[ansi 89] inline

[Charles Harrison Caudill]

Forgive the ignorance, I don't have k&r w/ me at the moment:

is there a way to do inline in ansi 89? 99?

 

[Régis Troadec]

Hi,

Forgive the ignorance, I don't have k&r w/ me at the moment:

is there a way to do inline in ansi 89? 99?

Yes in C99 with the inline keyword, no in C89.

Regards, Regis.

 

[Artie Gold]

Forgive the ignorance, I don't have k&r w/ me at the moment:

is there a way to do inline in ansi 89? 99?

C99 has `inline'. C89/90 does not.

The closest you can come to inlining in C89 is through the use of (some
often pretty ugly) macros, which you should *definitely* avoid if possible.

HTH,
--ag

[BTW - some C89 compilers offer `inline' as an extension.]

 

[Charles Harrison Caudill]

Forgive the ignorance, I don't have k&r w/ me at the moment:

is there a way to do inline in ansi 89? 99?

I was reading throught the GNU coding standards and I came across a good
suggestion for this problem:

#ifdef STD_C99_OR_WHATEVER
#define INLINE inline
#else
#define INLINE
#endif

 

[Gordon Burditt]

is there a way to do inline in ansi 89? 99?

In C89 it is not possible, in a standard way, to turn OFF inlining.

Gordon L. Burditt

 

[Dan Pop]

C99 has `inline'. C89/90 does not.

The closest you can come to inlining in C89 is through the use of (some
often pretty ugly) macros, which you should *definitely* avoid if possible.

Why avoid them? Function-like macros have been introduced into the
language for this very purpose.

Dan

 

[Dan Pop]

In C89 it is not possible, in a standard way, to turn OFF inlining.

It surely is: use a function pointer initialised in a different
translation unit.

Dan

 

[Old Wolf]

is there a way to do inline in ansi 89? 99?

It surely is: use a function pointer initialised in a different
translation unit.

Well, the linker could still inline the code. Also, the code
could be inlined at runtime (in a bytecode implementation),
or even by a CPU. More info:
http://www.cuj.com/documents/s=8906/cujexp0311sutter/sutter.htm

 

[Dan Pop]

Well, the linker could still inline the code. Also, the code
could be inlined at runtime (in a bytecode implementation),
or even by a CPU. More info:
http://www.cuj.com/documents/s=8906/cujexp0311sutter/sutter.htm

Not if you do it right:

In the translation unit containing the definition of foo:

typedef void func(void);
func *array[5] = {foo, foo, foo, foo, bar};

In the translation unit calling foo:

typedef void func(void);
extern func *array[5];
...
array[rand() % 4]();

Show me the system smart enough to realise that this expression can be
*always* substituted by a call to foo.

Dan

 

[Gordon Burditt]

In C89 it is not possible, in a standard way, to turn OFF inlining.

Well, the linker could still inline the code. Also, the code
could be inlined at runtime (in a bytecode implementation),
or even by a CPU. More info:
http://www.cuj.com/documents/s=8906/cujexp0311sutter/sutter.htm

Not if you do it right:

In the translation unit containing the definition of foo:

typedef void func(void);
func *array[5] = {foo, foo, foo, foo, bar};

In the translation unit calling foo:

typedef void func(void);
extern func *array[5];
...
array[rand() % 4]();

Show me the system smart enough to realise that this expression can be
*always* substituted by a call to foo.

Quote me the portion of the ANSI C standard that says it's not
allowed to make such a substitution.

As a practical matter, you've turned off inlining in at least 99.99%
of implementations. Probably 100% of existing ones. As a practical
matter, no existing implementation uses 97-bit longs. Nothing in
ANSI C prohibits either.

Gordon L. Burditt

 

[Dan Pop]

In C89 it is not possible, in a standard way, to turn OFF inlining.

It surely is: use a function pointer initialised in a different
translation unit.

Well, the linker could still inline the code. Also, the code
could be inlined at runtime (in a bytecode implementation),
or even by a CPU. More info:
http://www.cuj.com/documents/s=8906/cujexp0311sutter/sutter.htm

Not if you do it right:

In the translation unit containing the definition of foo:

typedef void func(void);
func *array[5] = {foo, foo, foo, foo, bar};

In the translation unit calling foo:

typedef void func(void);
extern func *array[5];
...
array[rand() % 4]();

Show me the system smart enough to realise that this expression can be
*always* substituted by a call to foo.

Quote me the portion of the ANSI C standard that says it's not
allowed to make such a substitution.

There is nothing preventing a conforming implementation from using ESP
(extra sensorial perception) in figuring out the program's output and only
generating the code involved in the generation of that output. It makes
precious little sense to talk about what an implementation can
*theoretically* do.

As a practical matter, you've turned off inlining in at least 99.99%
of implementations. Probably 100% of existing ones.

Looks good enough to me, considering that I only used standard code,
as required in the original statement, still included at the top of the
article.

Dan