On Tue, 25 Apr 2000, Ivo Timmermans wrote:
> On Mon Apr 24 2000, Mike MacNeill <mikem@accesscomm.ca> wrote:
> > I downloaded the cabal version as well and I seem to have discovered a
> > problem.
> > Whenever the regenerate key goes off it manages to eat memory
somewhere.
How do you know that? What are the symptoms?
> efence didn't show anything when I tried to use it a few weeks back.
> Guus discovered two bugs this weekend, so maybe that fixes your
> problem.
> Guus: how is that fix coming along?
Well... first all strange segfaults have stopped (checks for NULL pointers
everywhere now), and tinc works for networks with more than 3 hosts
correctly. I don't think it will fix the key regeneration bug though.
Bugs left:
- race condition when multiple clients want to connect at the same time
Client connects are not atomic, a client quitting halfway exchanging
basic_info/keys will induce a SEGV in the server. Locks?
- still a maximum for the number of clients
(We need to disable cheat mode and clean up the protocol handlers.)
- don't send structs directly (you never know when gcc will rearrange
them). Best would to sprintf() the structs and then write the string
to the socket. That also immediately fixes the host byte order
problem.
Met vriendelijke groet,
Guus Sliepen.
---
TINC development list, tinc-devel@nl.linux.org
Archive: http://mail.nl.linux.org/tinc-devel/