Debugger changes (was: Some changes..)
 

Hi Toni,

- can be added a 'memory block fill' command in debugger? (maybe there is but I've not found it..)
- command xx doesn't act the same from GUI and CLI: from CLI you can switch to GUI without reactivating emulation, not the vice versa (better w/o reactivating!).
- it's possible to add a command to directly jump to 'properties window' (F12) from debugger, to change some emulation property without neither code executed, and return back?

Thanks! :)
ross

It does not exist. I never needed it.

GUI is not supported so anything related to it is also unsupported :)

I thought debugger already had command to inject config changes (like uae-configuration) or input events but it seems it is still unimplemented.

GUI is only opened during emulated vblank so it isn't going to be instant.

Thanks for reply :)

It's useful, same as memcopy or even a little (raw) assembler..

Understandable.

Ok.


Bye!
ross

Fill and mem copy added. (check help message)

Assembler would be nice but I have no idea how to make it from CPU emulator tables. I won't accept separate external assembler, it must be based on "gencpu" tables.

EDIT: table68k should have enough information to create assembler, someday..

Added. I <custom event string> (for example "I show_leds=false")

A useful addition, thanks! :)

:great


Thanks Toni!

Cheers,
ross

Assembler seems to be very possible. Only need some annoying text parsing.

Warning: it probably will be a bit weird with some instructions, just like the disassembler..

This is a BIG news!

:)

Assembler is now partially done.

- two modes, immediate ("a address instruction") and interactive mode ("a address", return. empty line exits)
- "weird" instructions like MOVEM not yet supported (Weird syntax "MVMEL #mask,EA" and "MVMLE #mask,EA" do work..)
- 68020+ addressing modes not yet supported (scale factor will be implemented, not sure if I bother with other 68020+ modes)
- MMU and FPU instructions are not supported
- Only accepts original Motorola syntax even if disassembler uses 68020+ syntax only.. (Both will be supported later)
- Out of range values are truncated, no errors reported.
- ADD x,An is not supported, it needs to be ADDA (and so on)
- And more?

Test and report all issues.

:cool

how to exit from interactive mode?[edit]: empty is a 'space', maybe better a special char?

We need a glossary :)

..and bitfield instruction..

not really a prob.

Ok.

- shift/rotate do not work
- absolute addressing work but without $ sign
- Bcc only relative to start of 'interactive mode' address
- bit instruction do not work
(maybe they works but I did not find the syntax :))

This new feature is really useful!

Thanks Toni,
ross

Empty as in just press Return.

MOVEM will work normally later. It needs special handling. (It really is technically MOVEM #16-bit_immediate,EA)

Also have annoying encoding.. But these probably should be supported.

Lets check..

Hexadecimal is the default. (!val = decimal)

It is relative to current address, what else should it be? :)

Same

from CLI yes but not from GUI (i know, is unsupported ;), but a special char can solve both ambient)

is not relative to current, is relative to initial
(ex. suppose in interactive mode you insert various instructions and then a bcc, the branch is relative from the first instruction..)

mmh no i'm wrong :)

:great
ross

- "New" syntax (d16,An) and (d8,An,Xn) are now supported.
- Xn multiplier is supported (68020+ only)
- bit instruction/shifts with immediate value now work (any instruction that had immediate included with first opcode word and instruction was not somethingQ)

Note that btst and friends require .b or .l, it is not handled automatically. EDIT: automatic now.

EDIT2: ADD vs ADDA is now handled automatically.

You are too fast :)

All is working.
There is an exception, one of the most strange instruction in 68k ISA:
BTST dx,#immediate

:great

- BTST Dn,#x is now working (It didn't work because I forgot about that and I hardcoded destination immediate = word size instruction to make LINK working easily..)
- MOVEM is now sane :)

Awesome :)

I guess it is good enough now?

Yes! I've to deep test it but is great.

Thanks!!!
ross

I felt quite good having checked in two commits of roughly twenty lines today. Until I saw this.

How do you do it, Toni?

It is just boring code..

Assembler is only couple of hundred lines of code and 2/3 of code is ugly text parsing. (Get source and possible destination opmode and other parameters)

Main code is more or less a big loop that iterates already existing opcode table, multiple times, trying to find single matching entry using parsed data. All the opcode data already existed, it only needed to be used to make an assembler.

Writing the code was the easy part, finding working logic was more difficult.