I read the bug history for webapp-apache and userpriv and I'm not sure exactly what you did to fix it for userprive, Stuart. htdig still did not emerge successfully for me.
In any case, I made a ton of changes to webapp-apache. I don't know why the apache.conf greps were commented out. I made them better and I tested them with userpriv and usersandbox. It works very nicely. The big change is moving the fperms and fowner to webapp-pkg_install(). In order for userpriv to work, they have to be in pkg_install. This is going to mean some more changed ebuilds I bet.
I made a theoretical assumption about this eclass as well. The big question is do you want webapp-apache to force a depend on the apache ebuild? A package like htdig does not really need a dependency on apache. It does, however, need webapp-apache.eclass to emerge (for pathnames). We can certainly make webapp-apache work without having apache installed (as I did). We can have it supply default Gentoo values for file locations if apache isn't installed, otherwise read the values from config files (this is what I've done). Is this how you want it to act? I think it makes the most sense this way, in any case.
My fixes are really fixes, clean-up, and enhancements. I recommend running a diff.
I don't expect you to accept this entire modification to the webapp-apache.eclass as is, but please consider the ideas I present and my implementation.
I'm also commiting a bug fix for htdig that allows it to work with userpriv enabled that will depend on this bug (it depends on my modifications to webapp-apache).
Created attachment 22406 [details]
This is related to Bug 32636.
By the way, the word is "deprecated", not "depreciated" (from your comments in the current ebuild) :).
please submit diffs in future
*** This bug has been marked as a duplicate of 32636 ***
By marking this as a duplicate you closed the bug. Why? No one is going to look at it now and the eclass is definitely _not_ fixed.
I haven't seen your proposed changes anywhere else other than in this bug today. I didn't know they existed, and that's why I didn't consider them when I made the fixes I did to the webapp-apache eclass. Where have you posted them before now?
htdig-3.1.6-r4 has just emerged successfully for me - with userprivs in the FEATURES in /etc/make.conf. I cannot reproduce your problem.
Can you please post the following:
a) the captured output from running 'emerge htdig'
b) the $Header line from htdig-3.1.6-r4.ebuild
Thanks a lot,
Stuart, I didn't mean to alarm you. I just submitted my changes today. I was working on them in the last 3 days. I think some of them are good enhancements; no rush on reviewing them.
I am completely unable to reproduce my htdig problems, however, I did do an emerge rsync before reading this. Did you recently commit your webapp-apache.eclass updates? I completely debugged the problem when it occured. The fperms and fowners code in webapp-apache was running in the web-detect portion of the script, which gets called during src_unpack. This was causing a permissino denied error. I see now that you've created a webapp-apache-mkdirs function, which was not there when I assessed the problem originally, so I have to assume I just missed your updates! doh!
What do you think of my other changes?
It'd be a bit easier for me (because I'm at work and nowhere near a Linux box) if you could resubmit your webapp-apache changes as a diff against the latest webapp-apache.eclass file.
Created attachment 23111 [details]
Patch against webapp-apache.eclass v 1.19 2003/12/1 6 10:55:43
Sorry this took so long to submit
Okay thanks. If you've not heard back from me within a week, please drop me a friendly reminder ;-)
Friendly reminder ;-)
Heh. Thank you ;-)
I'm closing this bug because Stuart and I talked and decided it did not make sense to spend any time on this issue. Stuart's webapp-config utility and installation method will replace all of this confusion with an elegant solution.