[9fans] PCC - #if

[9fans] PCC - #if

[lucio]

I have the following diff:

	term% diff ../../../openldap-2.3.32/include/ac/time.h ../../include/ac
	20c20
	< #if defined(TIME_WITH_SYS_TIME)
	---
	> #if TIME_WITH_SYS_TIME
	23c23
	< #elif defined(HAVE_SYS_TIME_H)
	---
	> #elif HAVE_SYS_TIME_H

and I'm curious whether it's a slip in the OpenLDAP source or in PCC
that the one works and the other doesn't.  I do not have an ANSI
standard to check against, but my gut feel is that OpenLDAP is
stretching a limit.  If not, then PCC ought to be corrected, it
reports (the "elsif" is certainly mistaken):

	cpp: ../../include/ac/time.h:20 csn.c:47 Syntax error in #if/#elsif

++L

Re: [9fans] PCC - #if

[Russ Cox]

	term% diff ../../../openldap-2.3.32/include/ac/time.h ../../include/ac
	20c20
	< #if defined(TIME_WITH_SYS_TIME)
	---
	> #if TIME_WITH_SYS_TIME
	23c23
	< #elif defined(HAVE_SYS_TIME_H)
	---
	> #elif HAVE_SYS_TIME_H

The < version is always valid; the > version is often valid.

After

	#define A 1
	#define B 0
	#define C /* nothing! */
	#undef D

These are like #if 1:

	#if defined(A)
	#if defined(B)
	#if defined(C)
	#if A

These are like #if 0:

	#if defined(D)
	#if B
	#if D

and this is invalid:

	#if C

(The rule for variables appearing in #if is that macros
expand and names that make it through macro expansion
turn into zeros, so `#if D' is like `#if 0' but `#if C' is like `#if',
which is missing an expression.)

So if the config.h for this program has done, say,

	#define TIME_WITH_SYS_TIME 1
	#define HAVE_SYS_TIME_H 1

then the two versions you have are equivalent, but
if it has done

	#define TIME_WITH_SYS_TIME
	#define HAVE_SYS_TIME_H

then only the #if defined(...) versions are valid C.
The convention in autoconf etc. is to either
#define X 1 or #undef X for each variable X.

Russ

Re: [9fans] PCC - #if

[lucio]

> The convention in autoconf etc. is to either
> #define X 1 or #undef X for each variable X.

So if I have

	#undef	C

followed by

	#if		C

is it valid or invalid?  Your explanation leaves it unclear (or I
missed it, somehow).

++L

PS: CPP still needs "elsif" changed to "elif" in the error message.
Shall I submit a patch?

Re: [9fans] PCC - #if

[geoff]

No need for a patch; there's updated cpp source on sources now.

Re: [9fans] PCC - #if

[Russ Cox]

> So if I have
> 
> 	#undef	C
> 
> followed by
> 
> 	#if		C
> 
> is it valid or invalid?  Your explanation leaves it unclear (or I
> missed it, somehow).

It is valid.  `#if C' macro-expands to `#if C' (no change)
and then all names in #if expressions turn into 0s, so it
is treated as `#if 0'.

Russ

[9fans] PCC - #if

[lucio]

I have the following diff:

	term% diff ../../../openldap-2.3.32/include/ac/time.h ../../include/ac
	20c20
	< #if defined(TIME_WITH_SYS_TIME)
	---
	> #if TIME_WITH_SYS_TIME
	23c23
	< #elif defined(HAVE_SYS_TIME_H)
	---
	> #elif HAVE_SYS_TIME_H

and I'm curious whether it's a slip in the OpenLDAP source or in PCC
that the one works and the other doesn't.  I do not have an ANSI
standard to check against, but my gut feel is that OpenLDAP is
stretching a limit.  If not, then PCC ought to be corrected, it
reports (the "elsif" is certainly mistaken):

	cpp: ../../include/ac/time.h:20 csn.c:47 Syntax error in #if/#elsif

++L