On Thu, 2004-01-22 at 10:05, Henrik Nordstrom wrote: > Hello fellow coders. > > I am in desperate need of some advice in how to debug a intermettent stack > smashing problem. > > There is no problem detecting when this happens as Squid segfaults almost > immediately (on return from the function which smashed the stack), just > not where. As the stack is quite smashed a function return ends up at > address 0 and there is no stack frames to follow.. > > As the problem does not occur very frequently running with full debug > enabled is not really feasible. > > I have tried using stackshield, but that tool is simply not yet mature or > maintained and the current version has very many bugs, mostly due to very > weak parser.. (needs to be rewritten). Attempted to build Squid with > stackshield enabled (after fixing some obvious bugs in stackshield) but > the resulting binary is not even a fully valid binary.. > > Does any of you know about other tools which can be used to trap where the > stack gets smashed?
Not that I've used :[. There was a linux distro with stack-protection in gcc IIRC. I asked a gcc hacker I know and was pointed at 'mudflap' as 'promising'. Possibly valgrind or another profiler could give you some detail too. Hope that helps. Rob -- GPG key available at: <http://www.robertcollins.net/keys.txt>.
signature.asc
Description: This is a digitally signed message part
