[llvm-commits] PATCH: [Sanitizer] extend symbolizer code

Alexey Samsonov samsonov at google.com
Tue Jun 19 02:18:30 PDT 2012


On Tue, Jun 19, 2012 at 12:51 PM, Michael Spencer <bigcheesegs at gmail.com>wrote:

> On Mon, Jun 18, 2012 at 8:18 AM, Alexey Samsonov <samsonov at google.com>
> wrote:
> > This patch extends symbolizer code. In order to symbolize a module, a
> > symbolizer
> > maps it into memory and obtains pointers to necessary debug info sections
> > (this
> > part is implemented on Linux only).
> >
> > Later these pointers can be passed to constructor of DWARF
> context-in-memory
> > from LLVM DebugInfo lib.
> >
> > Codereview: http://codereview.appspot.com/6303097/
> >
> > --
> > Alexey Samsonov, MSK
>
> Why isn't this using lib/Object to read the object file?
> llvm-dwarfdump uses this to handle all three formats with a tiny bit
> of code.


Well, there is a trouble - currently I'm not aware of a way one can depend
on
LLVM code (i.e. include LLVM headers) in compiler-rt projects (and it can't
be solved
that easy - there is license issue, for example).

We'd need to find a way in any case if we're going to use code from
libLLVMDebugInfo
(and, hence, from libLLVMSupport) but I'd really like to depend on LLVM as
little as possible.
Maybe we would be able to build DebugInfo lib as .so and link against it.

Also, we can't run arbitrary code from ASan/TSan runtime (sometimes we can
even be hurt by
libc function calls or memory allocations).

-- 
Alexey Samsonov, MSK
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20120619/05337f7f/attachment.html>


More information about the llvm-commits mailing list