Backward compatibility

[Klod]

This macro works fine with version 8 but not with nasm-2.09rc4

%imacro uses 1-*.nolist
   %ifctx proc
      %push uses
      %assign %$j 0
      %rep %0
            %assign %$j %$j+1
            %xdefine %$reg_%$j %1
            push dword %1
            %rotate 1         
      %endrep
   %elifctx uses
      %rep %0
         %assign %$j %$j+1
         %xdefine %$reg_%$j %1
         push dword %1
         %rotate 1
      %endrep
   %else
      %error "'USES' must be used inside 'PROC/ENDPROC'"
   %endif
%endmacro

It produces error:`%xdefine' expects a macro identifier

Isn't %define and %xdefine the ways to define single line macros?

regards

Klod

[Frank Kotler]

Try 2.09, Klod - seems to work for me. Temporary glitch in 2.09rc4?

Best,
Frank

[Frank Kotler]

Whoops, I spoke too soon. If I actually "%push proc", I do get that message. Doing:

Code: [Select]

            %xdefine %$reg_%+%$j %1

seems to work...

O'course, Keith is, how's the manual put it? "improve it out of all recognition" to the preprocessor, so glitches could crop up, as it develops. I think "backward compatibility" is intended - except for that ill-advised "look in outer contexts for context-specific variables" misfeature. 2.09 is supposed to warn, if that change is about to bite ya.

Best,
Frank

[Cyrill Gorcunov]

yeah, please try 2.09 and if it fail -- please post the caller code not only macro itself, thanks.

[Klod]

I did compile with 2.09 and I do get the same error. I had compiled with both 2.09rc4 and 2.09rc7 with the same error.

I'm including here a copy of the program with the uses macro commented out and substituted with push/pop registers, it compiles and runs fine. This macro by the way I have adopted from one of Bryant Keller's post from a few years ago as are some of the other macros used in this example

See attached project
I compile with \nasm\bin\gorc,,,  \nasm\bin\NASM -f win32,,,   \nasm\bin\golink  @\nasm\bin\GFL.txt /entry Start

The errors I get:
\nasm\bin\NASM -f win32  "C:\NasmEd\Project\WorkInProgress\LatestExamples\BasicWindow2\BasicWindow.asm"
C:\NasmEd\Project\WorkInProgress\LatestExamples\BasicWindow2\BasicWindow.asm:82: error: `%xdefine' expects a macro identifier
C:\NasmEd\Project\WorkInProgress\LatestExamples\BasicWindow2\BasicWindow.asm:106: error: comma, colon or end of line expected
C:\NasmEd\Project\WorkInProgress\LatestExamples\BasicWindow2\BasicWindow.asm:106: error: comma, colon or end of line expected

This is Line 82:  uses ebx,edi
This is line 106:  ENDP
And this is the offending statement: %xdefine %$reg_%$j %1

I wonder if this topic is not related??
http://forum.nasm.us/index.php?topic=863.0

Possibly not the correct context?

Regards
Klod

[Keith Kanios]

See attached project

The first thing I see is...

Code: [Select]

%include '\nasm\inc\nasmmacros.asm'
%include '\nasm\inc\windows.inc'

... in which please provide those exact files as you have them, as a reply to this thread.

[Frank Kotler]

I don't think this is a "context" issue. I think it's a "token pasting"(?) issue. Both "$reg_ " and "%$j" are okay (I think) - both defined in this context. At one time (under some circumstances, at least) you could just jam 'em together:

Code: [Select]

%xdefine %$reg_%$j %1

"%+" has existed for some time as the "prefered"(?) way to do this...

Code: [Select]

%xdefine %$reg_%+%$j %1

Apparently, now you "have" to do it that way. I don't know if this is a "bug" or not. Since "%+" exists, I just figure "that's the way you're supposed to do it"... but it *is* a change...

At least, that works in my "test" code. Can you confirm whether that works or not in its "native habitat", Klod?

How "should" 2.10 do it, is the question...

Best,
Frank

[Keith Kanios]

I tend to use Macro Indirection for complex cases.

[Klod]

Thank you all for your replies.

Frank, I think you are correct. I had tried  %+ before and I still got compile errors and had therefore discounted this possibility. I have played around a bit and this is what I have come up.

Your suggestion %xdefine %$reg_%+%$j %1  solves the define issue.

BasicWindow.asm:82: warning: 1 ebx1
BasicWindow.asm:82: warning: 2 edi2
BasicWindow.asm:82: warning: 3 esi3

However there is one more issue:
This is in the ENDP macro, when restoring the registers
BasicWindow.asm:107: error: comma, colon or end of line expected
This line: pop  dword %$reg_ %+ %$j

BasicWindow.asm:107: warning: 3 esi3 3
BasicWindow.asm:107: warning: 3 esi2 2
BasicWindow.asm:107: warning: 3 esi1 1

I don't quite understand the leading number but assume this to be an internal counter??

Keith, sorry for uploading an incomplete project, I was tired and rushed through the "packaging". Here is a complete copy that should allow you to compile.

Regards
Klod

[Keith Kanios]

Leaning towards a bug in 2.09+.

%xdefine %%reg_%$j %1 works, as expected. Using the context-local counterpart should as well, but it doesn't.

%xdefine %{${reg_%$j} %1 works, and it shouldn't... notice the missing bracket.

Something is amiss.

[Klod]

I have been using a helper macro to circumvent this issue:

Code: [Select]

%macro def 2
        %xdefine %1 %2
%endmacro

def %$reg_%+%$j,%1

This seems to work with all versions of Nasm I have tested.

Regards,
Klod