Work Remaining on SCP Version 3
                         ===============================


============
Current Work
============


General
-------

 - Integrate device cloning into SCP.

 - Integrate expanded MTABs into SCP.

 - After SET CONSOLE SERIAL, exiting does not close the serial line.  Exiting
   takes care of serial lines connected to multiplexers when the detach routine
   is called for each port (because they don't have UNIT_ATTABLE).  It's just
   the console that's an orphan.  Could change "tmxr_close_master" thus:

     if (lp->conn                                        /* if the line is connected */
       && (tmxr_is_extended == NULL                      /*   and the line */
       || tmxr_is_extended (lp) == FALSE                 /*     is not extended */
    +  || sim_switches & SIM_SW_SHUT))                   /*     or the simulator is exiting */
         tmxr_disconnect_line (lp);                      /*       then disconnect it */

   Mux serial ports would not be closed twice because they would be closed
   during the "detach_all" processing, so "lp->conn" would be zero.

 - Add numeric overrides to ASSERT.  Currently uses REG or address radix
   exclusively.  But EXAMINE and DEPOSIT allow overrides of these.

 - [ETAB] Validation and display routines receive unbiased value or unbiased bit
   position conversion.  For SET, bit conversion is done before call and bias is
   applied after return; for SHOW, bias is removed before call and conversion is
   undone after call.  For bit position, bias is simply a shift right for SET
   and a shift left for SHOW.

 + [ETAB] Note in the manual that SHOW <option> matches "mstring" for ETAB but
   "pstring" for XTAB.  Also note that the print routine is responsible for
   printing only if it returns -1.  Otherwise, the value is printed using the
   "pstring" field if the print routine returns SCPE_OK or there is no print
   routine.

 + [ETAB] If radix > 0 and max <= min, then SCPE_IERR (Internal error).

 + [ETAB] Add MTAB_ESO for show-only.  Consider show-only for bits 12-15 LR
   position with bias 12 -- needs min 12 max 15, so these can't be used to
   suppress SET.

 + [ETAB] Add MTAB_ENO for show-named-only.  Same as MTAB_NMO for extended.
   Consider "CPU Speed=x15" which does not want to print with SHOW CPU.

 + [ETAB] Limit flags field modifications to just the user flags part.

 + [ETAB] Validation routine for a device MTAB should have the "uptr" parameter
   point at unit 0 and "desc" point at the device.  This would be the same as an
   extended device MTAB.

 + [ETAB] Reserve the upper four bits of the "expander" field for user use with
   instance adjust routines to identify the address bases, if needed. Provides
   up to 16 base identifiers (n = 0-15) that can index into an array of base
   pointers.  Otherwise would have to use heuristics to determine base addresses
   in the "desc" field if there were more than one.  [ The upper 16 bits of the
   REG "flags" field is reserved for user flags, some of which can be put to the
   same use. ]

 + [ETAB] Add MTAB_EMK to pass the mask instead of the match in the validation
   "value" parameter.  Needed to validate clearing an option (currently passes
   zero -- the match value -- which isn't helpful).

 + [ETAB] Add MTAB_EBT to store the parsed numeric value as a bit in a bit field
   whose width and location are specified by the mask value.  Bits are numbered
   right-to-left unless MTAB_ELR is included, in which case bits are numbered
   left to right, with bit number 0 corresponding to the bit indicated by the
   maximum value in the match field.  The set bit is then left-shifted by the
   LSB_NUMBER of the mask value, masked, and merged.  MTAB_EBI could also be
   applied to bias the bit numbering, e.g., bits 6-3 are rendered as 3-0.
   Needed to set the INTMASK value.

 + [ETAB] A validation routine should return -1 to indicate that the value
   should not be stored with no error reported.  Needed if the validation
   routine wants to store the value itself.  Return SCPE_OK to store the value
   or any other SCPE error to skip storing with the error reported.

 + [ETAB] If a range parse fails, and a validation routine exists, do not return
   SCPE_ARG but instead call the routine, passing -1 as the value (which is not
   a possible value, as ranges are always positive).  If the routine doesn't
   like the "cptr" token, it can return SCPE_ARG, or it can authorize 0 to be
   stored in the deposit field with SCPE_OK, or it can store its own value in the
   field by returning -1 (t_stat is an int) to skip the SCP deposit and return
   success.  If the parse is good but the value is out of range, or if the parse
   is bad and there is no validation routine, then return SCPE_ARG.  [ Allows
   UNIT=0/1/2/3/OFF or INTMASK=0-15/D/E. ]

 + [ETAB] A print routine should return -1 to indicate that the value should not
   be printed with no error reported.  Needed if the print routine wants to
   print the value itself, e.g., if a range value is out of range.  Return
   SCPE_OK to print the value or any other SCPE error to skip storing with the
   error reported.

