|
|
Joined: Feb 2014
Posts: 1,225 Likes: 224
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2014
Posts: 1,225 Likes: 224 |
I was trying to use install_write_tap in lua to watch writes to an io port,
the novag savant (driver name is savant) writes to 0xc1 and I wanted to watch this:
manager.machine.devices[":maincpu"].spaces["io"]:install_write_tap(0xc1,0xc1,"hi",function(offset,data,mask)print("testing",offset,data,mask) end)
It will run it a single time, then I don't see any more output.
testing 193 0 255
Am I doing something wrong?
(so every time you move a piece, the savant will write to c1 (which connects to an external clock timer))
Last edited by Golden Child; 06/17/25 07:24 PM.
|
|
|
|
|
Joined: Feb 2004
Posts: 2,641 Likes: 351
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2004
Posts: 2,641 Likes: 351 |
Write taps work properly in Lua. I’ve used them extensively for debugging the IGS gambling games. Think about what you’re neglecting.
|
|
|
|
|
Joined: Feb 2014
Posts: 1,225 Likes: 224
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2014
Posts: 1,225 Likes: 224 |
ok, got it. understood. The z80 io bus is weird, you can't just look at the low 8 bits, you have to consider 16bits and the data seems to get written to the high 8 bits. so in the io map, when we look at c1 there's a mirror: map(0xc1, 0xc1).mirror(0xff38).lw8([this](u8 data){printf("c1 data = %x %s\n", data, machine().describe_context().c_str()); }, "c1 write"); // printer and adjusting the write tap to look at any IO address, now I see the writes to c1:
m_maincpu->space(AS_IO).install_write_tap(
0x0,0xffff,
"maincpu_ext_w",
[this] (offs_t offset, u8 &data, u8 mem_mask)
{
if ((offset & 0xff) == 0xc1)
printf("WRITE TAP CPP = %x, %x\n",(u8) offset, (u8) data);
});
Last edited by Golden Child; 06/17/25 07:18 PM.
|
|
|
|
|
Joined: Feb 2014
Posts: 1,225 Likes: 224
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2014
Posts: 1,225 Likes: 224 |
ok, so if I run this:
./mame savant
sp = manager.machine.devices[":maincpu"].spaces["io"] sp:install_write_tap(0x0,0xffff,"hi5",function(offset,data) print(string.format("%x",offset),data)end)
It will run for awhile, then abruptly stop.
If I launch colecovision with
./mame coleco timeplt
and do
sp = manager.machine.devices[":maincpu"].spaces["io"] sp:install_write_tap(0x0,0xff,"hi5",function(offset,data) print(string.format("%x",offset),data)end)
(coleco has an address mask of 0xff)
it will run for awhile then just stop. I can start it up again with the same command line, but it will stop soon.
Shouldn't it keep running?
|
|
|
|
|
Joined: Feb 2004
Posts: 2,641 Likes: 351
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2004
Posts: 2,641 Likes: 351 |
It works. You’re neglecting something important.
|
|
|
|
|
Joined: Feb 2014
Posts: 1,225 Likes: 224
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2014
Posts: 1,225 Likes: 224 |
ok, so I call install_write_tap and it stops working, I put a printf in the remove routine and I can see where it gets removed.
void remove()
{
++m_installing;
try
{
printf("removing passthrough %s \n",name().c_str());
m_handler.remove();
}
catch (...)
{
--m_installing;
throw;
}
--m_installing;
}
function d (offset,data) c=c or 0 c=c+1 print("FUNCTION D",string.format("%x",offset),data,c,emu.time()) end
sp = manager.machine.devices[":maincpu"].spaces["io"] sp:install_write_tap(0x0,0xffff,"hi6",d)
[MAME]> function d (offset,data)c=c or 0 c=c+1 print("FUNCTION D",string.format("%x",offset),data,c,emu.time()) end [MAME]> sp = manager.machine.devices[":maincpu"].spaces["io"] sp:install_write_tap(0x0,0xffff,"hi6",d) INSTALLING WRITE TAP 0 [MAME]> FUNCTION D c5 0 1 0.00012799999999949 WRITE_TAP_EXECUTEd offset=c5 data=0 ... FUNCTION D 1fc0 31 232 0.56272483333108 WRITE_TAP_EXECUTEd offset=1fc0 data=1f FUNCTION D 1fc0 31 233 0.56648099999773 WRITE_TAP_EXECUTEd offset=1fc0 data=1f removing passthrough hi6 I don't understand why it gets removed. I have another tap installed in the savant.cpp machine_start that continues to run along without any problems.
|
|
|
|
|
Joined: Feb 2004
Posts: 2,641 Likes: 351
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2004
Posts: 2,641 Likes: 351 |
It’s something fundamental.
|
|
|
|
|
Joined: Jun 2001
Posts: 539 Likes: 45
Senior Member
|
|
Senior Member
Joined: Jun 2001
Posts: 539 Likes: 45 |
He's getting GC-ed into oblivion? One has to admit the documentation is quite unclear on the fact that the lua install_*_tap returns something, even less something important.
|
|
1 member likes this:
Golden Child |
|
|
|
|
Joined: Feb 2014
Posts: 1,225 Likes: 224
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2014
Posts: 1,225 Likes: 224 |
Ahhhh, ok. That makes sense.
function d (offset,data)c=c or 0 c=c+1 print("FUNCTION D",string.format("%x",offset),data,c,emu.time()) end
sp = manager.machine.devices[":maincpu"].spaces["io"]
holdit=sp:install_write_tap(0x0,0xffff,"hi6",d)
[MAME]> print(holdit)
sol.sol::d::u<lua_engine::tap_helper>: 0x561c31740048
so as long as holdit holds the value, it doesn't get GC'd!
and then holdit=nil will allow it to get GC'd.
thanks for the pointer, OG!
(you know I was thinking that it got GC'd possibly, didn't think to "grab" a handle to prevent that)
Last edited by Golden Child; 06/27/25 07:04 PM.
|
|
|
|
|
Joined: Feb 2004
Posts: 2,641 Likes: 351
Very Senior Member
|
|
Very Senior Member
Joined: Feb 2004
Posts: 2,641 Likes: 351 |
He's getting GC-ed into oblivion? One has to admit the documentation is quite unclear on the fact that the lua install_*_tap returns something, even less something important. Hey, it actually has a clickable link to the documentation for the class of the thing it returns!
|
|
|
|
Forums9
Topics9,381
Posts122,711
Members5,085
| |
Most Online1,529
Jun 7th, 2025
|
|
These forums are sponsored by Superior Solitaire, an ad-free card game collection for macOS and iOS. Download it today!
|
|
|
|
