[Bug c/63318] Hello World C program using inline assembly to invoke write(2) on amd64 Linux fails to print Hello World

[ryao at gentoo dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63318

Richard Yao <ryao at gentoo dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|INVALID                     |---

--- Comment #4 from Richard Yao <ryao at gentoo dot org> ---
Neither of the versions that you suggest work with my local GCC installation.
The first fails to compile, but will compile if I 's/string/hello/'. The second
compiles. Neither result in the compiler emitting "Hello World!\n". Clang does
work in all cases, except for the first before the correction is made.

Here is the output for the second version from GCC:

$ gcc -S -o - syscall.c -Wall -Wextra -fno-strict-aliasing -fwrapv
        .file   "syscall.c"
        .text
        .globl  main
        .type   main, @function
main:
.LFB0:
        .cfi_startproc
        pushq   %rbp
        .cfi_def_cfa_offset 16
        .cfi_offset 6, -16
        movq    %rsp, %rbp
        .cfi_def_cfa_register 6
        subq    $48, %rsp
        movq    %fs:40, %rax
        movq    %rax, -8(%rbp)
        xorl    %eax, %eax
        movabsq $8022916924116329800, %rax
        movq    %rax, -32(%rbp)
        movl    $560229490, -24(%rbp)
        movw    $10, -20(%rbp)
        movq    $14, -48(%rbp)
        leaq    -32(%rbp), %rax
#APP
# 8 "syscall.c" 1
        movl $1, %eax
        movl $1, %edi
        movq %rax, %rsi
        movl -48(%rbp), %edx
        syscall
# 0 "" 2
#NO_APP
        movq    %rax, -40(%rbp)
        movl    $0, %eax
        movq    -8(%rbp), %rdx
        xorq    %fs:40, %rdx
        je      .L3
        call    __stack_chk_fail
.L3:
        leave
        .cfi_def_cfa 7, 8
        ret
        .cfi_endproc
.LFE0:
        .size   main, .-main
        .ident  "GCC: (Gentoo 4.8.3 p1.1, pie-0.5.9) 4.8.3"
        .section        .note.GNU-stack,"",@progbits

It is clear that this used to work. If the failure is because the compiler has
become more strict, then it should be possible to rewrite this in a way that
produces a working program. If we cannot, I think this would qualify as a
regression.

Gentoo does not yet package newer versions of GCC, so it is possible that this
is fixed in a newer version. If that is the case, I would like confirmation of
that. I am setting this back to Unconfirmed lest that it be forgotten before we
are on the same page.