Re: [Help-gnunet] Major problems on RH7.2

[I. Wronsky]

On Mon, 22 Apr 2002, Christian Grothoff wrote:

> > (both use same version of openssl, btw)... I'll try to get the
> > other node upgraded to the same rh/gcc version and see what happens
> > then. If it starts working, I'll let you know that gcc2.96-98
> > is to be avoided.
> Could be. You may also want to try to play with compiler options. Sometimes 
> the right optimization level can do wonders. Or it may be OpenSSL, or a bug 
> in our code, really hard to tell. If you can demonstrate that the same code 
> works fine with compiler X but not with compiler Y, please tell me and I'll 
> document that problem.

No, it wasn't that, sorry. I've experimented a bit now, and 
its clear that the problem appears on both hosts. I didn't
notice that at first, because I didn't try to use exactly
same files, and the problem is related to file size: file 
of 28kb inserts and returns correctly, but files of 100kb 
don't come back, but give

Invalid packet for Inode-Node at position 0 (hash .,..

flood instead. My rough guess is that the tree stuff is 
somehow broken (I remember reading you have a tree struct)
and the problem appears after a certain filesize. I didn't 
try to track the exact boundary, though.

Of course this might be system dependent. Anyone else
seen anything similar? If not, any debugging hints?

And another issue. if I insert a file, I can't find it 
with gnunet-search until i've shut down gnunetd and 
restarted it ... this might be related?


btw, today is the first time that content has actually arrived 
at my node from outside. ;) Remote hosts are even gaining 
credit. Promising. Yesterday there was 300megs udp traffic but 
not a single file appeared to content/ dir, nor did 
anyone get any credit. 

> Eh, not quite. I just accidentially killed the script that generates the 
> hosts.tar.gz-tarball. There have been over 270 hosts online so far, and even 
> usually a couple at a time recently. Check out
> http://gecko.cs.purdue.edu/GNUnet/hosts/ to see the last time at which a host
> was seen online...

Yes, I found that after a while... a direct href-link to
that dir could help. ;)

> Well, if we can find out, sure. It's just pretty hard to be 100% certain that 
> it is the compiler and to know which versions... ;-)

True, true. I even myself jumped into false conclusions. 

> absolutely concur with you that there is a lot to be done (and there will 
> probably be a big push after the semester is over, most developers have a 
> life, too :-)

Whaat? Nooo... That can't be accepted. ;) 

> *right now*, we made the code available such that people can give us ideas 
> and/or help. 

I understand that. And in a way, thats just what I'm trying
to do, in my own insignificant capability. The other option
would be just to wait for the problems to disappear,
and that can require even more patience than bug
hunting.


 - I.