problems with runlevel implementation

[Netmonger]

Hi! Im a new gentoo user, and would like to report what appears to be some problems with the implementation of the runlevel scripts in gentoo linux.

I was noticing that my auto NFS mounts were not coming up automatically. Typing 'mount -a' after the system was booted up worked ok, but it was annoying. Both 'netmount' and 'portmap' are set to run in my default runlevel.

I added some debug statements to /etc/init.d/netmount, and discovered that the test for the symbolic link /mnt/.init.d/started/portmap, which indicates that portmap has been started, was failing. I figured that portmap wasnt being 'given enough time' to start up..

After chatting with a dude on #gentoo, I tried his slight modification to the 'netmount' script. Instead of checking for /mnt/.init.d/started/portmap, we repeated the previous check to see if any auto NFS mounts were in /etc/fstab..

Well.. it hung my system up cold.. I was pretty confused - because portmap is in my default runlevel. And I knew it was starting up before..

I added some debug statement to netmount - a while loop which repeatedly checked for the symlink while doing a directory listing on /mnt/.init.d/started - and found that now, portmap WASNT starting!

So I added some debug statements to /etc/init.d/portmap, and thats when things _really_ seemed to get out of hand..

I sat and watched the 'depend' section of portmap be executed 3 or 4 times.. However the 'start' section of portmap was never being executed.

The dependancy check for 'net' was now failing from 'portmap' - although this was working fine a minute ago!?!

Taking out all my comments, brought my system instantly back to the state, when portmap was runing ok, and both portmap and netmount's symlinks were set in /mnt/.init.d/started - and my auto NFS mounts were again, UN-mounted..

Basically, my conclusion/fear was/is this: (And I havent looked through all the scripts yet to know EXACTLY how this works. so dont flame me too bad if Im wrong somewhere here.. )

It seems that the scripts are executed somewhat haphazardly - depending on the specific scripts you have setup for that runlevel, and depending on the timing that they take to execute, that you can cause some dep checks to fail and thus even some scripts to not execute correctly - i.e. my auto nfs mounts done get mounted at startup..

It even seemed my debug statement affected the timing here!

I _do_ admire the sort of 'automagical' approach to the sequence for your startup runlevel scripts in gentoo linux. Its cool to be able to throw another script in with 'use' and have it just work. Especially in contrast to the truly archaic system V approach - which is so standard everywhere..

However, from what Ive seen, it really looks as though a little more work needs to be done here to make them truly foolproof. Im still looking into it myself..

Has anyone else seen anything like this happen?

Also, what _is_ the preferred channel for submiting bugfix code to the gentoo developers?

[pjp]