jbn@wdl1.UUCP (John B. Nagle) (06/14/84)
One problem with C is that there are not identified versions, such
as FORTRAN-77 or SMALLTALK-80. But there sure are different versions of
the language. At least the following are known to me, strictly from AT&T
Unix versions.
The language understood by the PWB UNIX compiler, about 1976.
(in which, for arrays, x = &x)
The language understood by the V7 UNIX compiler, about 1979.
(typing gets stronger, "short" appears, casts appear, etc.)
The language understood by the System V UNIX compiler, about 1983.
("unsigned char" appears, bit fields work)
I propose that we call these C-76, C-79, and C-83. Are there others?guy@rlgvax.UUCP (Guy Harris) (06/16/84)
> One problem with C is that there are not identified versions, such > as FORTRAN-77 or SMALLTALK-80. But there sure are different versions of > the language. At least the following are known to me, strictly from AT&T > Unix versions. > The language understood by the PWB UNIX compiler, about 1976. > (in which, for arrays, x = &x) > The language understood by the V7 UNIX compiler, about 1979. > (typing gets stronger, "short" appears, casts appear, etc.) > The language understood by the System V UNIX compiler, about 1983. > ("unsigned char" appears, bit fields work) > I propose that we call these C-76, C-79, and C-83. Are there others? "x" and "&x" for x an array is true for all versions of the compiler up to the System III compiler; the question is which compilers accept that construct without comment, which give a warning, and which (if any) reject it. The PWB/UNIX compiler had all the features listed there for the V7 compiler, as did the "Phototypesetter, Version Seven" compiler (which was almost the PWB/UNIX compiler; the source differed a little bit, but I don't remember whether the differences had any effect). I'd propose as versions of the language: C as of Sixth Edition UNIX - although this one is lacking things like "short" and "long", and casts, which makes writing portable code very difficult. (Actually, it did have "long", but it was undocumented, un-"supported", possibly buggy, and definitely incomplete - no long multiply/divide support.) C as of "Phototypesetter, V7"/PWB. This was the first "modern" C, in the sense of having most of the features we're used to, like "short", "long", "unsigned", casts, etc. I think that "#else" also appeared here first, and maybe macros with arguments. Bit fields originally appeared here, as did unions. The new forms of assignment operators ("+=" as opposed to "=+") and of initialization ("int i = 1;" rather than "int i 1") also appeared. C as of V7. This added structure assignment, structure-valued functions, and structure arguments, and it added enumerated data types. It also added several new preprocessor features, such as "#if" (which was, I think, available in PWB, but not in Photo V7) and the "__FILE__" and "__LINE__" builtins, and the "defined()" operator in "#if". C as of System III, which also came with 4.1BSD. This added "void" and "unsigned char" as official parts of the language (although I believe pre-S3 portable C compilers may have accepted them), and added new rules for structure member names making them local to the structure type rather than global. 4.1BSD also added long variable names. C as of System V (Release 1), which finally put the old assignment operators and initialization syntax to death. C as of System V (Release 2), which picked up Berkeley's long variable names. Bell's System V "Transition Aids" document incorrectly describes several changes that appeared in V7 and S3 as coming with S5. I don't deal here with versions of C not publicly released; given the multitude of UNIX development tracks within Bell, I suspect that story is more complicated than the one given here. Guy Harris {seismo,ihnp4,allegra}!rlgvax!guy