[LLVMbugs] [Bug 21176] New: option for light-weight poisoning of uninitialized variables

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=21176

            Bug ID: 21176
           Summary: option for light-weight poisoning of uninitialized
                    variables
           Product: clang
           Version: unspecified
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Static Analyzer
          Assignee: kremenek at apple.com
          Reporter: jhi at iki.fi
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

Tru64 cc used to have a flag called -trapuv which initialized uninitialized
values with 0xfff58005 (repeated, so 64-bit value was 0xfff58005fff58005).

This was a handy value in that it was an illegal pointer value (for Alpha
architecture), and as a floating point value (float or double) it was a signal
NaN.  So trying to deference the pointer value or use the fp value lead into a
quick failure (SEGV or FPE).

One cannot win always, of course, in that for integer values the value didn't
(and couldn't) cause a trap, but at least it was not zero, so there was some
change of people noticing that things aren't right.

Some poisoning like this would be a nice light-weight 'sanitizer'.  The exact
values for the 'illegal address' and 'signaling nan' would depend on the
architecture.

-- 
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/20141006/bcb7669a/attachment.html>