When running nntpcache chroot'd, because the configuration files are outside the chroot fail, a warm reload of the server will _always_ fail. At first it seems the solution would be to move the configuration files within the chroot jail, which partially defaults the purpose of the jail (but not completely). This isn't possible at present without modifying the code, and may not be the best way to address this. Perhaps, if desired, reader processes could chroot (assuming they were still uid==0 at this point, which they're not) into the cache directory... then give up perms? Hmm... this one is tricky the more I think about this. Having the config. files outside of the chroot jail is a good idea, and having the process give up perms asap is also good. A chroot jail for uid==0 is a waste of time. Maybe keep the fd open for the config. files, assuming the files are modified in place and retain the same inode, not renamed, this would work. Suggestions? -Chris
