[webkit-dev] Question about Heap::markConservatively.

[Holger Freyther]

On Monday 21 December 2009 07:31:12 pattin.shieh wrote:

> b. Cound somebody tell me the crash reason about?  I think maybe program
> access non-align(4bytes) memory address?


Hi,

this is this interesting. Normally we are interested in a reduction. This 
means the smallest possible test case that is showing the error? Could you 
come up with that?

Regarding your assumption on unaligned memory access. The Linux kernel has at 
least three different modes to handle the alignment trap. Which one do you have 
set? You could set the fixup+warn mode and then see in dmesg which PC was doing 
this operation and use addr2line to find the place...

For Linux please see Documentation/arm/mem_alignment in the linux tree.

z.