NOTES ON NESTING OF PREPROCESSOR DIRECTIVES IN BDS C.


Versions of the BDS C compiler from 1.45a+ do support
nesting of the preprocessor directives.

Leor specifies that  #if <exp> compiler directive is
supported with restrictions on <exp>, and that nesting of
conditional compile directives is permitted with the
limitation that #else directives must be followed by a
matching #endif.  This last point is a little unclear to me
but 'SEEMS' to say that

#ifdef	A

#ifdef	B
#else	/*B*/
#endif	/*B*/
#endif	/*A*/

    is permitted, but

#ifdef	A

#else	/*A*/
#ifdef	B
#endif	/*B*/
#endif	/*A*/

    is not.???

    He also states that the preprocessor does now generate a
valid error, rather that wandering off through source.

    Also parameterized #defines are supported as per K & R.

    K & R page 207pp indicates that both forms of the #if...
nesting above should be legal.	He implies that a
conditional compile directive (#if...) MUST be followed by
an #endif, but may have an optional #else before the #endif.
Where nesting is permitted (no restrictions) and #else's act
upon the most recently active #if..., #endif's act upon the
most recently active #if... or #else.

i.e.

#if...	A
.
.
#if...	B
.
.
#else	/*B*/
.
#if...	C
.
.
#else	/*C*/
.
.
#endif	/*C*/
.
.
#endif	/*B*/
.
.
#endif	/*A*/

Š

		is o.k.


    On testing of the compiler with (granted) a fairly
simple test for compiler nesting, the BDS C compiler 1.46 seems
to behave exactly as K & R describes, and as I would expect
it to.	When I tried to upset it, it upset me with 'Not in a
conditional block', or EOF when expecting #endif etc error
messages.  #defines that exist in an unenabled block of code
do not get seen at all, so nesting of #defines inside
#if...'s works as expected as well.


    If you know of any examples of nesting that do not work
in 1.46, or know any other (possibly clearer) descriptions
on the operation and nesting of these directives in C please
let me know.

    Hope this helps clear up a little of the confusion.


		    John Woolner.