jwasm 2.01 rc

[japheth]

Hello,

there's a new JWasm release candidate available at http://www.japheth.de/JWasm.html. It was intended to be a "bugfix" release only, but I was a bit curious and implemented support for SSE4.1 and SSE4.2 instruction sets. AFAIK that makes this release virtually compatible with Masm v9. The exception is the SSE4A instruction set, which isn't implemented yet ( mainly  because of the strange INSERTQ instruction which needs 4 arguments, a hard job for the JWasm parser ).

[UtillMasm]

 :U
PS. is jwasm written in masm language?

[hutch--]

C.

[Ficko]

This macro:

Code Select Expand


; -----------------------
;; INLINE [nX] OPCODE,...
; -----------------------
inline equ <INLINE>
Inline equ <INLINE>
INLINE MACRO v1,v2,v3,v4,v5,v6,v7,v8,v9,v10,v11,v12,v13,v14,v15
LOCAL pos
FOR arg,<v1,v2,v3,v4,v5,v6,v7,v8,v9,v10,v11,v12,v13,v14,v15>
  IFNB <arg>
   pos=@InStr(1,arg,<!:>)
   argy CATSTR <arg>
   IF pos
    argy CATSTR <@SubStr(<arg>,1,pos-1),@SubStr(<arg>,pos+1)>
   ENDIF
   IFIDN @SubStr(<arg>,2,1),<X>
    REPEAT @SubStr(<arg>,1,1)
     @SubStr(%argy,4)
    ENDM
   ELSE
    argy
   ENDIF
  ENDIF
ENDM
ENDM


won't work in JWASM.

Usage:

Code Select Expand


inline 2X inc eax,mov nEvents:eax


Looks like the using of ":" messing up something.

[japheth]

Looks like the using of ":" messing up something.

The ":" is innocent. Here's a simplified test case:

Code Select Expand


.386
.model flat

argy CATSTR <@SubStr(<mov nEvents:eax>,1,11) , @SubStr(<mov nEvents:eax>,13)>

.data

nEvents dd 0

.code

argy

End


the problem occurs if a text macro contains more than 1 macro function, and the evaluation of the macro functions is "delayed" ( = they are enclosed in <> ).

[japheth]

It is fixed.

[Ficko]

Thanx,

that was faaast! :bg

I just compiled a library with 408 modules.
No errors.  :U

[japheth]

I just compiled a library with 408 modules.
No errors.  :U

Good! So v2.01 is final now.

[Ficko]

Good! So v2.01 is final now.

Just slow down a little bit. :bg

I didn't say everything is ok.
Just sad that JWASM didn't report error. :P

Seriously there is something not working right in my library.
A dialog won't open but no exception is generated.

I did found something that is strange may causing the problem.

Portion of my code:

Code Select Expand


.686p
.model flat
; ------------------------------------------------------------------
LF_FACESIZE       equ 32
; ------------------------------------------------------------------
LOGFONT STRUCT
  lfHeight          DWORD ?
  lfWidth           DWORD ?
  lfEscapement      DWORD ?
  lfOrientation     DWORD ?
  lfWeight          DWORD ?
  lfItalic          BYTE ?
  lfUnderline       BYTE ?
  lfStrikeOut       BYTE ?
  lfCharSet         BYTE ?
  lfOutPrecision    BYTE ?
  lfClipPrecision   BYTE ?
  lfQuality         BYTE ?
  lfPitchAndFamily  BYTE ?
  lfFaceName        BYTE LF_FACESIZE dup(?)
LOGFONT ENDS
WNDCLASSEX STRUCT
   cbSize            DWORD ?
   style             DWORD ?
lpfnWndProc       DWORD ?
cbClsExtra        DWORD ?
cbWndExtra        DWORD ?
hInstance         DWORD ?
hIcon             DWORD ?
hCursor           DWORD ?
hbrBackground     DWORD ?
lpszMenuName      DWORD ?
lpszClassName     DWORD ?
hIconSm           DWORD ?
WNDCLASSEX ENDS
TEXTMETRICA STRUCT
tmHeight              DWORD ?
tmAscent              DWORD ?
tmDescent             DWORD ?
tmInternalLeading     DWORD ?
tmExternalLeading     DWORD ?
tmAveCharWidth        DWORD ?
tmMaxCharWidth        DWORD ?
tmWeight              DWORD ?
tmOverhang            DWORD ?
tmDigitizedAspectX    DWORD ?
tmDigitizedAspectY    DWORD ?
tmFirstChar           BYTE ?
tmLastChar            BYTE ?
tmDefaultChar         BYTE ?
tmBreakChar           BYTE ?
tmItalic              BYTE ?
tmUnderlined          BYTE ?
tmStruckOut           BYTE ?
tmPitchAndFamily      BYTE ?
tmCharSet             BYTE ?
TEXTMETRICA ENDS
DLGTEMPLATE STRUCT DWORD
style             DWORD ?
dwExtendedStyle   DWORD ?
cdit              WORD ?
x                 WORD ?
y                 WORD ?
lx                WORD ?
ly                WORD ?
DLGTEMPLATE ENDS
.const
MSSansSerif db "MS Sans Serif",0
DlgClassN db "MYDLGDlgClass",0
.code
; =============== S U B R O U T I N E =======================================
CREATEDIALOG proc near syscall public uses ebx edi esi win:DWORD
LOCAL lf :LOGFONT
LOCAL wc :WNDCLASSEX
LOCAL tm  :TEXTMETRICA
mov  eax, tm.tmAveCharWidth
mov  esi, SIZEOF(DLGTEMPLATE) + 21 + (SIZEOF DlgClassN)*2 + (SIZEOF MSSansSerif)*2
  ret  4
CREATEDIALOG endp
end


