one we'll use yet.
More layout issue fixes. When I changed the debugger fonts, everything
was redrawn correctly (just smaller). This is exactly what should
happen, and is a good first step to making the GUI be font-size aware.
Added change tracking to the CPU PS register (ToggleBitWidget).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@708 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Eventually, the debugger GUI will support multiple fonts, and the
layout will automatically resize itself.
(Hopefully) fixed bug whereby pressing '~' would not exit the debugger.
Small performance improvement to OpenGL GUI mode. No redraws will be
done if no widget has been changed, but if any changes must be made,
the whole screen is redrawn. So it's partial support for dirty updates.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@707 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
this means the text is pretty huge.
The 9x15 X font, BTW, is public domain.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@706 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
out of sync. Implemented dirty widget support for the GUI. That
means the widgets will only be drawn when necessary. There are still
a few gotcha's:
1) OpenGL mode hasn't been ported to this new scheme.
2) It's not totally finished, so some artifacts may appear onscreen.
3) Selecting active widgets with the mouse is borked.
4) Prompt commands that change the core aren't shown in the other
tabs/interface. Fixing this will require some infrastructure work
in Debugger and DebuggerParser.
5) A lot of print debug code has been left in; please ignore it for now.
Moved a lot of the debugger widgets to use non-proportional font (still
TODO is get a larger font) and not use 'magic numbers' for the layout.
That means when a new font is added, the layout should re-arrange itself.
Moved various Debugger tab widgets from 'src/gui' to 'src/debugger',
because they shouldn't be compiled when debugger support isn't
included. So now (for example), RamWidget and RamDebug are both in
the debugger directory.
Probably more stuff I'm forgetting about. It looks like the ScummVM
code can be made adequately fast, so the jump to Qt won't be necessary.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@705 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
or on the command line. See http://members.cox.net/rcolbert/ for details.
Cheetah codes are cool, but all they can do is modify ROM. We'll also be
adding some sort of extended Stella cheat code support to allow modifying
RAM, either once or once per frame, like the Game Genie does on an NES.
Most likely the actual codes will resemble Bob's codes, but with extra
digits. Unfortunately there won't be an easy way for the real Cheetah
to support these extended codes :(
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@704 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
expect the 'net will be flooded with hacked versions of Battlezone that
give you infinite lives :)
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@702 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Every cart type is going to have to support this by having a getImage()
method that returns the size of and pointer to its internal ROM image
array. Currently only Cartridge4K supports it, but it does work.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@700 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
common conditions such as "joystick 0 pressed to the left" (this would
be "_joy0left"). To see the list, type "help".
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@699 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
a modifier (such as Shift). This allows us to actually type a ~ character
in the debugger (our unary NOT operator!).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@695 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
...but we need them in the Expression classes and the DebuggerParser.
Expressions now return uInt16 instead of int. This gets us 6502-like
behaviour when dealing with numbers that don't fit into the 6502's 16-bit
address space. -1 in 6502-speak is equal to $ffff (twos' complement); this
is exactly what happens in C++ if you try to assign -1 to an unsigned
short (aka a uInt16). I believe the C++ standard doesn't *require*
conforming implementations to use twos' complement math, but I doubt
anyone will ever port Stella to any hardware old or esoteric enough to
use anything else.
Also, I've added casts to uInt8 to the debugger commands that set
registers. This also results in 6502-like handling of negative numbers:
using "a -2" to set the Accumulator will result in it having the value
$fe, which is exactly correct for a 6502. This may seem like undesirable
behaviour (and in a regular C++ program, when it happens by accident,
it IS undesired), but trust me, this is exactly what a 6502 assembly
programmer would expect.
Also also, I got rid of the distinction between commands that take a byte
or word argument. They all take words now. I had to do this to make the "a
-2" example work. A side effect is that you can now say something like "a
$1234" and the value will get truncated to $34 (due to the cast to uInt8).
What remains to be seen is whether all this behaves the way I think it
should on a big-endian platform (e.g. the Mac). My poor Mac is having
cooling issues, so I can't actually compile Stella any more :(
However, I can get it to run long enough to compile a little 5-10
line test program to see how these casts work. If I'm wrong, and they
behave differently on the Mac, I'll have to add platform-dependent
"uInt16_to_uInt8" type functions, which I'd really rather avoid...
But the debugger has GOT to treat negative and out-of-range values the
same way as a 6502 does.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@694 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
In the process, I've discovered a bug in the argument processing: any of
the commands that use kARG_FILE as their first argument will work *only*
if they're the first command issued after starting Stella, otherwise
they segfault. Am investigating this now
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@691 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
the TIA image to show frame/scanline related info. In the future,
this area will also show a zoomed region of the current TIA image,
as well as a small message window.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@689 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Currently they only show up in the prompt, but the TIA and TIADebug
have support for getting the current pixel clock (clocksThisLine()). The
pixel and CPU cyc are derived from the pixel clock.
Actually it's acting a bit funny, I need to find out why...
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@687 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
> function joy0right { !(*SWCHA & $80) }
> breakif joy0right
After deleting the breakif, joy0right is still defined, so you can use it
again...
At some point I plan to create a standard startup script full of useful
functions like joy0right. Also at some point functions will support
arguments...
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@686 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
today (I was supposed to be working on the TIA stuff and I
haven't even touched it yet). I think we have to accept that
software mode in Windows just isn't as fast as Linux (yay!),
and suggest not to use high resolutions in software mode.
