const from type

[nidud]

From the UCRT source (strcmp.asm):

CHAR_TYPE EQU BYTE
CHAR_SIZE = sizeof CHAR_TYPE

Test case:

.386
.model flat
CHAR_T EQU SBYTE
WCHAR_T EQU WORD
.code
 mov eax,size CHAR_T	    ; 1
 mov eax,size WCHAR_T	    ; 2
 mov eax,sizeof CHAR_T	    ; 1
 mov eax,sizeof WCHAR_T	    ; 2
 mov eax,type CHAR_T	    ; 0
 mov eax,type WCHAR_T	    ; 0
 mov eax,length CHAR_T	    ; 1
 mov eax,length WCHAR_T	    ; 2
 end

[Baron-von-Riedesel]

Masm returns 0 for TYPE CHAR_T, but 1 for TYPE BYTE - is this supposed to be a bug?

Also, operator LENGTH should never accept a TYPE, it needs a data label - what Masm version does accept it?

[nidud]

what Masm version does accept it?

Same result for these two:

Microsoft (R) Macro Assembler Version 6.14.8444
Microsoft (R) Macro Assembler Version 14.50.35730.0

Also, operator LENGTH should never accept a TYPE, it needs a data label

LENGTH is "technically" SIZEOF / TYPE but yes: LENGTH(variable), SIZEOF(variable or type), and TYPE(expression).

Masm returns 0 for TYPE CHAR_T, but 1 for TYPE BYTE - is this supposed to be a bug?

SIZEOF and LENGTH uses the underlying type there which seems a logical outcome, but TYPE(const) normally returns 0.

[Baron-von-Riedesel]

After thinking about this a bit more thoroughly I came to the opinion that the SIZE/SIZEOF value returned by Masm is a bug ( and TYPE behaves correctly ).

If you want to define an alias type, there's the TYPEDEF directive ( or, for just string substitution, there's EQU <> ). Using the plain EQU is supposed to hold the numeric value of an expression, and no type information.

What's definitely a JWasm bug, however, is that its TYPE operator doesn't return 0 for numeric equates ...

[nidud]

After thinking about this a bit more thoroughly I came to the opinion that the SIZE/SIZEOF value returned by Masm is a bug ( and TYPE behaves correctly ).

I still hold the view that ifdef rax is a bug but finally yielded on that one, and for similar reasons for this, as sizeof CHAR_TYPE is consistently used in publicly available sources written for Masm by Microsoft (at least from 2001 as far as I know). The solution is also similar to the signed const value issue where Masm store type information on conversion and actively use this.

If you want to define an alias type, there's the TYPEDEF directive ( or, for just string substitution, there's EQU <> ). Using the plain EQU is supposed to hold the numeric value of an expression, and no type information.

True, but reg EQU eax is also widely used and accepted.

[Baron-von-Riedesel]

True, but reg EQU eax is also widely used and accepted.

Yes, but that EQU defines a text macro. That's similar to format

CHAR_TYPE EQU <BYTE>

and everything is fine.

as sizeof CHAR_TYPE is consistently used in publicly available sources written for Masm by Microsoft

I guess the quality of MS code has significantly decreased in recent decades... :))

[nidud]

Yes, but that EQU defines a text macro.

I guess that was may point , that Masm do something similar to byte as JWasm do to eax there, but I leave you to it.