gcc-4.3 optimizer bugs?

William Cohen wcohen at redhat.com
Wed Feb 27 15:24:01 UTC 2008


Neal Becker wrote:
> Horst H. von Brand wrote:
> 
>> Neal Becker <ndbecker2 at gmail.com> wrote:
>>> We are trying to track down a numerical problem with unuran that occurs
>>> on
>>> gcc-4.3.  I know that gcc now has a very extensive test suite, and bugs
>>> are
>>> not common.  I'm just wondering, are others finding similar problems with
>>> gcc-4.3?
>>> We are working on preparing a bug report.  The symptoms are:
>>> 1) Problem disappears if -O0
>>> 2) Problem disappears if a printf is added
>> That probably means there is a problem with the memory layout (something
>> stomping on your variables?).
> 
> Even if valgrind is silent?
> 
> (BTW, this is not my code)
> 

Valgrind is instrumenting the code at a very low level. As a result, valgrind 
doesn't have knowledge where one structure or array ends and another starts. 
Valgrind will catch access to uninitialized location in memory but is unlikely 
to catch array out of bound references.

Would it be possible to compile the code with the mudflap option in gcc. Mudflap 
is in the compiler and has a better understanding of the data structures:

http://gcc.gnu.org/wiki/Mudflap_Pointer_Debugging
http://gcc.gnu.org/onlinedocs/gcc-4.2.3/gcc/Optimize-Options.html#index-fmudflap-502


-Will




More information about the fedora-devel-list mailing list