[Bug c/87085] New: with -march=i386, gcc should not generate code including endbr instruction

[chengming at bjuci dot com.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87085

            Bug ID: 87085
           Summary: with -march=i386, gcc should not generate code
                    including endbr instruction
           Product: gcc
           Version: 8.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: chengming at bjuci dot com.cn
  Target Milestone: ---

My gcc version is gcc (GCC) 8.2.1 20180801 (Red Hat 8.2.1-2), OS is Fedora 28.
I develop program for a CPU called Vortex86 on it. Vortex86 is an x86
artitecture CPU from Taiwan.

My problem is, even a program with only "return 0;", it will cause "Illegal
Instruction" on Vortex86. 

Based on dumped core file and objdump, the problem comes from the endbr32
instruction. 

Even though I add the -march=i386 and -fcf-protection=none option, it still
output endbr32 instruction.

So I think it's a bug. Because i386 CPU didn't support endbr32 instruction. If
I specified -march=i386, it should not generate endbr32 instruction.

There is no such problem on GCC 7 and below.

Thanks.