Re: [GNUnet-developers] Reproduceable failure

[Igor Wronsky]

On Mon, 31 Mar 2003, Tracy R Reed wrote:

> My gnunetd dies quite regularly. It won't run for more than an hour or
> two.
> Mar 31 04:52:42 FATAL: Bad file descriptor in select. Please report if you
> know how to reproduce this. (see Mantis #381)
> Mar 31 04:52:42 __BREAK__

/Time for truth (rant-time!)/

Acknowledged and experienced. In addition my gnunetd dies even
more regularly by "Killed", because the task ate too much memory
for the kernel's liking. Consequently, I'm not running gnunetd
at the moment.

Can't help you. Can't help even myself. If the problems do
not go away either by Debian fixing some of their packages
(haven't been even able to trace which) or other gnunet
developers accidentally fixing it back to behaving nice,
there will be one (useless/minor) developer less. I've
been tolerating this ultimate mem hogging for over a month
now and my patience with the involved parties, including
myself, is growing thin. I do not like to be constantly
reminded how helpless I am, and I'd rather drop something
than be constantly tormented by it - that was my original
reason for getting involved. I tried to get rid of something
that was just trouble, trouble, trouble, in exchange for
something that seemed to be just around the corner and
the intellectual victor of the match. I wonder now, as Freenet
doesn't constantly explode on my face these days, whether
the places have changed, and still if instead of a simple
'corner' we are talking about years of remaining development
for all candidates. With monotonically increasing usability,
'years' would be preposterously long prediction, but the
usability has just gone up and down, up and down, I can't
remember for how long. I can't even remember when gnunet or
freenet had such a version of which could be said "theres no
major bugs in this, no reason to upgrade whatsoever, unless
you want new features". Instead, as old bugs are got rid of,
new ones are introduced, or the new ways at efficiency or
whatever at the same time invent new bugs and make the
whole yet again unusable. And the cycle continues. Yet
most of the problems encountered seem to be just engineering
ones, those of ordinary software development; who eats the
memory, why select() craps, why datastore bangs, etc - those
have nothing to do with problems on anonymity or p2p.

Now how many bugs I could have fixed instead of ranting? None.
Ranting is like therapy, low-cost. Spending your time running
after bugs and incomprehensibilities and subsequently taking a
holiday at some mental ward, thats costly.


/Lets get back to pretending/

(silence)


Igor