timeout during hash_search

[Arne Jansen]

Hi,

I'm experiencing timeouts during synchronization of large files.
The scenario is as follows: Locally I have a large file (e.g. 20GB),
which is already present at the server. When I try to resync it with:

rsync --inplace --ignore-times -vvvv --compress bigfile rsync://xxx@...

The '--ignore-times' is only to force a comparison. The real world scenario
would be the sync of a truecrypt container.

This transfer aborts with a timeout. The reason is that during the hash_search
phase no packets are sent to the peer. If I omit the --compress option,
tokens are sent regularly, preventing the timeout, but send_deflated_token()
seems to collect an hours worth of responses before sending them out. Flushing
them on a regular basis to prevent a timeout would be good.

-Arne