[LLVMbugs] [Bug 19778] New: asm label with the same name as static variable leads to: Assertion failed: (Symbol->isUndefined() && "Cannot define a symbol twice!")

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=19778

            Bug ID: 19778
           Summary: asm label with the same name as static variable leads
                    to: Assertion failed: (Symbol->isUndefined() &&
                    "Cannot define a symbol twice!")
           Product: new-bugs
           Version: trunk
          Hardware: All
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: dimitry at andric.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

I received a bugreport about the following assertion failure in clang, when
building libglapi (a component of Mesa):

https://jenkins.freebsd.org/pci/head-i386/logs/bulk/headi386-default/753/logs/errors/libglapi-9.1.7.log

After some reducing, the testcase turns out to be very small:

__asm__("x86_entry_end:");
static x86_entry_end[];
entry_generate() {
  int *code_templ = x86_entry_end - 1;
  memcpy(0, code_templ, 32);
}

Compiling this with clang trunk r209086 (with assertions enabled), and any
optimization level above -O1, gives:

Assertion failed: (Symbol->isUndefined() && "Cannot define a symbol twice!"),
function EmitLabel, file /share/dim/src/llvm/trunk/lib/MC/MCAsmStreamer.cpp,
line 309.
[... etc ...]

The code in question originates here in the Mesa repository:

http://cgit.freedesktop.org/mesa/mesa/tree/src/mapi/entry_x86_tls.h#n46

Now the question is: is this "legal" in any sense?  E.g. defining your own
inline asm label, which has the same name as a static char array?  Mesa seems
to have this construction since some time, so gcc apparently has no trouble
with it.

In any case, I think clang should not assert on this.  Either reject the
construction (but break Mesa), or deal with it. :)

-- 
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/20140518/1658a16c/attachment.html>