[Verse-dev] pyverse r5

[Emil Brink]

On Fri, 18 Feb 2005 03:17:41 +0100 (CET)
"Eskil Steenberg" <eskil at obsession.se> wrote:

> Hi
> 
> > But, it has had nearly no testing.

Still, thanks a lot for making the effort of porting it over.

> >  The current r5 code isn't working very well for me. First, about
> > half of the commands got lost. After i
> > did a clean checkout, it stopped working entirely: the server only
> > receives garbage commands. Even the simple list-nodes.c in examples/
> > in cvs isn't working. I'll see if I can get it to work later on.
> 
> Thats very strange. I delibretly didnt get in to the outgoing history
> performance issiues to avoid this kind of problems. It sounds like the
> encrytion code. I will look in to it too.

Very strange indeed. I've seen this happen a few times, but that was
when I had the login and data key sizes not matching. They match in
v_encryption.h now, and I can run various simple clients without any
problems.

I did have issues with Purple yesterday afternoon, which were rather...
interesting, but I don't think that's anything that can happen in a
simple thing like list-nodes.c. The problem was that Purple tries to
connect to the host multiple times, doing repeated attempts until it
gets a connection. But trying now involves generating a key, which
pushed the time an attempt took to very close to my 5-second period.

So, it did a new attempt in the middle of the old one, which I think
messed up the state engine in the connection system, so that it never
actually managed to get a connection. The "fix" I did was to simply
increase the delay a little. Does _not_ feel optimal, but I don't
have the energy to look into stuff like that now, the exact working
of the connection-establishment code are still somewhat of a myst-
ery to me.

I'll try running list-nodes.c now, and see if it works out for me.

Regards,

/Emil