The cost of stack traces

[David Daney]

Anthony Green wrote:
> On Wed, 2006-05-10 at 18:01 -0400, Andrew Haley wrote:
> 
>>Yes.  All we have to do is keep a table that maps IP address -> filename.
> 
> 
> To be more explicit, one idea would be to walk the stack backwards until
> we find something "familiar" (using some heuristic for determining what
> is familiar or not), and then simply using memcmp on the rest of the
> stack, to see if it's what we think it is.
> 

This "familiar" stuff would necessarily contain an IP->source_line 
mapping.  So APH's idea to just maintain an IP->source_line map might be 
simpler to implement with no additional memory required.

David Daney,