[LLVMbugs] [Bug 8015] False use of undefined value warning due to not handling "array[symbolic value]" correctly

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=8015

Ted Kremenek <kremenek at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #5 from Ted Kremenek <kremenek at apple.com> 2010-09-01 18:29:20 CDT ---
(In reply to comment #2)
> Created an attachment (id=5423)
 --> (http://llvm.org/bugs/attachment.cgi?id=5423) [details]
> Use UnknownVal for stack-based ElementRegions with non-constant indexes
> 

I've gone ahead and applied this, since I think this is a good working base to
suppress the false positive.  We then should improve on from there.

Patch applied here:

http://llvm.org/viewvc/llvm-project?view=rev&revision=112766

In the patch I included a test case showing how this approach *doesn't* do the
right thing:

+// FIXME: This is a false positive due to not reasoning about symbolic
+// array indices correctly.  Discussion in PR 8015.
+void pr8015_D_FIXME() {
+  int number = pr8015_A();
+  const char *numbers[] = { "zero" };
+  if (number == 0) {
+    if (numbers[number] == numbers[0])
+      return;
+    int *p = 0;
+    *p = 0xDEADBEEF; // expected-warning{{Dereference of null pointer}}
+  }
+}
+

Here we actually flag a bogus null dereference warning.  I'm going to file a
follow-up bug to track this one, and mark this as fixed since the false
positive is gone.

-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.