPARC
    		adjust NPC logic
    		add in fixup for bizarre SPARC calling convention
    	... improve signal handling on SPARC/Linux
    		make a distinction between interrupt_handle_now() as called
    			from handlers, and interrupt_handle_now_handler()
    			installed as the sa_sigaction bit.
Commits on Jul 23, 2002
  1. 0.7.6.3:

    William Harold Newman authored
    	tweaked bsd-os.h to make the new sigaltstack(2) stuff build on
    		OpenBSD/x86
    	fixed bug 189: Now FLET and LABELS inlining respects NOTINLINE
    		declarations as required by ANSI.
    	While I'm at it, suppress FLET/LABELS inlining when (> DEBUG SPEED)
    		too.
  2. @telent

    0.7.6.2:

    telent authored
    	Revert random changes in my (Daniel's) checkout area that were
    	mistakenly committed in 0.7.6.1.  Sorry!
  3. @telent

    0.7.6.1:

    telent authored
    	Mostly-tested but still considered "experimental" non-invasive
    	stack exhaustion checking, using a guard page at the end of the
    	stack and an extra clause in the sigsegv (on some ports, sigbus)
    	handler.  One day there will be an internals doc with the
    	gory details: for now, try http://ww.telent.net/diary/2002/7/#23.59392
  4. 0.7.6:

    William Harold Newman authored
    	release, tagged as sbcl_0_7_6
Commits on Jul 22, 2002
  1. 0.7.5.22:

    William Harold Newman authored
    	deleted email addresses in the wake of autospam to cmucl-imp
    		("Attn: If you find bugs in CMU CL please report them to")
  2. @csrhodes

    0.7.5.21:

    csrhodes authored
    	Complilation fix for Alpha/Linux (from DB, fixing CSR bogosity)
  3. @csrhodes

    0.7.5.20:

    csrhodes authored
    	NEWS tweaking.
Commits on Jul 19, 2002
  1. 0.7.5.19:

    William Harold Newman authored
    	added FILES section of sbcl.1
    	screened CHECK-FULL-CALL with a test on *FAILURE-P* after one
    		too many times getting its "aiee! a compiler bug!"
    		complaints when the real problem is just that the
    		compiler is bravely trying to carry on despite
    		previously-detected programmer errors
    	renamed CHECK-FULL-CALL to PONDER-FULL-CALL because it has
    		side-effects, and even though they're very innocuous
    		side effects, naming a side-effectful function
    		CHECK-FOO just seems evil in principle
Commits on Jul 18, 2002
  1. @csrhodes

    0.7.5.18:

    csrhodes authored
    	Get the NAME initarg for UNDEFINED-FUNCTION right (patch CSR
    		"PPC and undefined functions" sbcl-devel 2002-07-17)
  2. @telent

    0.7.5.17: whitespace! there is no substantive change between this and

    telent authored
    	  0.7.5.16, but all of src/runtime/*.[ch] has been eyeballed and
    	  weird indenting fixed up in favour of a uniform 4 charcters
    	  everywhere.  We give this change its version number to make
    	  it easier to diff previous and subsequent changes without
    	  white(space) noise adding to the confusion.
  3. @telent

    0.7.5.17: whitespace! there is no substantive change between this and

    telent authored
    	  0.7.5.16, but all of src/runtime/*.[ch] has been eyeballed and
    	  weird indenting fixed up in favour of a uniform 4 charcters
    	  everywhere.  We give this change its version number to make
    	  it easier to diff previous and subsequent changes without
    	  white(space) noise adding to the confusion.
  4. @telent

    0.7.5.16:

    telent authored
    	on targets that use the stop'n'copy gc, fix the "gc trigger
    	mysteriously ends up unset after MAYBE-GC doesn't" workaround,
    	to put the new barrier near the end of dynamic space instead of
    	leaving it only a page away.
  5. 0.7.5.15:

    William Harold Newman authored
    	fixed handling of INTERSECTION-TYPE corner case in
    		DEFTRANSFORM %DATA-VECTOR-AND-INDEX (used to fail with
    		TYPE-ERROR at compile time) by making DEFTRANSFORMs
    		call UPGRADED-ELEMENT-TYPE-SPECIFIER-OR-GIVE-UP instead
    		of assuming ARRAY-TYPE and calling
    		ARRAY-TYPE-SPECIALIZED-ELEMENT-TYPE directly
Commits on Jul 17, 2002
  1. @csrhodes

    0.7.5.14:

    csrhodes authored
    	Add support for Solaris 9
    	... MAP_ANON is the same as in Solaris 8 (not that surprising).
    	Mention one or two other bits of NEWS
Commits on Jul 16, 2002
  1. @csrhodes

    0.7.5.13:

    csrhodes authored
    	Linux floating point fixes
    	... define an os_restore_fp_control() function in the runtime, and
    		use it in signal handlers (protected by #ifdef
    		LISP_FEATURE_LINUX in non-Linux-specific code)
    	... write useful definitions of it
    		... for x86 (probably correct)
    		... for ppc (works, but could do with fixing)
    	... delete a SET_FPU_CONTROL_WORD from initialization
    		... dunno why it was there; no observable symptoms. Shout
    			if some early x86/Linux system no longer works.
    	... export to C (via sbcl.h) Lisp's view of the floating point
    		control word
    	... delete some floating-point-related stale symbols and comments
    		from package-data-list.lisp-expr
    	Now floating point stuff mostly works (I think) on x86/Linux and
    		ppc/Linux, as well as sparc/SunOS. Other platforms still
    		probably don't work.
  2. 0.7.5.12:

    William Harold Newman authored
    	fixed (LOAD "foo.fasl" :IF-DOES-NOT-EXIST NIL) bug reported
    		by Gregory Wright sbcl-devel 2002-07-15
    	found bug 188. I hope entomotomy will have some logic to avoid
    		this kind of stupid fix-a-reported-bug-report-a-bug
    		primitivity!
Commits on Jul 15, 2002
  1. 0.7.5.11:

    William Harold Newman authored
    	finished translating the FFI chapter of the manual from LaTeX
    		to valid (though extremely clunky) DocBook up to date
    		(I hope) w.r.t. current SBCL behavior
    	made *SOFT-HEAP-LIMIT* default not NIL (after my application
    		overflowed the heap)
    	deleted bug 96 because it was implicitly a reference to the
    		old default implementation of TRACE in terms of
    		breakpointing the target function, and no longer
    		really applies now that the wrapper-based TRACE is the
    		default. TRACE still has the bug 67 problem which
    		keeps it from being used on functions which play a
    		role in the implementation of TRACE, but no longer
    		has a problem with functions just because they're
    		compiled into a particular physical representation.
Commits on Jul 11, 2002
  1. @csrhodes

    0.7.5.10:

    csrhodes authored
    	Begin to fix floating point exceptions on PPC
    	... write an os_context_fp_control function
    	... use it in CONTEXT-FLOATING-POINT-MODES
    	(this, unlike on the SPARC and Alpha ports, doesn't fix all the
    		normal-user problems.  After the first return from the Lisp
    		debugger on the PPC, the FLOATING-POINT-MODES are (almost)
    		cleared, which means that the second bad floating point
    		operation is not trapped).
  2. 0.7.5.9:

    William Harold Newman authored
    	&REST error-checking fix (cribbed from rtoy's post to
    		cmucl-imp today)
Something went wrong with that request. Please try again.