Write a memory around the instruction pointer from the crashing thread to the minidump on OS X

On Windows, MinidumpWriteDump includes a window of memory around
the instruction pointer from the crashing thread in the minidump,
which can be useful for debugging JIT crashes. This patch implements
the same behavior for the OS X minidump generator.

The patch includes some build shuffling to include a few processor
source files in the unittest build, as well as moving
src/common/linux/{memory.h,memory_unittest.cc} to src/common, with
some slight changes so memory.h is usable on OS X. I also hooked up
the memory_unittest tests that weren't being run anywhere.

[mochalatte]

LGTM

http://breakpad.appspot.com/200001/diff/1/8
File src/client/mac/handler/minidump_generator.cc (right):

http://breakpad.appspot.com/200001/diff/1/8#newcode581
Line 581: const size_t kIPMemorySize = 256;
could you comment about units or make it part of the var name?

http://breakpad.appspot.com/200001/diff/1/8#newcode615
Line 615: uintptr_t(ip - (kIPMemorySize / 2)));
i might be a little hazy on the details of mach_vm_region_Recurse but do you
want max or min here - shouldn't the lower bound be the {start of
region|ip-128}, depending on which is closest to the faulting address?


also, what does mach_vm_region_recurse return if the IP is not part of any
memory region? could it return the memory region that is valid and "above" the
IP? (again stressing that it's been awhile since I touched this stuff)

[Ted Mielczarek]

http://breakpad.appspot.com/200001/diff/1/8
File src/client/mac/handler/minidump_generator.cc (right):

http://breakpad.appspot.com/200001/diff/1/8#newcode615
Line 615: uintptr_t(ip - (kIPMemorySize / 2)));
On 2010/09/22 20:24:26, mochalatte wrote:
> i might be a little hazy on the details of mach_vm_region_Recurse but do you
> want max or min here - shouldn't the lower bound be the {start of
> region|ip-128}, depending on which is closest to the faulting address?

Oops, good catch. I added tests to check that we bound the memory appropriately.

> also, what does mach_vm_region_recurse return if the IP is not part of any
> memory region? could it return the memory region that is valid and "above" the
> IP? (again stressing that it's been awhile since I touched this stuff)

Also good, I added a test for this as well (made the instruction pointer NULL,
ensured that a memory region didn't get added to the dump).

Now of course I'm going to have to go back and write those tests for Linux to
make sure that code works too.