klaus2m5 / 6502_65c02_functional_tests Goto Github PK

A bug reported in the a6502 emulator project indicates a possible test hole, because I believe that project passes this test suite. See
BigEd/a6502#2

when carry flag is cleared and register value is 0x01, it fails to set zero flag.

(I haven't yet been able to test this myself.)

I'm writing a nes emulator which uses a custom NMOS 6502 that doesn't implement decimal mode ADC/SBC. It would be nice to have an option not to test this functionality.

Maybe allow for self modifying code to be an option?
See http://stackoverflow.com/q/31260415 where someone wants to run from ROM

As indicated in the source code, I followed the link to the AS65 assembler at the indicated link: http://www.kingswood-consulting.co.uk/assemblers/ which is unfortunately broken. As a retro-computing fan, I see a lot of these issues happening. In fact, nowadays it would be not a bad idea to simply include the compiler into the project to avoid any external dependencies. Another option would be to re-write the source-code to run on more widespread assemblers, or publish compiled hex-code listings alongside.

I'm writing a Lynx emulator and I've used this test suit many months ago and my CPU emulation eventually passed it (after many corrections as the tests are excellent and I've found many bugs thanks to them). It turns out that I had one bug more - in RMW operations addressing absolutely with indexing (e.g. inc $80ff,x) when page crossing occurred ( here x > 0 ) the read was from correct address but write to not corrected (i.e. to page $80 not $81 in above example). It wasn't cached by this test suit.

I tried to compile the .bin files myself on a Windows 10 VM, using the zipped as65.exe.

 ./as65.exe -l -m -w -h0 ../6502_functional_test.a65

Somehow, the resulting .bin file's codes segment starts 10bytes before 0x0400.

I fixed this using

dd if=6502_functional_test.bin of=6502_functional_test.bin.padded bs=1 seek=10 

Any idea why this happens?

Hi, I have a problem with the 65c02 test code:Two uninitialized memory spaces are compared and this leads to a trap.

Should really the index register y be used? x is cleared jut before. But then a loop over y is programmed...

; testing load / store accumulator LDA / STA (zp)
.
.
.

        trap_ne
        cpx #$99
        trap_ne     ;x altered during test
        cpy #$66                                                              <= Here it is clear that y is #$66
        trap_ne     ;y altered during testorg
        ldx #0
tstai2  lda abst,y                                                           <= Here $0205 + $66 = $026B is read, which is uninitialized
        eor #$c3
        cmp abs1,y                                                           <= Here $0210 + $66 = $0276 is read, which is uninitialized
        trap_ne     ;store to indirect data                           <= Compare fails and trap occurred
        txa
        sta abst,y  ;clear
        dey
        bpl tstai2
        tsx
        cpx #$ff
        trap_ne     ;sp push/pop mismatch
        next_test

Hi Klaus,

I have just been writing a small 6502 simulator in Python3 for supporting my own FPGA developments, and found your very useful test suite. I have made a little modification in your asm code to be able to produce a complete ROM image with reset vector pointing to the start of the testing code. The reset vector is written back to point original "res_trap" vector, right after the start. This feature is useful when you start the testing code on a simple virtual microcomputer which has no other executable code apart from this testing suite -> after "powering on" this machine, the test code is started immediately by the reset mechanism.

If you find this extension usable, please feel free to lift it into your test code. Attached zip contains the complete modified asm.

Thank you for attention,
Zoltan

6502_functional_test.zip

Hi

This is not a ticket - but I haven't got your email address, so ;-)

I am one of countless others who are having fun implementing a 6502 simulator. I have included
one bin file from your repo (6502_functional_test.bin). How do you want me to credit you?

Best

Morten

It appears that the files 6502_functional_test.bin/lst checked in under bin_files have not been updated to reflect the latest commit from 2020

By the way - Thanks for creating this—it's awesome!

Hi, I really like your 6502 testing suite but I found an error in report.i65 that converts values to hexadecimal characters for IO debugging output:

Line 189: bcc rchar

This line in the "rnib" subroutine should actually be a BCS instead. Thanks :)

Hi,

I'm adding 65C02 support to an emulator and currently stuck on the TRB tests for 65C02

It completes an op1 loop, but appears to fail when OP2/$0B = $01
As you can see, AND $0B does not set the Zero flag because A=1 and $0B = 1.

I should note that this emulator passes the 6502 Tests with no issues so I'm really confused as to what is going on here.

ZPG values at failure time:

0008-000F  00 00 01 01 00 00 01 81

Trace log:

1CC2  98        TYA           A:01 X:C0 Y:01 F:21 S:1FF [..-....C]
1CC3  25 0B     AND $0B       A:01 X:C0 Y:01 F:21 S:1FF [..-....C]
1CC5  08        PHP           A:01 X:C0 Y:01 F:21 S:1FE [..-....C]
1CC6  68        PLA           A:31 X:C0 Y:01 F:21 S:1FF [..-....C]
1CC7  29 02     AND #$02      A:00 X:C0 Y:01 F:23 S:1FF [..-...ZC]
1CC9  85 0C     STA $0C       A:00 X:C0 Y:01 F:23 S:1FF [..-...ZC]
1CCB  98        TYA           A:01 X:C0 Y:01 F:21 S:1FF [..-....C]
1CCC  49 FF     EOR #$FF      A:FE X:C0 Y:01 F:A1 S:1FF [N.-....C]
1CCE  05 0B     ORA $0B       A:FF X:C0 Y:01 F:A1 S:1FF [N.-....C]
1CD0  49 FF     EOR #$FF      A:00 X:C0 Y:01 F:23 S:1FF [..-...ZC]
1CD2  85 0D     STA $0D       A:00 X:C0 Y:01 F:23 S:1FF [..-...ZC]
1CD4  98        TYA           A:01 X:C0 Y:01 F:21 S:1FF [..-....C]
1CD5  05 0B     ORA $0B       A:01 X:C0 Y:01 F:21 S:1FF [..-....C]
1CD7  85 0E     STA $0E       A:01 X:C0 Y:01 F:21 S:1FF [..-....C]
1CD9  84 0A     STY $0A       A:01 X:C0 Y:01 F:21 S:1FF [..-....C]
1CDB  A9 FF     LDA #$FF      A:FF X:C0 Y:01 F:A1 S:1FF [N.-....C]
1CDD  48        PHA           A:FF X:C0 Y:01 F:A1 S:1FE [N.-....C]
1CDE  A5 0B     LDA $0B       A:01 X:C0 Y:01 F:21 S:1FE [..-....C]
1CE0  28        PLP           A:01 X:C0 Y:01 F:FF S:1FF [NV-BDIZC]
1CE1  14 0A     TRB $0A       A:01 X:C0 Y:01 F:FF S:1FF [NV-BDIZC]
1CE3  08        PHP           A:01 X:C0 Y:01 F:FF S:1FE [NV-BDIZC]
1CE4  C5 0B     CMP $0B       A:01 X:C0 Y:01 F:7F S:1FE [.V-BDIZC]
1CE6  D0 FE     BNE $FE       A:01 X:C0 Y:01 F:7F S:1FE [.V-BDIZC]
1CE8  68        PLA           A:FF X:C0 Y:01 F:FD S:1FF [NV-BDI.C]
1CE9  48        PHA           A:FF X:C0 Y:01 F:FD S:1FE [NV-BDI.C]
1CEA  09 02     ORA #$02      A:FF X:C0 Y:01 F:FD S:1FE [NV-BDI.C]
1CEC  C9 FF     CMP #$FF      A:FF X:C0 Y:01 F:7F S:1FE [.V-BDIZC]
1CEE  D0 FE     BNE $FE       A:FF X:C0 Y:01 F:7F S:1FE [.V-BDIZC]
1CF0  68        PLA           A:FF X:C0 Y:01 F:FD S:1FF [NV-BDI.C]
1CF1  29 02     AND #$02      A:02 X:C0 Y:01 F:7D S:1FF [.V-BDI.C]
1CF3  C5 0C     CMP $0C       A:02 X:C0 Y:01 F:7D S:1FF [.V-BDI.C]
1CF5  D0 FE     BNE $FE       A:02 X:C0 Y:01 F:7D S:1FF [.V-BDI.C]
1CF5  D0 FE     BNE $FE       A:02 X:C0 Y:01 F:7D S:1FF [.V-BDI.C]
1CF5  D0 FE     BNE $FE       A:02 X:C0 Y:01 F:7D S:1FF [.V-BDI.C]

Hello!
Could you add the missing bin file for the interrupt tests and the decimal tests?
(and seeing that the other tests got an update a few weeks ago, perhaps update the two other bin files with the recent changes?)
Thanks a lot!

I have a question about this portion of your functional test

37c3 : ba                       tsx             ;test break on stack
37c4 : bd0201                   lda $102,x
                                cmp_flag 0      ;break test should have B=1 & unused=1 on stack
37c7 : c930            >            cmp #(0      |fao)&m8    ;expected flags + always on bits

This is what I see in my emulator:

37C3: BA         TSX                   A=42 X=FB Y=48 S=FB P=B1 PC=$37C3 P={$B1 N=1 V=0 B=1 D=0 I=0 Z=0 C=1} SP={$FB stack:[$1FF:$09 $1FE:$CF $1FD:$00 $1FC:$30 ]}
37C4: BD 02 01   LDA $0102,X           A=00 X=FB Y=48 S=FB P=33 PC=$37C4 P={$33 N=0 V=0 B=1 D=0 I=0 Z=1 C=1} SP={$FB stack:[$1FF:$09 $1FE:$CF $1FD:$00 $1FC:$30 ]}
37C7: C9 30      CMP #$30              A=00 X=FB Y=48 S=FB P=B0 PC=$37C7 P={$B0 N=1 V=0 B=1 D=0 I=0 Z=0 C=0} SP={$FB stack:[$1FF:$09 $1FE:$CF $1FD:$00 $1FC:$30 ]}
37C7: D0 30      BNE $37C7             A=00 X=FB Y=48 S=FB P=B0 PC=$37C7 P={$B0 N=1 V=0 B=1 D=0 I=0 Z=0 C=0} SP={$FB stack:[$1FF:$09 $1FE:$CF $1FD:$00 $1FC:$30 ]}

As you can see, my stack contains the expected value (#$30), however, TSX followed by LDA $102,X points toward the next available spot of the stack, not the current value. So I get 0 and the test fails.

Am I missing something?

First, thank you very much for this fantastic suite.
This is not really an issue, but I've noticed that the compiled examples belong to a previous version. The 6502 LST file states 2015 and the 65c02 LST file states 2017.
I've compiled myself the 6502 and I've obtained a different HEX/BIN and LST file.

Raul

The 6502 test checks the BRK instruction only with the interrupt disable flag off. The additional 65C02 test makes the humble assumption that the BRK instruction works in general and does not check if the BRK is actually executed.

I will add a second pass with I flag on to the test and repeat the full BRK test for the 65C02.