[llvm-commits] PATCH: Stub for ASan/TSan symbolizer

[Alexey Samsonov]

On Thu, May 31, 2012 at 5:38 PM, Dmitry Vyukov <dvyukov at google.com> wrote:

> On Thu, May 31, 2012 at 5:32 PM, Alexey Samsonov <samsonov at google.com>wrote:
>
>>  What's the goal of this stub? Why is it needed, and how will it fit
>>>>> into the bigger picture?
>>>>
>>>>
>>>> Dynamic tools (AddressSanitizer and ThreadSanitizer) need to symbolize
>>>> code locations to produce fine stack traces.
>>>>
>>>
>>> I'm not sure I was sufficiently precise here: the key word in my query
>>> was "stub".
>>>
>>> I understand the need for symbolization. I was trying to get the review
>>> thread (and hopefully the comments in the source files) to reflect why the
>>> particular stub interface proposed in the location it was proposed was
>>> needed, and how it might interact with the symbolization library that is to
>>> grow inside of LLVM's core libraries.
>>>
>>
>> OIC. Location is /compiler-rt/lib/sanitizer_common as symbolizer would be
>> used by runtime libraries of both sanitizer tools, and would likely use
>> much of low-level code they share. It would also call LLVM debug info
>> library. In fact, this might be a concern, as that library should be
>> lightweight enough to be called from ASan/TSan runtime (and, afaiu from
>> dvukov@ we might have to fix it a bit).
>>
>>
> We don't use libstdc++, STL, libc, global ctors, big stack frames, etc.
>

These are serious constraints. Did you look if we can actually use anything
from LLVMSupport, for example?

-- 
Alexey Samsonov, MSK
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20120531/f7e7dff2/attachment.html>