probbable generator hang bug in rsync

Steve Sether steve at vellmont.com
Thu Sep 15 09:29:59 GMT 2005


> Wow, that's unexpected.  I had thought that the connection from the
> stunnel process to the rsync process would have been a normal (albeit
> local) socket connection.  Perhaps it is not?  Or perhaps there is a bug
> that got triggered by this setup.  The way rsync works, it has two
> separate processes accessing one socket (one reading, one writing), and
> both need to be notified of the EOF.  It sounds like the receiver got an
> EOF when reading (and terminated) but that the generator did not get
> told about the EOF when waiting around to write.
> 
> ..wayne..

The connection between stunnel and rsync is just a normal socket connection.  
I've also ran an stunnel server process on a different machine that then 
connects to rsync on the server machine in an attempt to isolate any strangeness from 
local connections.  I get the same results of a hung generator process.

for clarity, the above test looks like this:

[rsync client]<-localhost->[stunnel process on client]<--local lan-->[stunnel process on server 1]<-internet->[rsync process on server 2]


Any suggestions on how to find out more of what's going on?


More information about the rsync mailing list