mit-scheme.git
35 years agoMajor change:
Guillermo J. Rozas [Sun, 1 Apr 1990 22:31:56 +0000 (22:31 +0000)]
Major change:

- New drifting/undrifting code and closure analysis.

Minor changes:

- lap:make-uncoditional-branch now returns a sequence of instructions,
rather than a single instruction.

- Uses of `expt' in machin and lapgen have temporarily been commented
out because the `usual-definition' declaration does not work when
`compile-by-procedures?' is turned on.

- letrec-value now gives only a warning when binding a trivial
closure.

- close-procedure! in blktyp now only gives a warning when a
non-trivial closure becomes trivial.

35 years agoSubstitute constants for expressions computed with `expt'. The expt
Guillermo J. Rozas [Sun, 1 Apr 1990 22:28:28 +0000 (22:28 +0000)]
Substitute constants for expressions computed with `expt'.  The expt
constant-folding declaration is currently ignored (because of
compile-by-procedures? breaking compiler declarations) and since the
values are integrated, they were being computed in a zillion places.

35 years agolap:make-unconditional-branch now returns a sequence rather than a
Guillermo J. Rozas [Sun, 1 Apr 1990 22:26:21 +0000 (22:26 +0000)]
lap:make-unconditional-branch now returns a sequence rather than a
single instructions.  On machines with branch delay slots and no
nullification, a nop must be returned in addition to the branch
instruction.

35 years agolap:make-unconditional-branch now returns a sequence rather than a
Guillermo J. Rozas [Sun, 1 Apr 1990 22:26:01 +0000 (22:26 +0000)]
lap:make-unconditional-branch now returns a sequence rather than a
single instructions.  On machines with branch delay slots and no
nullification, a nop must be returned in addition to the branch
instruction.

35 years agoletrec-value no longer errors if a binding is a trivial closure.
Guillermo J. Rozas [Sun, 1 Apr 1990 22:24:35 +0000 (22:24 +0000)]
letrec-value no longer errors if a binding is a trivial closure.
There are perfectly legal programs where this happens, although they
are written in poor style.  It now just gives a warning.

35 years agoNth revision of this code.
Guillermo J. Rozas [Sun, 1 Apr 1990 22:23:16 +0000 (22:23 +0000)]
Nth revision of this code.

Improve the closing and undrifting code significantly.  Many of the
procedures that were (needlessly) closed by the previous version are
no longer closed.

Undrifting constraints are computed more precisely.

Additionally, closing checks connectivity of the environment chain
(generating undrifting constraints as necessary) to ensure that all
the free variables captured are in fact available, and it also ensures
that disowned children (spliced out by the drifting code) are
undrifted if they depend on the closed-over variables.

One potential non-improvement introduced:  Undrifting any procedure
forces its compatibility class to close, in order to avoid a harder
analysis.  This is a cop-out for now.

35 years agoGet rid of funny drifting rules. Let-like procedures and others that
Guillermo J. Rozas [Sun, 1 Apr 1990 22:19:41 +0000 (22:19 +0000)]
Get rid of funny drifting rules.  Let-like procedures and others that
were previously not allowed to drift are now allowed to.  The
undrifting code takes care of them.

35 years agoIt's OK for a non-trivial closure to become trivial, not the other way
Guillermo J. Rozas [Sun, 1 Apr 1990 22:18:32 +0000 (22:18 +0000)]
It's OK for a non-trivial closure to become trivial, not the other way
around.  The previous error is now split into a warning (in the
unexpected but bad case), and an error (in the bad case).

35 years agoFix bch system to run on machines with floating alignment.
Guillermo J. Rozas [Sun, 1 Apr 1990 20:32:02 +0000 (20:32 +0000)]
Fix bch system to run on machines with floating alignment.

