Chris Hanson [Sun, 30 Aug 1998 02:43:59 +0000 (02:43 +0000)]
Fix two bugs in command completion: (1) completions list contained
wrong entries in some circumstances, and (2) command-completion
procedure not returning #T in all circumstances that it did
completion.
Chris Hanson [Sun, 30 Aug 1998 01:52:39 +0000 (01:52 +0000)]
Redefine COMINT-DYNAMIC-COMPLETE so that it tries a list of completion
procedures, as in Emacs 19. Fix bug in COMINT-LINE-START: used
current value of COMINT-PROMPT-REGEXP rather than the one for the
buffer being examined.
Chris Hanson [Sun, 30 Aug 1998 01:50:29 +0000 (01:50 +0000)]
Erase message when inserting completion. That way,
STANDARD-COMPLETION consistently writes something in the message area,
eradicating any previous message.
Chris Hanson [Sat, 25 Jul 1998 05:46:56 +0000 (05:46 +0000)]
Explicitly code the Linux superblock magic constants in this file.
The previous trick of reading various kernel header files no longer
works when linking with glibc2, because there are symbol conflicts
between the kernel and glibc2.
Chris Hanson [Mon, 20 Jul 1998 04:08:23 +0000 (04:08 +0000)]
Must call mprotect to set the "execute" bit on Scheme's memory, so
compiled code can run. Also, implement mechanism to use mprotect to
flush the cache (which is disabled by default).
Chris Hanson [Sun, 19 Jul 1998 20:41:30 +0000 (20:41 +0000)]
More recent MIPS processors (and the corresponding assemblers) require
that the stack pointer always be a multiple of 8. Change the code to
make that so. Also, work around an assembler bug having to do with
".globl" and the "la" pseudo-instruction.
Chris Hanson [Sun, 19 Jul 1998 20:40:10 +0000 (20:40 +0000)]
Must call mprotect to set the "execute" bit on Scheme's memory, so
compiled code can run. Also, implement mechanism to use mprotect to
flush the cache (which is disabled by default).
Chris Hanson [Sun, 21 Jun 1998 09:05:17 +0000 (09:05 +0000)]
If NNTP server closes connection, attempt to notice this and recover
by reopening the connection. Previously, we just signalled an error
and the user had to manually reopen. Problem: the specific error code
to be returned by the server in this situation is not specified in the
RFC, so we must determine the value empirically.
Chris Hanson [Sun, 7 Jun 1998 08:18:13 +0000 (08:18 +0000)]
Generalize code that refers to process mark so that it can be used in
an inferior-repl buffer. In coming-flush-output, don't flush output
unless there is some.
Chris Hanson [Mon, 1 Jun 1998 05:49:43 +0000 (05:49 +0000)]
Use in-buffer prompt to handle errors in the REPL buffer. The
modeline prompt was very intrusive and couldn't be aborted out of.
The new mechanism reads input exactly as the in-buffer debugger does,
and can be aborted.
Chris Hanson [Tue, 5 May 1998 02:15:08 +0000 (02:15 +0000)]
Fix bugs in handling of ENVIRONMENT-ASSIGNABLE? and
ENVIRONMENT-ASSIGN! that are the analogs of the bug in
ENVIRONMENT-LOOKUP that was fixed in the previous revision.
Chris Hanson [Tue, 5 May 1998 00:24:29 +0000 (00:24 +0000)]
Fix bug: ENVIRONMENT-LOOKUP did not look at parent environments if the
environment being operated on was a compiled-code environment.
Instead, it bombed with a type error.
Chris Hanson [Thu, 30 Apr 1998 22:15:43 +0000 (22:15 +0000)]
Don't convert package names to environments when binding the
SCHEME-ENVIRONMENT buffer variable -- this is a screw when reloading
packages. Also, use new ENVIRONMENT-SYNTAX-TABLE to determine the
default syntax table to be used when evaluating.
Chris Hanson [Sat, 18 Apr 1998 05:40:42 +0000 (05:40 +0000)]
Add primitive WIN32-VIRTUAL-QUERY, which is a direct interface to the
Win32 VirtualQuery API call. This is very useful for examining
Scheme's address space.
Chris Hanson [Sat, 18 Apr 1998 05:39:44 +0000 (05:39 +0000)]
Add mechanism to reserve the largest available block of memory below
0x04000000 at the earliest possible moment. This maximizes the amount
of memory available for Scheme's heap. Unfortunately, it still isn't
very much space compared to other platforms, and under Visual C++ it
isn't even as much as that provided by Watcom C/C++. I don't know why
this is so poor, or whether there is anything to be done about it.
Chris Hanson [Sun, 8 Mar 1998 07:26:25 +0000 (07:26 +0000)]
Change interface procedure DEBUG-SCHEME-ERROR so that it returns if
the user opts not to enter the debugger. Also, do a better job of
presenting the error message to the user when asking whether to start
the debugger.
Chris Hanson [Sat, 7 Mar 1998 08:54:02 +0000 (08:54 +0000)]
Tweak error-reporting mechanism to make it more reliable. Trouble is
that this isn't the right place for it -- it should be integrated into
the debugger interface.
Chris Hanson [Mon, 23 Feb 1998 05:37:47 +0000 (05:37 +0000)]
Add stupid mechanism to break cycles in reference graphs. This
doesn't try to do a good job -- it just breaks them at the point where
they are discovered. After having used this program for over a year,
the cycle I just ran across was the first, out of perhaps 100,000
messages or more, so this is an unusual occurrence.
Chris Hanson [Sun, 22 Feb 1998 08:48:02 +0000 (08:48 +0000)]
Eliminate BKPT macro in favor of procedure. The macro used
(THE-ENVIRONMENT), which screwed the compiler. The procedure uses the
environment of the first subproblem of the continuation of the call,
which in many cases is the same thing.
The effect of the old BKPT macro can still be achieved by writing