this seems to be the 3rd attempt to place a thttpd-ebuild in portage; the latest is bug #35927 35927 seems a bit overloaded with attachments, patches, ... so I post my ebuild here. sorry for the duplicate bugs in advance. thomas Reproducible: Always Steps to Reproduce:
Created attachment 22468 [details] thttpd-2.25.ebuild
Created attachment 22469 [details] init-script
Created attachment 22470 [details] conf.d config
Created attachment 22471 [details] sample config-file
*** Bug 18266 has been marked as a duplicate of this bug. ***
Up and running on PPC.
*** Bug 35927 has been marked as a duplicate of this bug. ***
Sorry for the delay in looking at this. We're a bit short of volunteers to look at new packages. I've committed the ebuild into portage. It should show up on an rsync mirror near you in around an hour. Could you test this please, and let me know whether it works for you or not? I haven't had the chance to test this myself yet. Thanks, Stu
Forgot to change the bug's status (doh!)
Tested. Executive summary: looks good. Eyeballed scripts, MD5sum of source agrees with a copy I'd previously obtained directly from Acme, ran with default config and found to be working. CGI stuff in the distribution is safely disabled unless user really wants it. Works with my own (unique-situation) config too. Minor-quibble-1: the init.d script could itself pass "-i /var/run/thttpd.pid" to thttpd to ensure this never gets out of sync with the same pathname being given as a parameter to start-stop-daemon; you could then leave the pidfile entirely out of /etc/conf.d/thttpd and /etc/thttpd/thttpd.conf.sample. Minor-quibble-2: there's no logfile rotation happening, though Jef himself admits this is problematic with a chrooted thttpd. My preference is to log to a named pipe; http://www.leftmind.net/projects/logdated/ has the thing I put on the read end, but that would be an entire extra package and init script and all. Likely there should be some simpler solution enabled for the default case.
sorry for commenting late on this. not much to add to comment #7 -- it worked for me too. quibble-1: i put it in the con.d-file to make it changeable (e.g. for chroot- environments) but on 2nd thought it looks like a good idea to harde-code it in the initscript. quibble-2: is there a prefered way to do that on gentoo? some packages ship with app-admin/logrotate /etc/logrotate.d/-scripts -- would be better than nothing ... but sufficient?
It looks like the syslogs themselves use savelog(8); better that than a dependency on the logrotate package. However, writing to a separate logfile while chrooted means you have to pick between doing a full restart of thttpd to rotate logfiles, put the logfile inside the chroot tree, or write to a pipe or socket to an outside-the-chroot logging daemon. It might be best to just log to syslog in the basic config, and leave the complexity of rotating a native logfile to advanced users.
