Dear all,

actually I am quite new with SLC and don't know if it is the proper way. 
Since my job is to encourage people doing profiling of their algorithms, 
it would be nice to have a more "stable" valgrind.

Some errors in valgrind 3.6 occurs, which I did not encounter in my 
self-compiled valgrind 3.7.

Exemplary I am profiling Moore, the HLT software from LHCb with Gaudi. 
Memcheck had a segmentation violation and now it shows another problem 
with the 32 bit version:

**5942** new/new[] failed and should throw an exception, but Valgrind
**5942**    cannot throw exceptions and so is aborting instead.  Sorry.
==5942==    at 0x49AACC8: VALGRIND_PRINTF_BACKTRACE (valgrind.h:5028)
==5942==    by 0x49AB587: operator new(unsigned int) 
(vg_replace_malloc.c:255)
... and so on ...

However, since all this problems did not yet occur in the valgrind 3.7, 
it would be nice if the distribution could include the new version, to 
make it easier for users to profile more complex software frameworks.

Thanks,
Stefan