[llvm-bugs] [Bug 36016] New: GlobalOpt emits debug fragments with wrong offsets for padded fields

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=36016

            Bug ID: 36016
           Summary: GlobalOpt emits debug fragments with wrong offsets for
                    padded fields
           Product: new-bugs
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: david.stenberg at ericsson.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 19712
  --> https://bugs.llvm.org/attachment.cgi?id=19712&action=edit
IR reproducer.

LLVM commit used: r322927.

When running:

  opt -globalopt -S foo.ll -o -

on the attached reproducer, a fragment expression with an offset of 32 bits is
created for the second field in the struct:

  !14 = !DIGlobalVariableExpression(var: !7, expr:
!DIExpression(DW_OP_LLVM_fragment, 32, 64))

However, as the field is padded up to 64 bits:

  !12 = !DIDerivedType(tag: DW_TAG_member, name: "bar_p", scope: !8, file: !3,
line: 3, baseType: !13, size: 64, offset: 64)

that fragment offset is incorrect.

The reproducer was created by compiling the following program, foo.c:

  struct {
    int foo;
    int *bar_p;
  } static var;

  int *ptr;

  int main() {
    if (var.foo == 0)
      var.bar_p = ptr;
    return 0;
  }

using:

  clang -march=x86-64 -O0 -g2 -S -emit-llvm foo.c -o foo.ll

-- 
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/20180119/b31e1913/attachment.html>