Finding a character in a string - strchr.

[jj2007]

While I really have no idea what I'm doing, I can assemble and link this code, and it's interesting to me that it fails so arbitrarily on my old computer.

To illustrate what these "unorthodox" jumps do, here a snippet:

Code Select Expand

include \masm32\include\masm32rt.inc

.code
start: mov ebx, -1
jmp @F+3 ; attention "artwork"
@@: nop
L0: inc ebx
L1: inc ebx
L2: inc ebx
L3: inc ebx
L4: inc ebx
L5: inc ebx
MsgBox 0, str$(ebx), "I hit label No. :", MB_OK
exit
end start

That code works also without the L1, L2 etc labels. Now what happens in real life is that
- the clever author of the code simply writes his code with "normal" labels, e.g. jxx MyGoodLabel
- then launches Olly or whatever to see what is the offset against a reference label (@@, i.e. the nop in the example above)
- replaces jmp MyGoodLabel with jmp @F+123
- deletes MyGoodLabel in the hope that who reads his code is too stupid to get the trick.
The real fun comes up when people use different assemblers, because in complex code the offsets *may* differ on rare occasions. If that happens, the jmp @F+123 ends in no man's land, and you have "arbitrarily" behaving source code that is meant to drive you insane.

[Queue]

No, the jumps don't fail for me. That is something I actually do understand about assembler and jumps in general. Actually, as an exercise, I put in the labels Lingo had chopped out so I could try and follow what was going on. Also took a stab at converting the SSE to MMX with... some degree of success. What I meant by it's failing is the SSE-related stuff isn't working properly on my old Athlon; I don't know enough about SSE and my processor to know why, it's just how it is.

Queue

[jj2007]

So which SSE level does the first line give you?

Intel(R) Celeron(R) M CPU        420  @ 1.60GHz (SSE3)

[Queue]

SSE1

[dedndave]

that could be the problem   :P
most of the SSE code in here requires at least SSE2 support

[clive]

To illustrate what these "unorthodox" jumps do, here a snippet:

Code Select Expand

include \masm32\include\masm32rt.inc

.code
start: mov ebx, -1
jmp @F+3 ; attention "artwork"
@@: nop
L0: inc ebx
L1: inc ebx
L2: inc ebx
L3: inc ebx
L4: inc ebx
L5: inc ebx
MsgBox 0, str$(ebx), "I hit label No. :", MB_OK
exit
end start


A quick static analysis suggests this doesn't work the way you wanted it too. L1 will report 4, L0 5.

[KeepingRealBusy]

The real fun comes up when people use different assemblers, because in complex code the offsets *may* differ on rare occasions. If that happens, the jmp @F+123 ends in no man's land, and you have "arbitrarily" behaving source code that is meant to drive you insane.

Is there any way to tell from examining the .exe which assembler created the .obj, and which loader was used?

[Queue]

that could be the problem   :P
most of the SSE code in here requires at least SSE2 support

Well, that answers that. Why does an SSE1 processor not just outright crash the program when SSE2 opcodes are used?

Queue

[lingo]

"Is there any way to tell from examining the .exe which assembler created the .obj"

Yes, if you have a MS "Rich" signature after DOS Header of your exe file, ONLY... :wink 

[jj2007]

A quick static analysis suggests this doesn't work the way you wanted it too. L1 will report 4, L0 5.

Thanks, Clive. Here is the correct version.

Code Select Expand

include \masm32\include\masm32rt.inc

.code
start: mov ebx, 7
jmp @F+5 ; attention "artwork"
@@: nop
L0: dec ebx
L1: dec ebx
L2: dec ebx
L3: dec ebx
L4: dec ebx
L5: dec ebx
MsgBox 0, str$(ebx), "I hit label No. :", MB_OK
exit
end start


[Queue]

"Is there any way to tell from examining the .exe which assembler created the .obj"

Yes, if you have a MS "Rich" signature after DOS Header of your exe file, ONLY... :wink 

The Rich header is added by the Microsoft linker. I know of some differences in structure between linkers, but not between assemblers. For example, if the code and import sections are merged, that results in a different layout between MS's LINK and POLINK. The resource secton's organization can vary based on how the RES file was converted to object code and by which linker. Different linkers also tend to use slightly different section flags, especially when merging sections.

Queue

[jj2007]

I know of some differences in structure between linkers, but not between assemblers.

In general, there are no differences in the code generated by various versions of MASM, from ml v6.15 onwards.
Jwasm generates something else for .if eax, but that is not particularly relevant. However, there is one little bug that can make a difference for the jmp @F+123 "artwork": Jwasm generates long jumps in situations where a short jump would be sufficient. That can be cumbersome indeed :bg

[lingo]

"The Rich header is added by the Microsoft linker"

Congratulation, you just reinvent the wheel....
Will be better to tell as why and how to use it?
For your info Rich is just a part of PRODITEM structure for "end of tallies" rather than a "header"...  :wink

[Queue]

"The Rich header is added by the Microsoft linker"

Congratulation, you just reinvent the wheel....
Will be better to tell as why and how to use it?
For your info Rich is just a part of PRODITEM structure for "end of tallies" rather than a "header"...  :wink

Wait, what? You said a way to tell the difference between assemblers is the presence of the Rich header. I simply pointed out it's the linker that adds the Rich header, not the assembler.

The Rich header is used to identify which versions of various MS utilities were used to make the executable.
http://ntcore.com/files/richsign.htm

Queue

[KeepingRealBusy]

I wonder how many hits that site will get today?

Interesting read.

Dave.