> > >All starts fail until i delete the swap.state.new manually and > > > restart squid. > > > > The IPC code in the Windows port is not perfect. Probably some of > > the child process, may be pinger, don't terminate correctly, > > leaving some files locked. > > The child processes should not even have a clue there is a swap.state > or even less a swap.state.new file. These files are internal to the > cache_dir store. > > More likely there is yet anoter "rename while open" condition on dirty > cache rebuilds.
This happens while a logrotate is happening, so the cache should not be dirty. On the second and third attempt it is dirty, of course, because the previous starts failed. This is the reason for the swap.state.new file. Hermann