35 years agoInitial revision
Guillermo J. Rozas [Sun, 1 Apr 1990 20:19:57 +0000 (20:19 +0000)]
Initial revision

35 years agoInitial revision
Guillermo J. Rozas [Sun, 1 Apr 1990 20:14:51 +0000 (20:14 +0000)]
Initial revision

35 years agoAdd MIPS compiler support.
Guillermo J. Rozas [Sun, 1 Apr 1990 20:10:29 +0000 (20:10 +0000)]
Add MIPS compiler support.

35 years agoInitial revision
Guillermo J. Rozas [Wed, 28 Mar 1990 22:05:00 +0000 (22:05 +0000)]
Initial revision

35 years agoAdd new method for finding variable that is used by stack-overwrite
Guillermo J. Rozas [Wed, 28 Mar 1990 06:11:39 +0000 (06:11 +0000)]
Add new method for finding variable that is used by stack-overwrite
mechanism.  This new method does not use any tricks like variable
indirection to find the value, but produces a locative for that
particular variable.

35 years agoUndrifting constraints must propagate transitively to free callers.
Guillermo J. Rozas [Wed, 28 Mar 1990 06:07:59 +0000 (06:07 +0000)]
Undrifting constraints must propagate transitively to free callers.
Undrifted procedures must be re-analyzed for Exporting.

35 years agoMove switch.scm back to compiler/base, so it can be shared.
Guillermo J. Rozas [Tue, 27 Mar 1990 00:50:00 +0000 (00:50 +0000)]
Move switch.scm back to compiler/base, so it can be shared.
Update to match 68k version.

35 years agoAdd compiler:noisy?
Guillermo J. Rozas [Mon, 26 Mar 1990 23:46:08 +0000 (23:46 +0000)]
Add compiler:noisy?
Move switch.scm back to compiler/base, so it can be shared.
Update to match 68k version.

35 years agoAdd compiler:noisy?
Guillermo J. Rozas [Mon, 26 Mar 1990 23:45:38 +0000 (23:45 +0000)]
Add compiler:noisy?

35 years agoAdd compiler:noisy?
Guillermo J. Rozas [Mon, 26 Mar 1990 23:45:05 +0000 (23:45 +0000)]
Add compiler:noisy?

35 years agoMove switch.scm back to compiler/base, so it can be shared.
Guillermo J. Rozas [Mon, 26 Mar 1990 23:36:42 +0000 (23:36 +0000)]
Move switch.scm back to compiler/base, so it can be shared.
Update to match 68k version.

35 years agoMove switch.scm back to compiler/base, so it can be shared.
Guillermo J. Rozas [Mon, 26 Mar 1990 23:27:32 +0000 (23:27 +0000)]
Move switch.scm back to compiler/base, so it can be shared.

35 years agoAdd sf:noisy? flag, and syntax&integrate global procedure.
Guillermo J. Rozas [Mon, 26 Mar 1990 20:49:53 +0000 (20:49 +0000)]
Add sf:noisy? flag, and syntax&integrate global procedure.

35 years agoModify gc-statistic/meter so that gcnote will not be confused.
Guillermo J. Rozas [Mon, 26 Mar 1990 19:42:44 +0000 (19:42 +0000)]
Modify gc-statistic/meter so that gcnote will not be confused.

35 years agoAdd gc-timestamp for data structures that need to be updated after
Guillermo J. Rozas [Mon, 26 Mar 1990 19:38:43 +0000 (19:38 +0000)]
Add gc-timestamp for data structures that need to be updated after
gcs, but not necessarily after each gc.

35 years agoAdd arities to primitive definitions so that the file will syntax and
Guillermo J. Rozas [Sat, 24 Mar 1990 19:14:13 +0000 (19:14 +0000)]
Add arities to primitive definitions so that the file will syntax and
load in systems with no process primitives.

35 years agoFix register*constant methods for plus-fixnum and minus-fixnum.
Guillermo J. Rozas [Sat, 24 Mar 1990 13:22:42 +0000 (13:22 +0000)]
Fix register*constant methods for plus-fixnum and minus-fixnum.
When adding 0 with overflow tests, they were not loading the target.

35 years agoWhen accessing `block-procedure', make sure it is a procedure before
Chris Hanson [Wed, 21 Mar 1990 02:12:51 +0000 (02:12 +0000)]
When accessing `block-procedure', make sure it is a procedure before
using it as such; it might be an expression instead.

35 years agoFix unbound variable in `delete-process'. The `process/to-port' is an
Chris Hanson [Fri, 16 Mar 1990 22:43:33 +0000 (22:43 +0000)]
Fix unbound variable in `delete-process'.  The `process/to-port' is an
output port, not an input port, so use the appropriate operation to
close it.  Also do some slight reformatting.

35 years ago*** empty log message ***
Chris Hanson [Wed, 14 Mar 1990 21:06:31 +0000 (21:06 +0000)]
*** empty log message ***

35 years agoIn `decode-ordinary-floating-instruction', always specify at least two
Chris Hanson [Wed, 14 Mar 1990 21:06:06 +0000 (21:06 +0000)]
In `decode-ordinary-floating-instruction', always specify at least two
operands in the result, because the assembler does not understand the
one-operand syntax.

35 years agoIn `two-arg-register-operation', don't delete reused alias of first
Chris Hanson [Wed, 14 Mar 1990 21:05:05 +0000 (21:05 +0000)]
In `two-arg-register-operation', don't delete reused alias of first
operand until the second operand's reference has been computed.  This
prevents the reused alias from being mistakenly re-allocated for use
by the second operand.

35 years ago*** empty log message ***
Chris Hanson [Tue, 13 Mar 1990 00:27:17 +0000 (00:27 +0000)]
*** empty log message ***

35 years agoConditionalize floating-point register load instructions to give
Chris Hanson [Tue, 13 Mar 1990 00:20:45 +0000 (00:20 +0000)]
Conditionalize floating-point register load instructions to give
source type if loading from memory.

35 years agoConditionalize binary floating-point open-coded operations on whether
Chris Hanson [Mon, 12 Mar 1990 23:20:01 +0000 (23:20 +0000)]
Conditionalize binary floating-point open-coded operations on whether
the source is a register or not.

35 years agoRemove duplicate entry for fp6 from the list of initially available
Chris Hanson [Mon, 12 Mar 1990 21:38:18 +0000 (21:38 +0000)]
Remove duplicate entry for fp6 from the list of initially available
registers.

35 years agoAdd sun4 processor type.
Chris Hanson [Thu, 8 Mar 1990 19:26:43 +0000 (19:26 +0000)]
Add sun4 processor type.

35 years agoAdd ^B ^X ^U ^G to the ^C handler.
Guillermo J. Rozas [Mon, 5 Mar 1990 21:05:19 +0000 (21:05 +0000)]
Add ^B ^X ^U ^G to the ^C handler.

35 years agoAdded subprocess support.
Mark Friedman [Fri, 2 Mar 1990 20:48:01 +0000 (20:48 +0000)]
Added subprocess support.

35 years agoAdded subprocess support.
Mark Friedman [Fri, 2 Mar 1990 20:24:46 +0000 (20:24 +0000)]
Added subprocess support.

35 years agoAdded support for subprocesses.
Mark Friedman [Tue, 27 Feb 1990 19:53:29 +0000 (19:53 +0000)]
Added support for subprocesses.

35 years agoInitial revision
Mark Friedman [Tue, 27 Feb 1990 19:45:28 +0000 (19:45 +0000)]
Initial revision

35 years agoAdded support for subprocesses.
Mark Friedman [Tue, 27 Feb 1990 19:44:26 +0000 (19:44 +0000)]
Added support for subprocesses.

35 years agoAdded support for subprocesses to OPEN-CHANNEL-WRAPPER.
Mark Friedman [Tue, 27 Feb 1990 19:39:34 +0000 (19:39 +0000)]
Added support for subprocesses to OPEN-CHANNEL-WRAPPER.

35 years agoAdded support for unix subprocesses.
Mark Friedman [Tue, 27 Feb 1990 19:18:26 +0000 (19:18 +0000)]
Added support for unix subprocesses.

35 years agoAdded compiler flag ENABLE_SUBPROCESSES.
Mark Friedman [Tue, 27 Feb 1990 18:44:04 +0000 (18:44 +0000)]
Added compiler flag ENABLE_SUBPROCESSES.

35 years ago*** empty log message ***
Chris Hanson [Sun, 25 Feb 1990 23:57:57 +0000 (23:57 +0000)]
*** empty log message ***

35 years ago*** empty log message ***
Chris Hanson [Sun, 25 Feb 1990 23:23:57 +0000 (23:23 +0000)]
*** empty log message ***

35 years agoOpen-code vector-8b ref and set operations.
Chris Hanson [Sat, 24 Feb 1990 04:01:42 +0000 (04:01 +0000)]
Open-code vector-8b ref and set operations.

35 years agoDon't use arithmetic right shift for division.
Chris Hanson [Fri, 23 Feb 1990 21:43:42 +0000 (21:43 +0000)]
Don't use arithmetic right shift for division.

35 years agoEliminate some potential name conflicts in the expansion of a
Chris Hanson [Fri, 23 Feb 1990 18:47:56 +0000 (18:47 +0000)]
Eliminate some potential name conflicts in the expansion of a
`define-structure' macro.

35 years agoAdd more handlers for INTERNAL-APPLY-VAL types. Somehow some were
Guillermo J. Rozas [Wed, 21 Feb 1990 23:24:25 +0000 (23:24 +0000)]
Add more handlers for INTERNAL-APPLY-VAL types.  Somehow some were
missed last time around.

35 years agoMake debugger evaluation commands print out more like normal
Guillermo J. Rozas [Tue, 20 Feb 1990 16:15:16 +0000 (16:15 +0000)]
Make debugger evaluation commands print out more like normal
evaluations in a rep-loop.

35 years agoAllow strings as arguments to generate-uninterned-symbol.
Guillermo J. Rozas [Tue, 20 Feb 1990 15:58:32 +0000 (15:58 +0000)]
Allow strings as arguments to generate-uninterned-symbol.

35 years agoAdd definition of `alist?' that detects circularity.
Chris Hanson [Wed, 14 Feb 1990 01:56:12 +0000 (01:56 +0000)]
Add definition of `alist?' that detects circularity.

35 years agoFix definition of list? to terminate even in the presence of circular
Guillermo J. Rozas [Wed, 14 Feb 1990 00:20:38 +0000 (00:20 +0000)]
Fix definition of list? to terminate even in the presence of circular
structure, as required by the standard.

35 years agoInitialize `Prev_Restore_History_Stacklet' to NULL rather than
Chris Hanson [Tue, 13 Feb 1990 16:11:07 +0000 (16:11 +0000)]
Initialize `Prev_Restore_History_Stacklet' to NULL rather than
SHARP_F.

35 years agoAdd \n\ to newlines in comments.
Chris Hanson [Tue, 13 Feb 1990 16:00:20 +0000 (16:00 +0000)]
Add \n\ to newlines in comments.

35 years agoFix minor bug in last change.
Chris Hanson [Mon, 12 Feb 1990 23:07:31 +0000 (23:07 +0000)]
Fix minor bug in last change.

35 years agoChanges from JMiller that are needed to get this stuff running on the
Chris Hanson [Sun, 11 Feb 1990 22:34:12 +0000 (22:34 +0000)]
Changes from JMiller that are needed to get this stuff running on the
DECStation 3100.

35 years agoAdd new hash-table abstraction; this is loaded by
Chris Hanson [Sat, 10 Feb 1990 23:45:11 +0000 (23:45 +0000)]
Add new hash-table abstraction; this is loaded by

(load-option 'hash-table)

This abstraction is very similar to that of Common Lisp.

35 years agoInitial revision
Chris Hanson [Sat, 10 Feb 1990 23:43:09 +0000 (23:43 +0000)]
Initial revision

35 years agoDefine `symbol-hash-mod'.
Chris Hanson [Fri, 9 Feb 1990 19:11:24 +0000 (19:11 +0000)]
Define `symbol-hash-mod'.

35 years agoRun scheme-mode-hook before scheme-interaction-mode-hook.
Chris Hanson [Fri, 9 Feb 1990 00:57:43 +0000 (00:57 +0000)]
Run scheme-mode-hook before scheme-interaction-mode-hook.

35 years ago"fixnum.o" should depend on "mul.c", not "os.o".
Chris Hanson [Thu, 8 Feb 1990 02:50:03 +0000 (02:50 +0000)]
"fixnum.o" should depend on "mul.c", not "os.o".

35 years agoAdd changes to support gcc on the vax.
Chris Hanson [Thu, 8 Feb 1990 00:39:02 +0000 (00:39 +0000)]
Add changes to support gcc on the vax.

35 years agoFix cold-load problems.
Chris Hanson [Thu, 8 Feb 1990 00:36:31 +0000 (00:36 +0000)]
Fix cold-load problems.

35 years agoFix cold-load problems.
Chris Hanson [Thu, 8 Feb 1990 00:05:26 +0000 (00:05 +0000)]
Fix cold-load problems.

35 years ago*** empty log message ***
Chris Hanson [Wed, 7 Feb 1990 23:26:29 +0000 (23:26 +0000)]
*** empty log message ***

35 years agoImplement new record proposal.
Chris Hanson [Wed, 7 Feb 1990 23:25:58 +0000 (23:25 +0000)]
Implement new record proposal.

35 years agoInitial revision
Guillermo J. Rozas [Wed, 7 Feb 1990 21:33:20 +0000 (21:33 +0000)]
Initial revision

35 years agoAdd in sun4 (sparc) descriptions for Morry Katz.
Guillermo J. Rozas [Wed, 7 Feb 1990 21:31:56 +0000 (21:31 +0000)]
Add in sun4 (sparc) descriptions for Morry Katz.

35 years ago*** empty log message ***
Panayotis Skordos [Tue, 6 Feb 1990 22:04:04 +0000 (22:04 +0000)]
*** empty log message ***

35 years agoChanges by JMiller to make it work under HP cc and Sun cc (PCC?).
Guillermo J. Rozas [Tue, 6 Feb 1990 16:22:17 +0000 (16:22 +0000)]
Changes by JMiller to make it work under HP cc and Sun cc (PCC?).
Structure returns are different for those two compilers and GCC.

35 years agoChanges by JMiller to make it work under HP cc and Sun cc (PCC?).
Guillermo J. Rozas [Tue, 6 Feb 1990 16:15:20 +0000 (16:15 +0000)]
Changes by JMiller to make it work under HP cc and Sun cc (PCC?).
Structure returns are different for those two compilers and GCC.

35 years ago* Add new operations to register allocator that determine whether
Chris Hanson [Fri, 2 Feb 1990 18:40:04 +0000 (18:40 +0000)]
* Add new operations to register allocator that determine whether
allocation will cause unloading or spilling of registers.  Change the
operation `standard-register-reference' to use these operations when
deciding whether or not to refer to a register's home rather than
allocation a new alias for it.

* For an inline-coded procedure (e.g. LET), create an association
between each parameter of that procedure and the FG node that supplies
that parameter's value in the call.  This association is used to
optimize the initialization of variables that will be stored in cells:
the cell for such a variable is created during the call rather than
after it.  Previously, a stack-allocated parameter was initialized by
pushing its initial value, and then the contents of the stack location
were removed, placed in a new cell, and the cell stored back into the
stack location.  Now, the parameter's value is wrapped in a cell
before being pushed.

* RTL output files have been changed to print uninterned symbols by
name.

* The code generated for assignments in value position has been
slightly changed to guarantee the correct order of events.
Previously, the order of the computation of the new value and the
fetching of the old value was indeterminate; now it is guaranteed that
the new value is computed before the old value is fetched.

* The bit-string representation of register sets has been restored.
This has a time penalty for small register sets, but guarantees that
access to the register sets is independent of the number of registers.
Certain programs with large numbers of registers were being
unreasonably penalized by the list-based representation.  Also, the
dependencies for the file "rtlbase/regset" were adjusted to reflect
the files that actually refer to it.

* The RTL generated for cached variable assignments has been changed
to precompute the value of the assignment and store it in a pseudo
register.  Previously, the code was replicated.

35 years agoNew definition of `complex:acos' preserves more precision and is more
Chris Hanson [Fri, 2 Feb 1990 17:25:50 +0000 (17:25 +0000)]
New definition of `complex:acos' preserves more precision and is more
efficient.  This one also does not introduce inexact real or imaginary
zero components in certain (important) boundary cases, such as:

(acos 2)
;Value: 0.+1.3169578969248166i

35 years agoFix bug in fasdump by which zero length files were left around when
Guillermo J. Rozas [Wed, 31 Jan 1990 05:01:53 +0000 (05:01 +0000)]
Fix bug in fasdump by which zero length files were left around when
the fasdump failed.

35 years agoFix bug in load/internal by which empty files caused an error in the
Guillermo J. Rozas [Wed, 31 Jan 1990 02:03:13 +0000 (02:03 +0000)]
Fix bug in load/internal by which empty files caused an error in the
fasl check.  peek-char returned an eof-object, and char->ascii barfed
on that.

Now, if the file is empty, it is loaded as a source file, whatever
that means.

35 years agoFix bug in Prepare_Apply_Interrupt noticed after last set of changes
Guillermo J. Rozas [Tue, 30 Jan 1990 14:44:25 +0000 (14:44 +0000)]
Fix bug in Prepare_Apply_Interrupt noticed after last set of changes
(addition of RC_INTERNAL_APPLY_VAL).

35 years agoFix bug in Prepare_Apply_Interrupt noticed after last set of changes
Guillermo J. Rozas [Tue, 30 Jan 1990 14:35:28 +0000 (14:35 +0000)]
Fix bug in Prepare_Apply_Interrupt noticed after last set of changes
(addition of RC_INTERNAL_APPLY_VAL).

35 years agoAdd the INTERNAL-APPLY-VAL return code.
Guillermo J. Rozas [Mon, 29 Jan 1990 22:35:32 +0000 (22:35 +0000)]
Add the INTERNAL-APPLY-VAL return code.

35 years agoAdd the INTERNAL-APPLY-VAL return code. It replaces the procedure
Guillermo J. Rozas [Mon, 29 Jan 1990 22:32:57 +0000 (22:32 +0000)]
Add the INTERNAL-APPLY-VAL return code.  It replaces the procedure
with the current value of Val and then proceeds to apply.  Apply
errors use this return code rather than INTERNAL-APPLY to restart.
Under normal operation there is no change since a POP-RETURN-ERROR
return code (which restores Val) is pushed as well, so to make use of
this feature the top few frames of the stack must be eliminated before
invoking the continuation.

35 years agoAdd the INTERNAL-APPLY-VAL return code.
Guillermo J. Rozas [Mon, 29 Jan 1990 21:30:36 +0000 (21:30 +0000)]
Add the INTERNAL-APPLY-VAL return code.

35 years agoInitial revision
Guillermo J. Rozas [Mon, 29 Jan 1990 21:29:09 +0000 (21:29 +0000)]
Initial revision

35 years agoAdd the INTERNAL-APPLY-VAL return code.
Guillermo J. Rozas [Mon, 29 Jan 1990 21:14:15 +0000 (21:14 +0000)]
Add the INTERNAL-APPLY-VAL return code.

35 years agoInitial revision
Guillermo J. Rozas [Thu, 25 Jan 1990 16:46:40 +0000 (16:46 +0000)]
Initial revision

35 years agoNew compiler port.
Guillermo J. Rozas [Thu, 25 Jan 1990 16:45:49 +0000 (16:45 +0000)]
New compiler port.

35 years agoInitial revision
Guillermo J. Rozas [Thu, 25 Jan 1990 16:37:05 +0000 (16:37 +0000)]
Initial revision

35 years agoNew compiler port.
Guillermo J. Rozas [Thu, 25 Jan 1990 16:21:29 +0000 (16:21 +0000)]
New compiler port.

35 years agoMake setup-image do what the comment at the top of the file says when
Guillermo J. Rozas [Wed, 24 Jan 1990 16:45:07 +0000 (16:45 +0000)]
Make setup-image do what the comment at the top of the file says when
identify is #f.

35 years ago`hash-table-copy', when copying the elements in a hash bucket, was
Chris Hanson [Tue, 23 Jan 1990 22:44:12 +0000 (22:44 +0000)]
`hash-table-copy', when copying the elements in a hash bucket, was
copying each element and then setting its previous-hash link to that
of the original element rather than its copy.  Don't entirely
understand why this wasn't noticed before.

35 years agoConditionalize constant space sizes for the Spectrum.
Guillermo J. Rozas [Tue, 23 Jan 1990 08:35:09 +0000 (08:35 +0000)]
Conditionalize constant space sizes for the Spectrum.
Fix floating alignment bugs.
Fix bchdmp bug.

35 years agoFix floating alignment.
Guillermo J. Rozas [Tue, 23 Jan 1990 08:30:51 +0000 (08:30 +0000)]
Fix floating alignment.

35 years agoAdd floating alignment back.
Guillermo J. Rozas [Tue, 23 Jan 1990 08:30:29 +0000 (08:30 +0000)]
Add floating alignment back.

35 years agoFix bug by which current_disk_position was not reset from the last
Gerald Jay Sussman [Tue, 23 Jan 1990 03:00:23 +0000 (03:00 +0000)]
Fix bug by which current_disk_position was not reset from the last
gc-like operation.  When the files were swapped for fasdump, the
system might not lseek thinking that the pointer into the file was at
the correct place while it doesn't.  Typical failure would be
"Could not read into the scan buffer" since the file was shorter than
expected.

35 years agoMake changes to use more space-efficient representation for dbg-label
Chris Hanson [Mon, 22 Jan 1990 23:45:31 +0000 (23:45 +0000)]
Make changes to use more space-efficient representation for dbg-label
objects.  These changes require runtime version 14.72 or later.

35 years agoAdd new representation for `dbg-label', that is optimized for minimum
Chris Hanson [Mon, 22 Jan 1990 23:41:39 +0000 (23:41 +0000)]
Add new representation for `dbg-label', that is optimized for minimum
space consumption.  Make the dbg-label operations generic so that they
handle both the old and new representations.  By default, the
constructor makes the old representation.  These changes are
upwards-compatible for old programs and old versions of the compiler.

35 years agoChange `operation/open' to handle #F as a result from
Chris Hanson [Mon, 22 Jan 1990 23:39:37 +0000 (23:39 +0000)]
Change `operation/open' to handle #F as a result from
`starbase-open-device'.

35 years agoMake gr2 be compile code's val register.
Guillermo J. Rozas [Mon, 22 Jan 1990 22:33:22 +0000 (22:33 +0000)]
Make gr2 be compile code's val register.
Fix trampoline_to_interface.
Fix the number of NOPs before hook_jump_table.
Fix the closure code.
Fix various comments.

35 years agoFix hppa_extract_absolute_address, hppa_store_absolute_address,
Guillermo J. Rozas [Mon, 22 Jan 1990 22:31:26 +0000 (22:31 +0000)]
Fix hppa_extract_absolute_address, hppa_store_absolute_address,
CLOSURE_SKIPPED_CHECK_OFFSET, and various comments.