68HC05: Fix #7064 - incorrect handling of X-indexed JMP and JSR

[philpem]

Fix for #7064 (incorrect handling of X-indexed JMP and JSR)

[GhidorahRex]

This doesn't fix the issue for the HC05/HC08/HCS08. The issue is that the subconstructor is exporting an address, so the goto [ADDRI]; tries to de-reference the address, whereas goto ADDRI tries to go to the unique varnode U500:1. The solution is to change the subconstructor to export the value, rather than the address. But the subconstructors oprx8_8_X, oprx16_8_X, and comma_X are also used in OP1 which should export the address.

So this is what I came up with, and it seems to work well, leaving the brackets in ADDRI. I imagine the same changes would be needed for the 6805:

@if defined(HC05)
oprx8_8_X:		imm8,X   is imm8 & X   { address:2 = zext(X) + imm8;   export  address; }
oprx16_8_X:		imm16,X  is imm16 & X  { address:2 = zext(X) + imm16;  export  address; }
comma_X:		","X     is X		   { address:2 = zext(X);          export  address; }
@endif

@if defined(HCS08) || defined(HC08)
oprx8_8_X:		imm8,X   is imm8 & X   { address:2 = HIX + imm8;       export address; }
oprx16_8_X:		imm16,X  is imm16 & X  { address:2 = HIX + imm16;      export address; }
comma_X:		","X     is X		   { address:2 = HIX;              export address; }
@endif

and

OP1: iopr8i		is op4_6=2; iopr8i       { export iopr8i; }
OP1: opr8a_8	is op4_6=3; opr8a_8      { export opr8a_8; }
OP1: opr16a_8	is op4_6=4; opr16a_8     { export opr16a_8; }
OP1: oprx16_8_X	is op4_6=5; oprx16_8_X   { export *:1 oprx16_8_X; }
OP1: oprx8_8_X	is op4_6=6; oprx8_8_X    { export *:1 oprx8_8_X; }
OP1: comma_X	is op4_6=7 & comma_X     { export *:1 comma_X; }

[byt3n33dl3]

JMP and JSR Code Review

[byt3n33dl3]

Format / Master

maybe back to "[ADDRI]" ?

[philpem]

This doesn't fix the issue for the HC05/HC08/HCS08. The issue is that the subconstructor is exporting an address, so the goto [ADDRI]; tries to de-reference the address, whereas goto ADDRI tries to go to the unique varnode U500:1. The solution is to change the subconstructor to export the value, rather than the address. But the subconstructors oprx8_8_X, oprx16_8_X, and comma_X are also used in OP1 which should export the address.

So this is what I came up with, and it seems to work well, leaving the brackets in ADDRI. I imagine the same changes would be needed for the 6805:

Yep. I agree that this is the correct fix - it seems to work for me too, so I've pulled it into my branch.

I've applied the same fix to the 6805 SLASPEC too -- this also had a bug where the non-offset indexed syntax would be incorrect -- it would show as LDA X instead of LDA ,X.

I'll have to read the SLEIGH spec more closely in regard to what the *:1 does. Mea culpa for pushing this without properly testing it!