|
From: Alexander P. <gl...@go...> - 2010-06-11 08:18:57
|
On Thu, Jun 10, 2010 at 9:22 PM, JoSH Lehan <kr...@gm...> wrote: > Thanks, I will do that. Interestingly, the instruction is the very > *first* instruction encountered by the executable as it begins to run, > as the address of the offending instruction is exactly the same > address that execution begins at. This leads me to suspect a setup or > configuration problem, with my particular platform. That could be an instruction inside ld.so or the kernel itself. For example, some time ago Valgrind didn't recognize userspace memory barriers used by the kernel built with CONFIG_SMP enabled. This was almost the first instruction and appeared in every program. That's why fixing a particular bug may help with all programs running on your system. > > Is there a FAQ for bringing Valgrind up on ARM? I looked around, and > found multiple conflicting sources. I'm not currently sure of the > state of ARM support. Any tips as to how to set up Valgrind on ARM > would be appreciated. > I don't think so, but I suppose it's rather straight, if you're building V natively on the ARM board: you just check out the code and run: ./autogen.sh ./configure --prefix=<path to install Valgrind> make && make install You can also try the self-contained Memcheck binary located at data-race-test.googlecode.com: wget http://data-race-test.googlecode.com/svn/trunk/tsan_binary/Linux-ARM/memcheck chmod +x memcheck ./memcheck your_test Please note that this is an unofficial Memcheck build and may contain bugs already fixed in the trunk. Alex |