 + [ETAB] The "pstring" field for ranges should carry the equals sign if
   desired.  Otherwise, printing "unit 0", e.g., is not possible for a range.
   Sometimes an equals sign is proper, in which case the "pstring" field should
   supply it.

 - Alter calibrated timers so that initial calibration is not redone when the
   simulator stops for a moment (e.g., BREAK, EXAMINE, GO, or GO UNTIL, REPLY).
   V4 handles this by determining the wall-time interval between executions.

 - Does "sim_prog_name" need to be encoded?  Currenty, it is not, which means
   that backslashes in the path are "bare."  This means that a path, such as
   "\release\hp2100.exe" contains an escape ("\r") that will be interpreted
   as a CR if the string is used in a command that takes a quoted string
   (e.g., REPLY "%SIM_BIN_PATH%").  Is this a problem?  Note that for
   non-quoted-string use (e.g., "! %SIM_BIN_PATH%"), this works as expected.
   If we encode the string, must we then decode it for non-quoted-strings?

 - Doing SET CONSOLE PCHAR=1000423640 is very obscure.  It would be much nicer
   to do SET CONSOLE PCHAR=ESC;DC1;CR;LF;HT;BS;BEL;ENQ (and similarly for SHOW).

 - GCC Version: 11.3.0.
   scp.c: In function ‘spawn_cmd’:
   scp.c:916:1: warning: ignoring return value of ‘system’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
   916 | system (cptr);

   sim_console.c: In function ‘sim_ttrun’:
   sim_console.c:1054:5: warning: ignoring return value of ‘nice’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
   1054 | nice (10); /* try to lower pri */

   sim_console.c: In function ‘sim_ttcmd’:
   sim_console.c:1066:5: warning: ignoring return value of ‘nice’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
   1066 | nice (-10); /* try to raise pri */

   sim_console.c: In function ‘sim_os_putchar’:
   sim_console.c:1101:1: warning: ignoring return value of ‘write’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
   1101 | write (1, &c, 1);

   I1401/i1401_cd.c: In function ‘cdr_read_file’:
   I1401/i1401_cd.c:414:1: warning: ignoring return value of ‘fgets’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
   414 | fgets (buf, sz, cdr_unit.fileref); /* rd bin/char card */

   Needs -Wno-unused-result, as the opposite is now the default.


Supplement
----------



==============
Work Completed
==============

General
-------

 + Allow multiple SET ENV with quotes, e.g., SET ENV '<name>=<value>','...'
   like SET <dev> <name>=<value>,<name>=<value> or SET CONSOLE WRU,BRK,DEL.

 + Change my makefile to NOT turn warnings into errors for debug builds.
   Failures for unused variables, e.g., are annoying during development,
   especially when code is temporarily commented out for testing.

 + Implement expanded modifiers.

 + Implement device cloning.

 + Decide how to handle GetDefaultCommConfig failure.

 + Allow REALTIME values for BREAK/REPLY DELAY?  Currently, we have to pick
   some arbitrary number of instructions to allow time for I/O cards to turn
   around.  Maybe BREAK DELAY 15 MSEC would be better.

 + Change "GO UNTIL 15 MSEC" to "GO FOR 15 MSEC" for ease of parsing
   and better consistency.

 + Add a PAUSE command to give a timed suspension.  Needed for a platform-
   independent way to delay while waiting for a terminal emulator to start
   up and connect.

 + Add a FLUSH command to flush debug log.  Avoids having to CTRL+E twice
   and then CONTINUE just to examine the log.  FLUSH does console and debug
   logs, FLUSH <dev> also does all units on the specified device, and
   FLUSH ALL also does all units on all devices (in addition to the logs).

 + Add flush to mux lines too.  Must track mptrs when attached, so that we
   can locate TMLN arrays.  Note that FLUSH <unit> won't work with 2100 MUXL
   device, because TMXR is attached to MUX!

