[Bug target/83641] -fstack-clash-protection generates incorrect CFI on i386

[fw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83641

Florian Weimer <fw at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #42995|0                           |1
        is obsolete|                            |

--- Comment #1 from Florian Weimer <fw at gcc dot gnu.org> ---
Created attachment 42999
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=42999&action=edit
cfi.c

Reduced test case.  Compile with -O2 -fstack-clash-protection.

f3:
.LFB0:
        .cfi_startproc
        pushl   %esi
        .cfi_def_cfa_offset 8
        .cfi_offset 6, -8
        pushl   %ebx
        .cfi_def_cfa_offset 12
        .cfi_offset 3, -12
        pushl   %esi
        .cfi_def_cfa_offset 16
        .cfi_offset 6, -16
        popl    %esi
        .cfi_restore 6
        .cfi_def_cfa_offset 12
        subl    $4, %esp
        .cfi_def_cfa_offset 16
        movl    x0, %ebx
        movl    x1, %esi


It looks related to the probing for noreturn functions.  The noreturn attribute
can even be removed as long as the infinite loop at the end is present (which I
added to suppress the warning).