VASM assert directive

[hop]

Is there a way to use ASSERT directives with the vasm motorola syntax module?

I wrote a ZX Spectrum Next game using sjasmplus and it was great to be able to do stuff like this:

; 9-bit colour palette

spritePalette:
;        RRR'GGG'BB  B
    DB  %000'000'00,%0  ; $00 black
    DB  %111'111'11,%1  ; $01 white
    DB  %011'011'11,%1  ; $02 blue-grey
    DB  %011'111'01,%1  ; $03 green-grey
    DB  %111'011'01,%1  ; $04 red-grey
    DB  %111'000'00,%0  ; $05 red
    DB  %000'111'00,%0  ; $06 green
    DB  %000'000'11,%1  ; $07 blue
    DB  %111'111'00,%0  ; $08 yellow
    DB  %111'101'00,%0  ; $09 orange
    DB  %101'101'10,%1  ; $0a light grey
    DB  %010'010'01,%0  ; $0b dark grey
    DB  %111'100'10,%0  ; $0c pink (invader bullets)
    ASSERT ($ - spritePalette) % 2 == 0 ; expect even number of bytes
SPRITE_PALETTE_COUNT EQU ($ - spritePalette) / 2 ; number of colours
    ASSERT SPRITE_PALETTE_COUNT <= 256

ufoPattern:
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$00,$00,$0a,$0a,$0a,$0a,$0a,$0a,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$00,$00,$00
    DB  $00,$00,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$00,$00
    DB  $00,$0a,$0a,$0a,$09,$09,$0a,$08,$08,$0a,$09,$09,$0a,$0a,$0a,$00
    DB  $0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a,$0a
    DB  $0a,$0b,$0a,$0b,$0b,$0a,$0b,$0b,$0b,$0b,$0a,$0b,$0b,$0a,$0b,$0a
    DB  $0a,$00,$0a,$00,$00,$0a,$00,$00,$00,$00,$0a,$00,$00,$0a,$00,$0a
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    DB  $00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00,$00
    ASSERT $-ufoPattern == 256

I can see that the vasm MadMac module suports 'assert' and hoping there is a way to enable it for MOT.

[phx]

Originally Posted by hop

Is there a way to use ASSERT directives with the vasm motorola syntax module?

It could be added in a minute. Probably 2 or 3 lines of C code. The reason why it's not there is that I have never seen an Assert directive in common Amiga assemblers, like Devpac, etc., and I don't like to introduce new vasm-only directives, which make source texts incompatible.

With some conditional assembly you could easily reach the same effect:

        ifne    (*-spritePalette)%2
        fail    odd address
        endif

EDIT: I was too fast. This is a bad workaround to check label differences in an optimizing assembler. See below.

[a/b]

I'm using this in asm-one/pro, similar to assert (IF2 = assembler is in pass 2):

_CHECK	MACRO	(CC, Value1, Value2)
	IF2
		IF\1	(\2)-(\3)
			PRINTT	"\2 is *\1* \3."
			PRINTV	(\2)-(\3)
			FAIL
		ENDIF
	ENDIF
	ENDM

CHECKEQ	MACRO	(Value1, Value2)
	_CHECK	NE,\1,\2
	ENDM

CHECKNE	MACRO	(Value1, Value2)
	_CHECK	EQ,\1,\2
	ENDM

CHECKGT	MACRO	(Value1, Value2)
	_CHECK	LE,\1,\2
	ENDM

CHECKLT	MACRO	(Value1, Value2)
	_CHECK	GE,\1,\2
	ENDM

[phx]

Originally Posted by a/b

I'm using this in asm-one/pro, similar to assert (IF2 = assembler is in pass 2):

Oh yes, IF2...
Only works with two-pass assemblers. vasm parses the source only once, but uses a variable number of passes until the optimization is perfect (can be hundreds of passes).

But it's good that you mentioned IF2, because the reason for its existence is that conditional assembly doesn't work very well with optimizations and multiple passes. So the workaround I have shown above it useless. The IFNE is evaluated only during initial parsing and later changes due to optimizations are not recognized (which is bad when testing label differences).

So there is a reason for an ASSERT directive to exist. Because it is not affected by these problems. It evaluates the expression in the final pass only.

As vasm doesn't really support IF2, an ASSERT directive might make sense...

[meynaf]

I would do the above example this way for 68k :

SPRITE_PALETTE_COUNT set 0
dbx macro
 dc.w \1
SPRITE_PALETTE_COUNT set SPRITE_PALETTE_COUNT +1
 endm

spritePalette:
    dbx  %000000000  ; $00 black
    dbx  %111111111  ; $01 white
    dbx  %011011111  ; $02 blue-grey
    dbx  %011111011  ; $03 green-grey
    dbx  %111011011  ; $04 red-grey
    dbx  %111000000  ; $05 red
    dbx  %000111000  ; $06 green
    dbx  %000000111  ; $07 blue
    dbx  %111111000  ; $08 yellow
    dbx  %111101000  ; $09 orange
    dbx  %101101101  ; $0a light grey
    dbx  %010010010  ; $0b dark grey
    dbx  %111100100  ; $0c pink (invader bullets)
; use of dc.w implies even number of bytes
    ifgt SPRITE_PALETTE_COUNT - 256
    fail
    endc

[hop]

Thanks for the insight and suggestions. I don't want to encourage divergence in assembler syntax, so I think macros sound like the way to go.

I usually use asserts to verify data size, struct member offsets (for code that makes assumptions) in which case optimisations would not affect them. This is assuming that like a C compiler, the assembler will never modify data structures or data.