 + BUG: ATTACH MUXL 1052 attaches port to MUX device, even though MUXL device
   is specified.  Problem is "ex_tmxr_[at|de]tach_unit" attaches to supplied
   poll unit if device name specified but doesn't check that poll unit is on
   same device as passed unit.  Actually, calls from "muxl_attach" should
   pass NULL poll unit, and this should be taken as meaning "cannot attach
   if device name specified" rather than internal error as it is currently.

 + makefile must detect "clang" in pathname instead of direct comparison.
   use "notdir" and "findstring" functions.

 + Replace "tmxr_set_modem_control_passthru" and "tmxr_set_get_modem_bits"
   with "tmxr_control_line" and "tmxr_line_status" extensions that work with
   both Telnet and serial ports (former don't work with serial ports in 3.x).

 + Add hook to print name of unit in SHOW QUEUE (for string breakpoint timer,
   keyboard poll timer printed in "show_queue").

 + Always include "sim_extension.h", regardless of whether ".c" is present,
   and supply a second (different) one for Bob only that remaps.

 + Add IF...IN and IF...NOT IN, e.g.:

     if "%1" not in "","2100","21MX","7900","7905" set env ERROR=1

   Comparative expression is:

     IN <quoted-string> { , <quoted-string> ... }
     NOT IN <quoted-string> { , <quoted-string> ... }

 + Add a SIM_RUNNING substitution variable (0/1) that reflects whether the VM
   is running when the variable is referenced.  This would be only when
   concurrent mode is enabled and the DO was entered at the "scp>" prompt.
   This allows a DO file to determine if a final GO is needed to resume
   execution.

 + Allow multiple string breakpoints.

 + Eliminate parens around IF statement, change SIM_BIN_PATH to SIM_EXEC
   (4.x accommodations).

 + Allow unrestricted operation for a DO within a concurrent session, so that
   prompt/response can be done.  So DO must stop simulator and then execute DO
   command file.  Last command in file must be a CONTINUE, if that's what's
   wanted.  But we don't want to drop out if DO file doesn't exist.

   Could have "execute_file" test "sim_is_running" and, if set, simply return
   SCPE_EXEC if file exists and SCPE_OPENERR if not.  Then, in "[ex_]sim_instr",
   if "reason" is SCPE_EXEC, point "sim_brk_act" at command line buffer, clean
   up, and leave normally.  This unwinds RUN command.  Whether return is to
   command line or to nested DO, the breakpoint action will re-execute the DO
   command with "sim_is_running" clear.

   One catch: "execute_file" must make a copy of the incoming command line in
   a local buffer because "do_arg" points into this buffer.  Arguments therefore
   depend on the buffer not changing during DO execution, but if it points at
   "cmd_buf", that will get overwritten as soon as the next concurrent command
   is entered.

   Another catch: if last cmd is CONTINUE, command file remains open until
   the sim stops.  Better to do auto CONTINUE at EOF on concurrent DO?  Maybe
   DO from ex_run_cmd?

 + With the console in CONCURRENT mode and CPU tracing on, doing CTRL+E prints
   "simulation stop: (null)" in the trace log.  Cause is SCPE_EXEC (equal to
   SCPE_BASE - 1) status being interpreted, and there is no corresponding
   string in the "sim_stop_messages" array.  The trace should be skipped for
   an SCPE_EXEC return.  Wanted to define SCPE_EXEC as "SCPE_LAST + 1", but
   there isn't one.  Can't define as "SCPE_KFLAG + 1", as that would be seen
   as a keystroke rather than an abort status (because test is >= KFLAG).
   Might do SCPE_KFLAG - 1 or maybe SCPE_KFLAG >> 1.  Want to ensure that it
   resides in an area that will never be assgined but can be detected easily.

 + Don't need to define REF_NONE, etc. in hp_defs.h?  They're a part of SCP,
   regardless whether or not extensions are included!

 + Don't copy sim_ether.* into my distribution.

 + Set sim_quiet in DO files.  Override with DO -A.

 + Add SET <dev> BINARY command and -B option for binary EXAMINE / DEPOSIT.
   There is currently no easy way to do this via extensions without fully
   parsing the entire EX/DEP line (to temporarily change the target device
   data or register radix to 2 if -B is present) or putting REG_VMIO on
   every register of every device, so that the VM parse/print routines
   are called.  Unfortunately, the PDP simulators use -B for "byte access",
   but the binary options could be conditional on a USE_BINARY define.

