[core-dev] possible bug in downloader code

I noticed today using LimeWire (current cvs) that a download sometimes 
remains in a 'connecting' state for a long time (at least several 
minutes), only switching to 'waiting for busy' for no more than a few 
seconds every now and then. I didn't have a chance to take a closer look 
at what was happening yet but I can tell you this much from the extended 
tooltips:
The problem affected two downloaders, one with 7 busy sources and 14 
invalid sources and one with 3 busy sources and 0 invalid sources. When 
the first downloader entered 'connecting' state, it would always show 20 
possible sources. This figure would slowly drop to the number of busy 
sources (7), then the downloader would enter 'waiting for busy' state 
for a moment and switch back to 'connecting' state with 20 possible sources.

The problem might in part be caused by our university's router which 
does not allow outgoing TCP connections to ports 6346 & 6347 (I hope 
future LimeWire versions may encourage the user to choose another random 
default port)  so it is entirely possible, that I receive HeadPong 
messages via UDP from hosts that I cannot connect to via TCP, a 
situation which may not have been anticipated in the downloader code.

mfg
gregorio
_______________________________________________
core-dev mailing list
[email protected]
http://www.limewire.org/mailman/listinfo/core-dev