🚀 go-pugleaf

Please cite ANSI or any standard or any non-standard or any compiler
that accepts that code without complaints.

You probably cannot.  If you can, it is probably a broken compiler.

If you even tried gcc with -pedantic -ansi you would get errors:

t10.c: In function `main':
t10.c:4: warning: ANSI C forbids const or volatile functions

If you remove the volatile keyword, the code works.

Please see the documentation of volatile function in the gcc
distribution for exact semantics.

> Message-Id: <9204150101.AA19311@oz.NeXT.COM>
> Date: Tue, 14 Apr 92 18:01:38 PDT
> From: Bruce_Martin@next.com
> Received: by NeXT Mailer (1.63)
> To: bug-gcc@prep.ai.mit.edu
> Subject: GCC 2.1 codegen bug when function is declared volatile

> Gcc Version: 2.1
> Configuration: i386-mach
> Reported by: Bruce Martin (Bruce_Martin@NeXT.COM)

> Description: if a function is declared volatile, gcc 2.1 produces
> completely bogus code when the optimizer is used.  I'm not at all
> sure what the correct response to the following C code should be,
> but the current behavior is definitely wrong (I'm assuming that
> the lack of any error messages means that the compiler should
> succeed in the compilation).

> Things work correctly when the optimizer is not used.

> Test case:

> 	compiled with: cc1 -O -o foo.s foo.c

> foo.c:

> main()
> {
>     volatile double     result1, result2;
>     volatile double     one_over(double);

>     do {
>         result1 = one_over(12);
>         result2 = one_over(12);
>     } while (result1 == result2);
>     exit(0);
> }

> double
> one_over(       double  x)
> {
>     return (double)( 1.0 / x );
> }




> Result of the compilation (note the lack of any code in main to
> test the return value of one_over, or a call to exit()):

>         .file   "foo.c"
> gcc2_compiled.:
> .text
>         .align 2
> .globl _main
> _main:
>         pushl %ebp
>         movl %esp,%ebp
>         subl $16,%esp
>         call ___main
>         pushl $1076363264
>         pushl $0
>         call _one_over
>         fstp %st(0)
>         .align 2,0x90
>         .align 2
> LC0:
>         .double 0d1.00000000000000000000e+00
>         .align 2
> .globl _one_over
> _one_over:
>         pushl %ebp
>         movl %esp,%ebp
>         fldl LC0
>         fdivl 8(%ebp)
>         leave
>         ret
>         .align 2,0x90