On Wed, 5 Nov 2008, Andreas Wurf wrote:

> To make it short: I could reproduce my problem (one channel stuck in passive
> mode) with the attached source on windows. But only on approx. every 20th
> test run. And I had to increase the number of files and their size to
> reproduce the error. Maybe we should implement a true dynamic way of faking
> a large number of files with a random size in a well-known range, to make
> this a real stress test.

Using this version I too could reproduce some kind of hang (roughly on one
attempt out of five or so) but it doesn't seem to be identical to your case. I
only get it after 58 transfers and after a 221 (transfer complete) is
received.

I'll need to experiment some more with a remote server (I only tried with one
on localhost so far) and see if I can get it with less number of files
transferred.

It is a bit to track down when it is this unreliable to trigger.

Oh and it built perfectly on Linux even with your changes.

-- 
  / daniel.haxx.se