It would however be nice to assert section sizes which would be affected by optimisations though.

[phx]

Originally Posted by hop

This is assuming that like a C compiler, the assembler will never modify data structures or data.

Yes. Usually there is nothing to optimize in pure data. But directives to declare storage may use expressions depending on labels:

        ds.b    label2-label1

It would however be nice to assert section sizes which would be affected by optimisations though.

I guess I will add an assert to mot-syntax. If anybody knows such a directive from a 68k assembler, then please tell me the syntax, so I don't invent something new.

Otherwise I would probably settle for

assert <expression>[,<optional message on failure>]

[hop]

Originally Posted by phx

I guess I will add an assert to mot-syntax.

That would be great thanks, as long as you are sure it's the right decision. Perhaps a command line option would be required to enable it in case pre-existing projects that define 'assert' break.

[phx]

Done.

assert <expr>[,<msg>]

will be in tomorrow's daily source snapshot (and the next release).

[hop]

Excellent. Many thanks.

How do you decide when to release?

[phx]

Hard to explain. A combination of enough new features or important bug fixes, and the feeling that it has been sufficiently tested.

[hop]

Originally Posted by phx

Done.

assert <expr>[,<msg>]

will be in tomorrow's daily source snapshot (and the next release).

Please forgive me but I am a bit confused by all the combinations of downloads. I approached Amiga OS programming from C, and have now descended into assembly. My "update_dependencies.bat" script is set to:

set VBCC_VERSION=2022-03-23
...
SET VBCC_BIN_ZIP=vbcc_bin_win64.zip
...
curl -O http://phoenix.owl.de/vbcc/%VBCC_VERSION%/%VBCC_BIN_ZIP%

So I am using vasm from the vbcc binary package.

I can't see vbcc daily binaries at http://sun.hasenbraten.de/vbcc/ so I guess I need to download the vasm daily binaries from http://sun.hasenbraten.de/vasm/index.php?view=bincur but which one do I want from the list to get the equivalent (to cross-compile on win64 for classic Amiga OS (1.3/2.0/3.2))?

1. vasmm68k_mot_os3.lha 22 Apr 2022 02:17:03 (190506 bytes)
2. vasmm68k_std_os3.lha 22 Apr 2022 02:19:25 (186955 bytes)
3. vasmppc_std_os3.lha 22 Apr 2022 02:20:18 (154705 bytes)
4. vasmm68k_mot_mos.lha 22 Apr 2022 02:22:55 (230575 bytes)
5. vasmm68k_std_mos.lha 22 Apr 2022 02:25:30 (226048 bytes)
6. vasmppc_std_mos.lha 22 Apr 2022 02:26:23 (191261 bytes)
7. vasmm68k_mot_os4.lha 22 Apr 2022 02:29:02 (243367 bytes)
8. vasmm68k_std_os4.lha 22 Apr 2022 02:31:37 (237368 bytes)
9. vasmppc_std_os4.lha 22 Apr 2022 02:32:32 (201691 bytes)
10. vasmm68k_mot_TOS.zip 22 Apr 2022 02:35:11 (183026 bytes)
11. vasmm68k_std_TOS.zip 22 Apr 2022 02:37:49 (179669 bytes)
12. vasmm68k_mot_MiNT.zip 22 Apr 2022 02:40:11 (186157 bytes)
13. vasmm68k_std_MiNT.zip 22 Apr 2022 02:42:32 (182488 bytes)

Or do I need to build from source?

[phx]

Originally Posted by hop

so I guess I need to download the vasm daily binaries from http://sun.hasenbraten.de/vasm/index.php?view=bincur

Correct.

but which one do I want from the list to get the equivalent (to cross-compile on win64 for classic Amiga OS (1.3/2.0/3.2))?

Unfortunately there are no Windows executables for daily source snapshots. We only build those for official releases.

Or do I need to build from source?

Yes. Or wait for the release. Or ask somebody to compile it. But not me, because I have no Windows cross compiler installed. Volker does that, whenever there is a release.

[paraj]

Attached is a build for 64-bit windows. Only did some very light testing to check out the feature, so there might be issues with it, that won't be present in an official release. Obviously completely unsupported, so use at your own risk.

Also note that you can't use it in devpac or phxass compatibility mode, and if you're used to the latter you probably want to add "-spaces" to the command line

[phx]

Originally Posted by paraj

Attached is a build for 64-bit windows.

Tak skal du have!

Also note that you can't use it in devpac or phxass compatibility mode

That's the least I could do to make people aware that they are using a vasm-only directive.

[paraj]

Originally Posted by phx

That's the least I could do to make people aware that they are using a vasm-only directive.

Yes, absolutely reasonable, I was just trying to preempt possible questions based on my own quick interaction. I tried it a random project and it didn't work => Ah, phxass mode enabled => Nothing works => Need "-spaces"

[hop]

Thanks very much. It will be very reassuring to be able to plonk the binary in place and (static_)assert

[hop]

ASSERT seems to be doing the job. Many thanks.

[hop]

ASSERT in action

	move.l (a2)+,d1				; hunkIndex

	; calcuate byte offset into RootHunk array
	IF 0
	mulu #RootHunk_SIZEOF,d1	
	ELSE	
	ASSERT RootHunk_SIZEOF==8
	lsl.w #3,d1				; more optimal, but brittle
	ENDIF
	movea.l #rootHunks,a4			; first RootHunk
	movea.l RootHunk_Address(a4,d1.w),a4	; referenced hunk address

[phx]

Although you could also write

IF RootHunk_SIZEOF==8

in this example, so there would be no need for ASSERT.