Normal resolutions of 800x600 or below are fine; it's just
1024x768 and above that causes problems.
The good news is that OpenGL mode in Windows works better than
it ever did. Hopefully this will take care of that other weird
OpenGL problem that I could never track down ...
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@685 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
finally got it this time. And as a nice side effect, it seems
that switching between debugger/emulation/launcher is working
in fullscreen Windows mode.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@683 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
quite fast, and is approaching software rendering speed for lower
resolutions. In higher resolutions, OpenGL always beats software
mode.
Screen redraws are now done as rarely as possible. For example,
when switching to menu/debugger mode or pausing emulation, CPU
usage normally drops to almost nothing.
This hasn't been tested in Windows yet, so I'm sure there'll be
some issues (there always are ...)
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@682 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
and contains the entire asm source, plus line number/file and address
for each line.
Currently, you can either "loadlist file.lst" or else name your listing
file "romname.lst" (e.g. "dasm foo.asm -ofoo.bin -sfoo.sym -lfoo.lst").
Once loaded, you can say "list myAddress" and see that address's source
plus a few lines before & after for context.
This is a long way from true source-level debugging, but it's a start.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@681 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
pretty much only works for my specific setup, I think. In the future,
the regular configure script will support cross-compiling.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@680 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
screen, which means that the area right of the TIA image can be used
for widgets (that's the next thing I'll be doing).
Created TiaOutputWidget, which is contained in the DebuggerDialog
and handles all drawing/updating of the TIA image while in
debugger mode. Some advantages of this are a greatly simplified
FrameBuffer::update() wrt debugger mode and instant updates for frame
and scanline advance. Another big advantage is that the TIA image
is now a widget, which means it can receive key/mouse events. These
events can trigger updates in other parts of the interface (imagine
clicking on a certain line and having scanlines filled to that
area, etc). The biggest disadvantage right now is that the TIA is
redrawn each time the dialog changes (ie, when a key is pressed).
This needs to be optimized.
Fixed bug whereby toggling video modes while in the debugger
sometimes screwed up the palette.
Removed some FIXME's, since the functionality has been provided.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@678 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
when we hit a trap or a breakif. This fix is kind of hackish, I've got
to come up with a better way to do it.
Improved behaviour of greying-out frame in updateScanline(). It's still
not quite right.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@677 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
to *(foo+bar), just like it is in C.
Added _bank special "variable" to expression syntax. It returns the current
bank, so you can set bankswitching breakpoints:
breakif {_bank==1 && pc==whatever}
It looks like I'm going to be using _ as a prefix that means "special
meaning; reserved; don't use in your program", just like C does. Some
existing games might need to be rewritten if they use _bank or _scan as
a label... but it'll be well worth the price, to the game author.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@676 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
"breakif _scan==whatever" (break on given scanline). Eventually it'll
let us get at a lot of internal TIA state.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@675 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
need the numbers so you can refer to them (e.g. delete them).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@673 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
configure-cross is my temporary work-in-progress of the configure
script, I don't want to replace the real configure script yet because
configure-cross is probably broken :(
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@672 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
must have static versions of all the libraries (like libSDL.a, libz.a,
libpng.a), or else you'll end up with the dynamic version(s) of them... so
you'd have a stella binary that's almost-static.
Also, the only way I could get things to link properly was to include
-lz in the LIBS twice: once before and once after -lpng. Removing either
one causes the link to fail on my Linux box. This could cause issues
for other systems though, am not sure...
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@671 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
file named "romfile.stella" in the current directory, it will be run
at startup.
Fixed memory leakage in YaccParser, when there was a parse error.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@669 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Ladies and gentlemen, we have working conditional breakpoints! And they're
*fast*!
The command to use is "breakif". It takes one argument, the condition.
You can use curly braces if you want to include spaces in your arguments
(they don't do anything except help readability).
On the Athlon 2100, I was able to run the emulator at full speed, using
100% of the CPU, with *20* expressions of the form:
breakif a==101||a==102||a==103||a==104||a==105
Around 25 expressions, the game started getting noticeably jerky, but was
still playable. With 1 to 5 expressions, there was *no* noticeable increase
in CPU usage!
Actually, this can probably be improved on: I used gcc 3.2.2 for
this. With 3.3.x or 4.0.x, it should be even faster.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@665 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
command; no argument advances one scanline, else advances given #
of scanlines.
Made some methods in TIA inline. Don't know if this will have any
effect, but we'll try to be as efficient as possible.
Bumped version for the next pending alpha release.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@664 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Added rudimentary error-checking to parser and lexer: we no longer accept
invalid digits. Also fixed bug that kept YaccParser from handling hex
constants correctly.
Still TODO is to provide more friendly error messages.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@663 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
The only slight weirdness is: if you're on frame X, and you hit Frame+1,
the prompt will tell you you're on scanline 262 of frame X. It actually
should probably say you're on scanline 0 of frame X+1. If you do a Scan+1
from that point, it'll say you're on scanline 1 of frame X (still X,
not X+1).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@661 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
advance and frame advance. There are still some issues though.
Specifically, scanline advancing past the end of the frame shouldn't
be allowed, and this is partially causing the problems with frame
'jumping' when doing a scanline advance after a frame advance.
Still TODO is add some visual cue to indicate which scanline is
being drawn, and which onscreen data is 'old' data.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@660 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
except it lets you keep advancing past the end of the frame.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@659 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
stephena