Direct Connect Pro / News: Recent posts

Just to tell you, I will not do anything more on dcpro. I have not given up my DC aspirations, but I have decided to rewrite it from scratch. Once I get a complete system, search for Dolda Connect instead.

I compensated for the CLONE_THREAD flaw in linuxthreads, so the issue with zombie filters seems to be solved now. The new release is available as dcprod 0.2-2.

I finally discovered why dcprod fails to wait for the filter processes. I suspected from the beginning that it was a flaw in the LinuxThreads implementation, but I thought it was that the signal handlers weren't correctly cloned. When I looked deeper, it turned out that LinuxThreads puts the cloned processes in different thread groups, even though Linux has had support for thread groups since the release of 2.4.
I was going to recommend everyone to upgrade their glibc, but then I found that not even the latest CVS version uses thread groups correctly.
I have begun writing a compensation for this issue in dcprod, so that the filter are both forked and waited for in the same thread, and hopefully I'll be done later tonight.

I've released a GUI in the dcpro (as opposed to dcprod) package. It runs with gtk+-1.2. Much is to be changed with this GUI. I will remake the make process, too. The current release is truly an alpha release, but it's far better than nothing.

I finally took my time to add list locking for interfaces and transfers. It seems to have paid off, though, becuase I'm not getting any of these strange, register changing crashes anymore. I've fixed some deadlocks, but some may be left.
The current bugs are more of the resource-devouring kind. For example, zombie filters aren't waited for, and there still seems to be a slight memory leak.
It seems stable enough to warrant a minor version number change, though.

I released 0.1-9 today, in which I fixed som emore bugs, most significantly of which that I made interface fd:s nonblocking, something that used to block threads arbitrarily when connected to laggy hubs.
I've also added virtual file support for the share, which I plan to use for the file list one day. Unfortunately, the crappy DC protocol imposes some limitations on this, but it might still be usable for some purposes.

I released 0.1-8. I think I've finally found out why the uploads tended to block. I never even considered that send() could differ form write() in that it blocks until _all_ bytes are written...

After a month of debugging, I believe that I have finally found the cause of the constant segfaults.
After using valgrind, it seemed they were due to that commands issued by a closing hub remained in the command queue, and were executed with an invalid interface reference when the hub interface was closed. I have not yet done any exhaustive testing, but it has made it through a situation that made it crash almost every time before, namely to be able to connect to a large number of hubs in sequence.
The problem was fixed simply by adding mutexes to the interfaces and deleting commands from the command queue when the interface they were associated with closes.
It took some pretty long time, though. I just hope it was worth it.

Just wanted to tell everyone that today I got kicked from my first hub for not having the file list available. The bastards!

I finally took the time to write manpages for dcprod. I have added them to release 0.1-6 along with some extra bugfixes. I finally saw what I had made that could just delete transfers in the middle of the file.

I have started working on a HTML/HTTP based GUI, since I think it's too boring to make an ordinary, windowing system based GUI.
A such would, however, still be extremely useful, providing someone made one. If you feel up to it, please contact me at fredrik@dolda2000.cjb.net
Windows, Gnome/GTK+ or Java would be preferred.

I made some internal improvements which resulted in a minor bugfix. The most important to this release is, though, autoconf/automake. My thanks stretch out to Kapelonis Kostis (kkapelon@freemail.gr) for making this possible.
There does seem to be a slight memory leak, but my dcprod has been running for some days without causing problem, so it doesn't seem to be urgent.

OK, dcprod has been running for many hours now without any signs of crashing or leaking memory.
It seems it can actually be used. At least I have done so.

I think the transfer system is fixed now.
The daemon seems stable enough to actually be used now.
I released it as 0.1-3

I just released version 0.1-2 after some hours of debugging. Even though I haven't run any long run tests on the results yet, it seems much more stable than before. At the very least, it was a large step in the right direction.

I've been working on this for some weeks, and I'm quite proud of the result, even though there seem to be some bugs left to get rid of.