Description

The C compiler issued with AT&T Unix.

This compiler doesn't have any identifying macros, nor seemingly do any header files provide any realistic means of identifying either the compiler or the underlying AT&T Unix.

I think it might be somewhat 'safe' to assume that an actual AT&T Unix system would be the "porting base" for that relevant version.

Porting Bases

1. There is a release of AT&T Unix System V Release 4 for 386, but I have not played around with that yet.
2. There is also a release of System V Release 4.0 for the 3B2.

Macros

As AT&T's compiler lacks any means of identification via preprocessor, we need alternates.

These are identifiers provided to identify the underlaying platform.

On SVR2, /lib/cpp defines vax and pdp11.
On SVR3, /lib/cpp defines u3b2, u3b5, and u3b20d.

The following are defined for cross-compilation environments: unix, gcos, ibm, DMERT.
The following are defined "for historical reasons" only: tss, os, mert, RT, RES.
The following are defined for both target machines: interdata, pdp11, vax, u370, u3b, u3b5, u3b2, u3b20d, ns32000, i386.

Example

Note: this is untested.

/*
 * Other Unix systems should be detected first, as all of them
 * will have 'unix' defined -- e.g. with this code, ULTRIX on VAX will
 * magically appear to be a System V which is wrong, as it's BSD-derived.
 */
#ifdef unix
# if defined(pdp11)
    printf("AT&T Unix System V Release 1\n");
# elif defined(vax) || defined(ns32000)
    printf("AT&T Unix System V Release 2\n");
# elif defined(u3b2)
    printf("AT&T Unix System V Release 3\n");
# elif defined(i386) || defined(sparc)
    printf("AT&T Unix System V Release 4\n");
# else
    printf("Probably isn't AT&T Unix.\n");
# endif
#endif

This seems to work on AT&T System V Release 3.2.3 for 3B2

$ uname -a; ./test
borman borman 3.2.3 3 3B2
AT&T Unix System V Release 3
$ 

It might seem strange to have code to detect a Unix system here under compiler macros, but this is probably the best way to handle the AT&T compiler. If you match AT&T Unix, then you can assume that the compiler is the AT&T compiler.

Notes

These sorts of shenanigans continues with all derivatives of System V Release 3, notably Xenix and SCO Unix.

SCO System V Release 5 also slightly conflates this. I think the mentality was "This is Unix, we do not need to define anything that identifies what sort of Unix, because this is Unix."