[llvm-bugs] [Bug 26389] New: [x86-64] clang generate wrong instruction for cygwin

via llvm-bugs llvm-bugs at lists.llvm.org
Fri Jan 29 20:36:34 PST 2016


            Bug ID: 26389
           Summary: [x86-64] clang generate wrong instruction for cygwin
           Product: clang
           Version: 3.8
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: LLVM Codegen
          Assignee: unassignedclangbugs at nondot.org
          Reporter: swigger at gmail.com
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

when access data that may be far away (than 4G) on x86-64, clang generate a
wrong instruction cause program fail to run (Segment fault).

sample code:

struct OSTREAM { char buf[100]; } ;
extern OSTREAM cout;

extern void show(void*);
int main()

command: clang -target x86_64-pc-cygwin -S -o -  ext.cpp

        .def     main;
        .scl    2;
        .type   32;
        .globl  main
        .align  16, 0x90
main:                                   # @main
.seh_proc main
# BB#0:
        pushq   %rbp
        .seh_pushreg 5
        subq    $32, %rsp
        .seh_stackalloc 32
        leaq    32(%rsp), %rbp
        .seh_setframe 5, 32
        callq   __main
        leaq    cout(%rip), %rcx     <=============== WRONG!!
        callq   _Z4showPv
        xorl    %eax, %eax
        addq    $32, %rsp
        popq    %rbp

look at the leaq instruction, when cout is far than 4G away, this will take an
invalid address and then cause a segment fault.

If target is x86_64-pc-linux-gnu, the generated instruction is movabsq that

This is a reduced test case. I can show it on clang 3.8 on ubuntu. the orignal
case is that I found clang compiled program failed to run cout<<"hello"<<endl
on 64 bits msys2 on windows7 x64.

$ clang --version
Ubuntu clang version 3.8.0-svn257311-1~exp1 (trunk) (based on LLVM 3.8.0)
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/lib/llvm-3.8/bin

You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20160130/c89ca47a/attachment-0001.html>

More information about the llvm-bugs mailing list