Compiles into with MASM:

Code Select Expand


public CREATEDIALOG
CREATEDIALOG proc near
push    ebp
mov     ebp, esp
add     esp, 0FFFFFF5Ch
push    ebx
push    edi
push    esi
mov     eax, [ebp-8Dh]
mov     esi, 61h
pop     esi
pop     edi
pop     ebx
leave
retn    4
CREATEDIALOG endp


With JWASM (/Zg):

Code Select Expand


public CREATEDIALOG
CREATEDIALOG proc near
push    ebp
mov     ebp, esp
add     esp, 0FFFFFF5Ch
push    ebx
push    edi
push    esi
mov     eax, [ebp-90h]
mov     esi, 5Fh
pop     esi
pop     edi
pop     ebx
leave
retn    4
CREATEDIALOG endp


Like I sad I am not sure that this is causing the problem but likelly.

[japheth]

Just slow down a little bit. :bg

I didn't say everything is ok.
Just sad that JWASM didn't report error. :P

Too late.

Seriously there is something not working right in my library.
A dialog won't open but no exception is generated.

You have to provide a full test case.

I did found something that is strange may causing the problem.
...
Like I sad I am not sure that this is causing the problem but likelly.

I doubt it. AFAICS the little difference between Masm's and JWasm's code generation is due to a small flaw in your definition of TEXTMETRICA. The correct, C-compatible size of this structure is to be 56, but yours is 53 only. You should add a DWORD alignment argument to the struct definition.

[Ficko]

I doubt it. AFAICS the little difference between Masm's and JWasm's code generation is due to a small flaw in your definition of TEXTMETRICA. ...

Yes I figured out that one.
But the second difference is more questionable.

I am building a dialoge template in memory and the calculations will be screwed up if I am getting different sizes.

Code Select Expand


mov  esi, SIZEOF(DLGTEMPLATE)


give me MASM equivalent size only without (/Zg)

Looks like the alignment

Code Select Expand


DLGTEMPLATE STRUCT DWORD

get ignored.

[japheth]

I am building a dialoge template in memory and the calculations will be screwed up if I am getting different sizes.

Code Select Expand


mov  esi, SIZEOF(DLGTEMPLATE)


give me MASM equivalent size only without (/Zg)

Looks like the alignment

Code Select Expand


DLGTEMPLATE STRUCT DWORD

get ignored.

DLGTEMPLATE is to be handled differently according to MS PSDK:

Code Select Expand


/*
* WARNING:
* The following structures must NOT be DWORD padded because they are
* followed by strings, etc that do not have to be DWORD aligned.
*/
#include <pshpack2.h>

/*
* original NT 32 bit dialog template:
*/
typedef struct {
    DWORD style;
    DWORD dwExtendedStyle;
    WORD cdit;
    short x;
    short y;
    short cx;
    short cy;
} DLGTEMPLATE;
typedef DLGTEMPLATE *LPDLGTEMPLATEA;
typedef DLGTEMPLATE *LPDLGTEMPLATEW;
...


So sizeof DLGTEMPLATE must be 18, not 20.

[japheth]

give me MASM equivalent size only without (/Zg)

I overread this sentence at first glance. While it is correct that sizeof DLGTEMPLATE is to be 18, JWasm's option /Zg must not have any impact on structure sizes.  However, in v2.01 - and also in v2.00 - it has an impact, which makes option /Zg unusable in these versions!

[Ficko]

I overread this sentence...

Damn!

You were faster I was almost ready with a "nasty" answer.  :green :bg

[japheth]

Damn!

You were faster I was almost ready with a "nasty" answer.  :green :bg

Don't worry! I'm known for arrogant and/or annoying replies, so I'm sure there will be good opportunities for "nasty" answers in the future.

there's a prelim. v2.02 now where's the -Zg switch has no impact on structure "padding":

http://www.japheth.de/Download/JWasm/JWasm202b.zip