DOS ain't dead


> D:\scratch\zzz>hexdump \msvc15\bin\cl.exe | more
> 000000  4D5A8000 33000E00 08004102 FFFF6106  MZ..3.....A...a.
> 000010  00200000 18212000 40000000 00000000  . ...! .@.......
> 000020  00000000 00000000 00000000 00000000  ................
> 000030  00000000 00000000 00000000 80660000  .............f..

The PE signature is at 0x6680.

> I thought that Windows 10 had no 16-bit support and
> only ran PE executables.

Win 10 (32-bit) used to have NTVDM. But Win 11 is x64 (or ARM64) only.

> Also it is unclear to me whether my CDROM is 1.52
> or 1.52c. I don't know how to tell the difference.

Dunno, check "_MSC_VER" and/or file dates and times.

> Also, I don't know how to assemble that assembler.
>
> D:\scratch\zzz>ml foo.asm
> Microsoft (R) Macro Assembler Version 6.15.8803
> Copyright (C) Microsoft Corp 1981-2000. All rights reserved.

Try "/Zm".

> I was surprised that masm doesn't
> appear to come with Visual Studio 1.52.
> Although checking the physical CD I don't see
> any mention of "professional", so maybe that
> is why.

You can always use Japheth's JWasm.

> Visual C++ 1.0 came with Phar Lap 286.

It's just a "pharlap lite" version, limited to 2 MB, IIRC - hardly useful for anything.

> No mention of whether 1.52 switched to Phar Lap 386.

No.

> But unfortunately the method of doing huge pointers
> effectively hardcodes a 4-bit shift into executables.

VC 1.52 uses "external" constants _AHINCR and _AHSHIFT, that work for both real-mode and protected-mode ( assuming that the protected-mode OS uses "selector tiling" for blocks > 64 kB ).

> Also, I don't know how to assemble that assembler.

The -Fa option generates assembly code, but without "red tape" cmds, so the generated code has to be adjusted before feeding the assembler with it.

---
MS-DOS forever!

In the full pharlap, they have their own assembler

I suspect you'd need that.

> Microsoft C++ 5.1 which I also recently received on floppy
> disks has huge pointer support, but not huge memory model.
>
> ie I have to use the "huge" keyword everywhere.

I don't remember what compiler options I was using
when doing that previous test - or maybe I was
drunk as a skunk - but I repeated the test as part
of my vlog on pdos.org (I bought a Book 8088 with
an NEC V20 (so I can do 8080 (8-bit) with some effort)
running msdos 6.22) and was shocked to see an
unexpected result - huge memory model works!!!

I have repeated the test under Win95, but I was not
able to redirect the stderr, so can't capture the
version number (5.1) and Bochs wouldn't do a
character screenshot either, so you'll have to trust
me that I didn't manage to stuff up the number 5.1.
Anyway - it works!

Here's the rest of what I ran:

D:\scratch\xxx>type doit.bat
del foo.asm >>output.txt
type foo.c >>output.txt
type foo.asm >>output.txt
cl /c /AH /Fa /Gs foo.c >>output.txt
type foo.asm >>output.txt

char *foo(char *a)
{
return (a + 5);
}
foo.c
; Static Name Aliases
;
TITLE foo.c
NAME foo

.8087
FOO_TEXT SEGMENT WORD PUBLIC 'CODE'
FOO_TEXT ENDS
_DATA SEGMENT WORD PUBLIC 'DATA'
_DATA ENDS
CONST SEGMENT WORD PUBLIC 'CONST'
CONST ENDS
_BSS SEGMENT WORD PUBLIC 'BSS'
_BSS ENDS
DGROUP GROUP CONST, _BSS, _DATA
ASSUME CS: FOO_TEXT, DS: DGROUP, SS: DGROUP
EXTRN __acrtused:ABS
EXTRN __AHSHIFT:FAR
FOO_TEXT SEGMENT
ASSUME CS: FOO_TEXT
; Line 2
PUBLIC _foo
_foo PROC FAR
push bp
mov bp,sp
; a = 6
; Line 3
mov ax,5
cwd
add ax,WORD PTR [bp+6] ;a
adc dx,0
mov cx,OFFSET __AHSHIFT
shl dx,cl
add dx,WORD PTR [bp+8]
pop bp
ret
nop

_foo ENDP
FOO_TEXT ENDS
END


Doing cl /help shows:

C COMPILER OPTIONS

-MEMORY MODEL-
/AS small model (default) /AC compact model
/AM medium model /AL large model
/AH huge model
-OPTIMIZATION-
/O enable optimization (same as /Ot) /Oa ignore aliasing
/Od disable optimizations /Oi enable intrinsic functions
/Ol enable loop optimizations /On disable "unsafe" optimizations
/Op enable precision optimizations /Or disable in_line return
/Os optimize for space /Ot optimize for speed (default)
/Ox max. optimization (/Oailt /Gs)
-CODE GENERATION-
/G0 8086 instructions (default) /G1 186 instructions
/G2 286 instructions /Gm put strings in constant segment
/Gc Pascal style function calls /Gs no stack checking
/Gt[number] data size threshold
-OUTPUT FILES-
/Fa[assembly listing file] /Fb[bound executable file]
/Fc[mixed source/object listing file] /Fe<executable file>
/Fl[object listing file] /Fm[map file]
/Fo<object file> /Fs[source listing file]
(press <return> to continue)
-PREPROCESSOR-
/C don't strip comments /D<name>[=text] define macro
/E preprocess to stdout /EP same as /E but no #line
/I<name> add #include path /P preprocess to file
/U<name> remove predefined macro /u remove all predefined macros
/X ignore "standard places"
-LANGUAGE-
/Za disable extensions /Zd line number information
/Ze enable extensions (default) /Zg generate declarations
/Zi symbolic debugging information /Zl remove default library info
/Zp[n] pack structs on n-byte boundary /Zs syntax check only
-FLOATING POINT-
/FPa calls with altmath /FPc calls with emulator
/FPc87 calls with 8087 library /FPi inline with emulator (default)
/FPi87 inline with 8087
-SOURCE LISTING-
/Sl<columns> set line width /Sp<lines> set page length
/St<string> set title string /Ss<string> set subtitle string
-MISCELLANEOUS-
/c compile only, no link /H<number> external name length
/J default char type is unsigned /Tc<file> compile file without .c
/V<string> set version string /W<number> warning level
(press <return> to continue)
-LINKING-
/F<hex_number> stack size (hex. bytes) /Lc link compatibility mode executable
/Lr link compatibility mode executable /Lp link protect mode executable
/link [linker_options_and_libraries]


but the version number must have gone to stderr
there too.

BFN. Paul.