Combining Identical String Literal

[msqweasm]

If I have something that C++ would call it "string literal" in my src file,  I know I should define it like this:

Code Select Expand

.data
mymsg DB 'hello world', 0aH, 00H
My code later on will print this hello world message using the mymsg label.  But if I have many src files each is having the same hello world "string literal".  There would be many duplicated const string.  Is it possible for me to instruct the linker (MS Visual C++ 2010) to combine them automatically into 1 constant string?  C++ compiler is able to do this automatically.  How do we do this automatically in assembly?  I don't want to manually verify the messages are the same and manually reference the same external label.  I just want the linker to recognize identical constant strings and combine them into 1.

[hutch--]

Nope. You do as you have shown, make a single common strin then reference it by its address from as many places as you need.

[jj2007]

Assemble this as MyStrings.asm to MyStrings.obj:

Code Select Expand

.386
.model flat

ExternDef _AppName:BYTE
ExternDef _Hello:BYTE

.data
_AppName db "Masm32:", 0
_Hello db "Hello World", 0

.code
dummy:
end dummy


Call it like this, adding MyStrings.obj to the linker's commandline:

Code Select Expand

include \masm32\include\masm32rt.inc

ExternDef AppName:BYTE
ExternDef Hello:BYTE

.code
start: MsgBox 0, addr Hello, addr AppName, MB_OK
exit

end start


[drizz]

You must use macros for this. There are ready made macros by Four-F, "Strings.mac" part of KmdKit.
http://www.freewebs.com/four-f/

The following macros try to eliminate duplicate strings. So only a single copy of identical strings
will present in the program image, resulting in smaller programs.

$TA / $CTA / $TW / $CTW / $T / $CT
$TA0 / $CTA0 / $TW0 / $CTW0 / $T0 / $CT0

Every time you define a string using one of the above listed macros the string is stored in database.
If somewhere later in your code you use the very same macro to define the very same string, the macro
will remember its offset from the database instead of defining it second time.

[jj2007]

You must use macros for this.

MasmBasic does it by default for some functions:

include \masm32\MasmBasic\MasmBasic.inc   ; Download
   Init
   .if Exist("\Masm32\include\Windows.inc")
      MsgBox 0, "\Masm32\include\Windows.inc", "Load this file?", MB_YESNO
      .if eax==IDYES
         Recall "\Masm32\include\Windows.inc", L$()
         Print Str$("%i lines read", eax), CrLf$
         Print Str$("%i bytes read", Lof("\Masm32\include\Windows.inc")), CrLf$, CrLf$
         For_ n=0 To 9
            .if Len(L$(n))
               PrintLine Str$(n), Tb$, L$(n)
            .endif
         Next
      .endif
   .endif
   Inkey Str$("\n\n%i identical strings condensed to one - launch Olly and look for 'push 404000' ...", found+1)
   Exit
end start

Code Select Expand

4 identical strings condensed to one - launch Olly and look for 'push 404000' ...

[xandaz]

   Lol... dummy as a starting point.

[jj2007]

   Lol... dummy as a starting point.

It won't assemble without a .code section... if anybody knows how to create a pure .data object file, please show up :thumbu

[hutch--]

JJ,

FDA and FDA32.

[msqweasm]

I've done some research and found that MS VC++ combine identical strings by placing some information into something called COMDAT section.  It is the linker that combines the strings using the COMDAT section.  I don't know albout all the details yet.  But does ML.exe support this COMDAT section thing?

[dedndave]

that might cause problems
what if 2 strings are identical at assembly-time, but one of them gets modified during execution
the linker would have no way of knowing this

[jj2007]

I've done some research and found that MS VC++ combine identical strings by placing some information into something called COMDAT section.  It is the linker that combines the strings using the COMDAT section.

See above, reply #2, adding MyStrings.obj to the linker's commandline

[drizz]

I've done some research and found that MS VC++ combine identical strings by placing some information into something called COMDAT section.  It is the linker that combines the strings using the COMDAT section.  I don't know albout all the details yet.  But does ML.exe support this COMDAT section thing?

I think it doesn't. You would still need to use macros if it did.
There is an interesting topic about it here.
Unfortunately the tool mentioned for marking comdat is not available anymore:
http://web.archive.bibalex.org/web/20040609211043/http://launcherasm.com/technical/comdats.html

Japheth is probably the best person to ask about this. Feel free to start a discussion on comdat segments for jwasm here

It won't assemble without a .code section... if anybody knows how to create a pure .data object file, please show up :thumbu

You don't need entry point labels for every module. You mustn't put any when linking with C runtime.

This is is the smallest compilable asm file :

Code Select Expand

end

[sinsi]

One way to do it is to create a seperate .asm for each string, assemble to a .obj then add each .obj to a .lib
Link with the .lib and the linker will pick out the strings you want.