Svadev mailing list

[Will Dietz <willdtz AT gmail.com>]

I just committed some minor changes to SC that should fix this issue.

The underlying issue was the same as
http://lists.cs.uiuc.edu/pipermail/llvmdev/2011-August/042684.html and
the fix was what was described there.

Unexpected trick was having to set the alignment, since otherwise the
memset calls would segfault (separate commit to highlight this
aspect).

Hope this helps!

~Will

On Mon, Jun 11, 2012 at 11:14 PM, John Criswell 
<criswell AT illinois.edu>
 wrote:
> On 6/11/12 10:05 PM, Matthew Wala wrote:
>> [snip]
>>>> John, I know you wanted me to try to compile Apache, but this issue is
>>>> getting in the way. Can anyone else reproduce this?
>>>
>>> Matt, can you do the following?
>>>
>>> 1) File a bug with the reduced test case as an attachment.
>>>
>> I haven't done this yet. Should the bug be filed against SAFECode or clang?
>
> It should be filed against SAFECode as it doesn't appear when the
> -fmemsafety option is not used (therefore, it may not trigger in the
> release_30 branch of LLVM/Clang).
>
> You can think of it as a SAFECode bug that occurs because SAFECode is
> using LLVM 3.0 instead of LLVM 3.1.  The fix is to upgrade SAFECode to
> use LLVM 3.1 (assuming that LLVM 3.1 fixes the code generator bug).
>
> -- John T.
>
> _______________________________________________
> svadev mailing list
> svadev AT cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/svadev