Let Matt Arsenault take a look at this. GCC is probably trying to generate
code to call __memcpy_chk with LLVM's objectsize intrinsic.
On Mon, Feb 10, 2014 at 1:23 PM, Scott Pakin <pakin at lanl.gov> wrote:
> It took a while to track down, but it looks like revision 192117
> (dated 2013-10-07) is the source of a DragonEgg segfault I've been
> fighting with for quite some time. For details and a reproducer (less
> than a page of C++ code) please see my Bugzilla report:
>
> Bug 18358 - "internal compiler error: Segmentation fault" on
C++ code
> http://llvm.org/bugs/show_bug.cgi?id=18358
>
> The real code that this affects is my Byfl tool,
> https://github.com/losalamos/Byfl, which compiles with LLVM revision
> 192116 and earlier but not with revision 192117 and later.
>
> I must admit I don't understand exactly what r192117 is doing, but
> I've attached a patch that simply and stupidly reverses the changes
> performed by r192117. Applying this patch to trunk (r201101) gets
> Byfl and my reproducer compiling again.
>
> Any chance this patch -- or, even better, a "real" fix for
whatever
> problem r192117 is introducing -- can make it into LLVM? (I've never
> submitted a patch before and don't know what the formal procedure is.)
>
> Thanks,
> -- Scott
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20140210/1e2fb193/attachment.html>