Bug#900437: Erroneous generated code

[Mario Xerxes Castelán Castro <marioxcc.MT@yandex.com>]

Package: gcc-6
Version: 6.3.0-18+deb9u

GCC produces incorrect code for the following program. The C11 standard
requires (§ F.8.5) that the floating point exception flag is set by the
following code, yet the generated assembly does not set it. These
results are for an x86-64 system running Debian 9.

The interaction is:

“““
$ gcc -W -Wall -S -std=c11 -O3 -march=native -masm=intel -fverbose-asm
test.c
test.c:2:0: warning: ignoring #pragma STDC FENV_ACCESS [-Wunknown-pragmas]
 #pragma STDC FENV_ACCESS ON

test.c: In function ‘main’:
test.c:11:9: warning: unused variable ‘x’ [-Wunused-variable]
   float x = 0.0 / 0.0;
         ^
”””

The file “test.c” is:

“““
#include <fenv.h>
#pragma STDC FENV_ACCESS ON

#if !defined(__STDC_IEC_559__)
#error "Requires IEC 60559 compliance"
#endif

int f();

int main(void) {
  float x = 0.0 / 0.0;
  f();
  return 0;
}
”””

The output file “test.s” is (excluding comments) is:

“““
	.section	.rodata.str1.1,"aMS",@progbits,1
.LC0:
	.string	"Hello world!"
	.section	.text.startup,"ax",@progbits
	.p2align 4,,15
	.globl	main
	.type	main, @function
main:
.LFB5:
	.cfi_startproc
	sub	rsp, 8	#,
	.cfi_def_cfa_offset 16
	xor	eax, eax	#
	call	f@PLT	#
	lea	rdi, .LC0[rip]	#,
	xor	eax, eax	#
	call	printf@PLT	#
	xor	eax, eax	#
	add	rsp, 8	#,
	.cfi_def_cfa_offset 8
	ret
	.cfi_endproc
.LFE5:
	.size	main, .-main
	.ident	"GCC: (Debian 6.3.0-18+deb9u1) 6.3.0 20170516"
	.section	.note.GNU-stack,"",@progbits
”””