 + Add a global "simh.ini" file that is unilaterally loaded.  4.x has this.
   Problems with sim-specific initializer files (e.g., "hp2100.ini" are that
   they are dependent on the executable name (so renaming the executable to,
   say, "hp2100d.exe" for a debug version demands an "hp2100d.ini" file),
   and if a file is supplied on the command line, the initializer is not run
   (so there's no place to put override commands that ALWAYS want to be
   processed, e.g., SET CONSOLE DEL=177).  This could be done as an extension
   by trapping the first DO and processing the "simh.ini" file before the
   indicated file (DO is always called at startup, either for the sim-specific
   file, or for the command-line file).

 + Should the default mode be concurrent or non-concurrent?  This would be
   less important if a global "simh.ini" file could override.

 + Add serial support to SET CONSOLE.

 + Entering QUIT in concurrent mode either quits execution (same as second
   CTRL+E) or maybe quits simulator (ends execution and puts QUIT in as
   "breakpoint" action).  Currently, I enter QUIT a lot and am greeted with
   Command not allowed.

 + Add ASSERT checking for CPU memory if no register specified.
   4.x allows symbols for address value, e.g., ASSERT 0=NOP
   (but NOT ASSERT 0=LDA 1, e.g. because it uses get_glyph for value!).
    - must trap ASSERT in hp3000_sys and set parse_config = apcBank_Offset.
    - exdep_addr_loop does test_search with sim_eval[0] only.
    - permit symbol for value?  if so, compare 1st addr values only?
      4.x does mask and comparisons across full sim_eval count.

 + Parse a quoted string (for escapes) for ECHO if leading quote?

 + Redo extensions as add-ons to standard 3.10 SCP (using hooks, etc.)

 + Strip old console halt stuff; use a string breakpoint table.

 + Need to keep close to 4.x syntax?
    - BREAK <string> { DELAY <delay> }, ...
    - REPLY <string> { DELAY <delay> }
    - BREAK DELAY <delay>
    - REPLY DELAY <delay>
    - GO 5 UNTIL "a" DELAY 5,"b" DELAY 10,"c"

 x Add an EQUATE command to substitute tokens (via get_glyph)?
   Would allow 4.x command files to work on 3.x without changes.

 + Add SCP command entry without stopping simulator.  With proper mode set,
   CTRL+E prints "scp>" prompt and sets flag to divert console input into
   a command buffer.  CR at end clears diversion and returns SCPE_STOP to drop
   out of command loop.  End of "run_cmd" sees command buffer and executes
   command there before looping back and calling "sim_instr" again.  Original
   entry into "run_cmd" clears command buffer, so that if no chars in buffer,
   then SCPE_STOP actually drops out of "run_cmd".  Dropping out of "scp>"
   prompt requires either ABORT command or maybe second CTRL+E after first.

 x Should concurrent DO save the string breakpoint context, i.e., save the
   current list and start a new one that's abandoned when DO completes?
   Problem is that a DO file might define a string breakpoint for its own
   use (e.g., MOUNT TAPE) that's already defined in the current execution.
   When the DO file completes, the original string break is lost.  Saving
   and restoring context is easy enough -- just save and restore "sb_head"
   -- but that would mean that a concurrent DO file can't set any string
   breakpoints that would persist after completing (unless any breakpoints
   left in the new context are added to the old one -- but watch for
   duplicates!).

   NO! The problem with this is that breakpoints in a DO file would behave
   differently, depending on whether concurrent mode was on or off.  Moreover,
   they'd behave differently from address breakpoints.

 x Might need to suppress (or reduce) the priority drop of the IOP instance
   when two CPU instances are running (i.e., for TSB).  The SPs of all TSBs
   have timeouts on IOP operations.  If the IOP instance is preempted, the
   SP may fail.  Ensuring that the IOP has priority over the SP means that
   the SP will be preempted if preemption of one process is needed.

   NO! Use instruction interlock instead.


Supplement
----------

 + New SIM_RUNNING substitution variable.

 + Multiple string breakpoints, same string replaces, NOBREAK "", NOBREAK ALL,
   "Break delay timer" in queue, it's not EXPECT.

 + No parens around IF statement.

 + IF...IN.

 + SIM_BIN_PATH changed to SIM_EXEC.

 + DATE_19XX_YYYY changed to DATE_RRRR (and DATE_RR).

 + Unrestricted operation for a DO within a concurrent session, DO behavior
   within a concurrent session, auto continue.

 + No 4.x accommodations.