Tell HN: Compiling with Turbo C++ on MS-DOS 2.0 (inside Windows 95, via DOSBox-X)

I spent the last days rebuilding a very old toolchain stack, mostly out of curiosity and partly for education/reverse-engineering reasons.

The stack:

Turbo C++ (Borland)

Target: MS-DOS 2.0–era constraints

Running inside Windows 95

Emulated with DOSBox-X (not vanilla DOSBox)

Why DOSBox-X? Because it actually cares about:

real DOS boot flows

disk images (FAT12/FAT16) instead of fake mounts

Windows 9x compatibility

correct INT/DPMI/EMS/XMS behavior

Vanilla DOSBox is great for games, but it lies too much for this kind of experiment.

---

Why do this in 2026?

Because modern toolchains hide everything.

I wanted:

to see what the compiler actually emits

to understand what a “hello world” really costs

to watch how 16-bit memory models, segments, and relocation work

to verify that the binary actually runs under real DOS/Win9x expectations, not a modern CRT fantasy

Turbo C++ is perfect for this because:

it’s brutally honest

it emits readable ASM

it forces you to think in memory models (tiny/small/medium/large/huge)

it doesn’t protect you from yourself

---

Setup notes (important)

I’m not using mount c folder

I’m using a real HDD image (.img) formatted as FAT

DOSBox-X boots DOS → Windows 95 → Turbo C++

AUTOEXEC.BAT and CONFIG.SYS are real, not simulated

This matters if you care about:

file layout

executable headers

segment limits

boot realism

---

What surprised me

How little you need to get useful binaries

How fast Turbo C++ compiles compared to modern compilers

How transparent the whole pipeline is

How much knowledge we lost by abstracting this away

Modern dev feels like programming through 5 layers of noise. This feels like touching the metal — even in emulation.

---

Why this might interest HN

If you care about:

compiler education

retro-computing

OS boot chains

reproducible builds

understanding what your tools actually do

…this kind of setup is incredibly enlightening.

Not nostalgia. More like archaeology with instruments.

If there’s interest, I can follow up with:

exact DOSBox-X config

disk image layout

Turbo C++ version quirks

memory model comparisons

ASM output examples

Thanks for reading.