• Bill Baxter (3/3) Dec 03 2007 It seems like maybe it's building the full string describing the stack
• Jascha Wetzel (12/16) Dec 04 2007 hm, in naive tests the printing starts instantly and the speed is

Bill Baxter <dnewsgroup billbaxter.com> writes:

It seems like maybe it's building the full string describing the stack 
in memory and then printing it out.

--bb

Jascha Wetzel <firstname mainia.de> writes:

Bill Baxter wrote:
 It seems like maybe it's building the full string describing the stack 
 in memory and then printing it out.
 
 --bb

hm, in naive tests the printing starts instantly and the speed is 
limited by the console output.
what changed in 0.11 is that ddbg goes through the stack twice in order 
to find the "last known source location" for stack frames that don't 
have source line info. unfortunately, i don't see a way to do it in one 
pass, since the stack frames can only be unwound in one direction.

there still is potential for optimization, though, since for every stack 
frame i do a tree search for the source location of the return address. 
assuming that in a stack overflow situation there are only few locations 
repeated many times, hashing the lookup results could speed it up.

if you can send me a testcase, i'll do some profiling.