> From: "Jay K (jayk123@hotmail.com) [via djgpp@delorie.com]" <djgpp@delorie.com>
> Date: Wed, 26 Jan 2022 09:53:31 +0000
> suggested_attachment_session_id: 69785034-0787-e0af-85a1-df4fb0b67acf
>
> I am trying to port my program to MS-DOS. For fun, granted.
> I figured DJGPP would help.
> Esp. because it seems to have a useful "sigalarm" functionality,
> I might use to implement preemptive threads.
> I have a prototype that suggests that will work:
>
> https://github.com/modula3/cm3/blob/3be819ed8e2fa60885aff0cb6278934a17b30702/m3-libs/m3core/src/thread/POSIX/test_dos_threads.cpp
>
>
> The rest of the Posix support helps also.
> I'm not sure about network/sockets, maybe give up.
>
> I have experimented with a few "DOS"s.
> FreeDOS, DosBos-X, 32bit Windows 10 VM.
>
> They all have advantages and disadvantages.
>
> Granted I could spend more time configuring each and none is "really" MS-DOS.
>
> My program is a compiler, that produces .cpp files and would run g++ or gxx, etc.
> (I might experiment with e.g. OpenWatcom, might, like if it can link the djpp libraries.)
>
> I am seeing, I guess, command line length limitations.
>
> gcc and collect2 crash, but if I run them manually with shorter equivalent-enough commands, they work.
>
> Advise?
> [...]
> bash:
> ---------------------------
> 16 bit MS-DOS Subsystem
> ---------------------------
> Administrator: C:\Windows\system32\cmd.exe
> NTVDM has encountered a System Error
> NTVDM has encountered a System Error 4001h Choose 'Close' to terminate the application.
> ---------------------------
> Close Ignore
> ---------------------------

On what version of Windows is that?

> Exiting due to signal SIGSEGV
> Stack Fault at eip=00000a11

This is stack fault, not necessarily related to the command-line
length.

What does go32-v2 report there if run without arguments?

> eax=00010001 ebx=01ad0080 ecx=01549400 edx=004207bf esi=00001223 edi=00001213
> ebp=0042773a esp=0042773a program=C:\DOS\BIN\GCC.EXE
> cs: sel=00cf base=000d3a50 limit=00000db0
> ds: sel=00b7 base=000cc160 limit=0000ffff
> es: sel=0040 invalid
> fs: sel=0000
> gs: sel=0000
> ss: sel=00b7 base=000cc160 limit=0000ffff
> App stack: [00423b78..00223b7c] Exceptn stack: [00223ac8..00221b88]

Something is very wrong in the run environment: the SS segment has
very small limit, ES is invalid...

On 1/26/22 15:08, Eli Zaretskii (eliz@gnu.org) [via djgpp@delorie.com] wrote:
>
> On what version of Windows is that?
>
>> Exiting due to signal SIGSEGV
>> Stack Fault at eip=00000a11
> This is stack fault, not necessarily related to the command-line
> length.
>
> What does go32-v2 report there if run without arguments?
>
>> eax=00010001 ebx=01ad0080 ecx=01549400 edx=004207bf esi=00001223 edi=00001213
>> ebp=0042773a esp=0042773a program=C:\DOS\BIN\GCC.EXE
>> cs: sel=00cf base=000d3a50 limit=00000db0
>> ds: sel=00b7 base=000cc160 limit=0000ffff
>> es: sel=0040 invalid
>> fs: sel=0000
>> gs: sel=0000
>> ss: sel=00b7 base=000cc160 limit=0000ffff
>> App stack: [00423b78..00223b7c] Exceptn stack: [00223ac8..00221b88]
> Something is very wrong in the run environment: the SS segment has
> very small limit, ES is invalid...

Is that under Windows 10 32 bit?

GCC does not work in all recent builds of Win 10 32 bit.

This will be (stack trace above) how any attempt to execute DJGPP program from another one will end
in this case. It down not matter whether its executed directly or through some other non DJGPP
program or bat file. I do not know whether it is intentional or got broken accidentally. It is
perhaps extremely unlikely that it could be ever fixed.

Andris