Please describe the problem:
The changes in esound 0.2.37 introduce a nasty bug in which esd will start to spin rapidly in a read/select/accept loop taking up 100% of the CPU until it is killed.  The problem stems from the fact that EOF is no longer properly detected on the player socket.  The attached patch corrects the problem simply by making sure we return on EOF instead of breaking out of the inner do/while loop.

Steps to reproduce:
1. Start esd
2. Run a test pattern using the Sound capplet
3. Stop the test pattern


Actual results:
esd will start to clime in CPU usage very rapidly after the test pattern completes.

Expected results:
esd should go to sleep

Does this happen every time?
yes

Other information: