Some bug fixes

I added some checks to cas to prevent segfaults (someone should port it to C++...), added an option to specify the location of the amuleweb binary for autostart and fixed the wxCas save dialog.

lfroen removed mutexts from amuleweb as we don't use threads anymore.

amuleweb-assert, php-standalone, compat

With developers either desperately lacking time and/or not feeling like working during the summer vacation only few changes occured during my absence.

aMuleweb gained support for creating basic backtraces on crash (just like aMule does, but using gdb is still necessary to properly identify problems). I think Vollstrecker, phoenix and GonoszTopi worked on this.

Lfroen changed the build process to allow for php standalone compilation.

GonoszTopi committed some patches to improve compatibility with different shells.

All quiet on the southern front

whilst I haven't updated lately (I blame flower vendors.), there have been some changes to aMule.

It was mainly Vollstrecker going on a patch spree and changing everything which had no svn lock on it.

He worked on Copyright messages, po file encoding, fixed make clean and make distclean (fixing a real bug) and updated the debian folder.

Myth contributed a patch for Feedback from shared files while morph had a look at our man pages.

Oh, and we made a release. But I think you know that already.

amuleGui, Windows Resource leaks, SX

Thanks to xor the amuleGUI desynchronization should be fixed now in 2.2.2 and SVN.

Also Stu Redman was working on fixing GDI resource leaks on the Windows build which should improve long term stability on Windows. (Hopefully this can be fixed on wx side, too.)

Muleview supports now v3 nodes.dat files, thanks to a patch by GonoszTopi.

At last, Kry committed a fix for incorrect source exchange packets.

Kad bugfixes and improvements.

GonoszTopi made more improvements to the aMule Kad implementation. (Will be included in 2.2.2)

• Implemented a special nodes.dat version, which is more suited for mass distribution without causing significant additional traffic for the included nodes. This is to allow eMule to add a nodes.dat to the installer, so that new user can connect to Kad without having to connect to servers or downloading a nodes.dat themself. We also need to be able to read and process this nodes.dat file.


• Implemented an alternate method to estimate the total user count in the Kad network, which is not based on our local routing table. It is experimental and only in trunk.


• Fixed a minor bug with kad obfuscation when contacting answering unbofuscated hello-requests.


• Made sure that only one challenge to verify a contact is sent at a time to the same IP.

Kad v8

GonoszTopi ported Kademlia version 8, which will be intregrated into our next release. I quote the Changelog:

• Kad now ignores multiple IDs pointing to one IP in routing request answers


• When calculating the total user count, Kad now uses (if available) the real number of firewalled clients for the calculation instead a fixed value. This will take more effect when 0.49+ nodes become common


• Kad no longer tries to add UDP firewalled (and therefore mostly unreachable) nodes into its routing table, reducing overhead and increasing reliability [Idea: Netfinity]


• Optimized the Kad packettracking code to only track packets which we check later, to save resources


• Fixed a bug which let Kad sometimes assume a Kad2 contact to be Kad1 after probing for both versions


• Did first small step in dropping Kad1 support by not probing unknown Kad contacts for Kad1 anymore to save overhead. Kad1 clients will nevertheless still be work for now.


• Several changes were made to Kad in order to defy routing attacks researched by University of Minnesota guys [Peng Wang, James Tyra, Eric Chan-Tin, Tyson Malchow, Denis Foo Kune, Nicholas Hopper, Yongdae Kim], in particular:
  
  • Kad contacts will only be able to update themself in others routing tables if they provide the proper key (supported by 0.49a+ nodes) in order to make it impossible to hijack them
  
  
  • Kad uses now a three-way-handshake (or for older version a similar check) for new contacts, making sure they do not use a spoofed IP
  
  
  • Unverified contacts are not used for routing tasks
  
  
  • When reading a nodes.dat without any verified contacts (possible an old file version), aMule will assume all contacts to be verified to avoid slowing down the connecting process
  
  
  
  
  

aMule development blog

This blog serves to inform users about important changes in the developmental version of aMule. This includes new features as well as changes to the application like new configuration options.

The blog won't be updated daily or immediately after every change, but expect to be informed of the most important changes in a timely manner.

I happily accept contributions like feature explanations, please contact me.

kind regards

Johannes Krampf <wuischke@amule.org>