http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47311
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #20 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-01-17
20:02:11 UTC ---
I think what valgrind complains is that it is reading past end of malloced
buffer, it doesn't figure out that eventhough some bits are set based on those
bytes past end of malloced buffer, it doesn't make a difference on the program
behavior.
I think the easiest workaround would be if we could make the allocations of
buffers on which the SSE4.2/SSE2 fast lexing is called are 16 bytes larger than
actually needed, either always, or just if (RUNNING_ON_VALGRIND) (after
#include <valgrind.h>).
