OpenBSD support thread

Website · Post by **Z-Man** » Sat Mar 25, 2006 2:02 pm

The last one who broke it was you

Removing half of the $DESTDIRs, tsk, tsk.

Luke-Jr · Post by **Luke-Jr** » Sun Mar 26, 2006 12:07 am

Nope, from the looks of it, sysinstall has been broken since it was created. And my removals of DESTDIR are only in locations where it shouldn't be. DESTDIR should be used for nothing except where the files are copied and certainly never hard-coded into scripts.

Website · Post by **Z-Man** » Sun Mar 26, 2006 1:58 am

Darn, you're right about DESTDIR. Your change was standard compliant, my change was not. BUT:
Your change broke "make installcheck", which is our most important test. Actually, it's our only test

We need it. If I see it break without getting a note somewhere "I did this change here, it's absolutely required, but now make installcheck doesn't work anymore and needs to be adapted", I just make the test work again any way I see fit.
Cause in my book, the "does not work" broken is worse than the "not standard compliant" broken.

Luke-Jr · Post by **Luke-Jr** » Sun Mar 26, 2006 2:38 am

z-man wrote:Darn, you're right about DESTDIR. Your change was standard compliant, my change was not. BUT:
Your change broke "make installcheck", which is our most important test. Actually, it's our only test We need it. If I see it break without getting a note somewhere "I did this change here, it's absolutely required, but now make installcheck doesn't work anymore and needs to be adapted", I just make the test work again any way I see fit.
Cause in my book, the "does not work" broken is worse than the "not standard compliant" broken.

In my book, the "does not work in real-world scenario" broken is worse than the "does not work in a test" broken. I only found the bugs because they broke the ebuild. I'm pretty sure this won't be the second time I ask people not to revert a fix especially without contacting the fixer... if the fix breaks something else, then fix that problem, don't re-break it.

Website · Post by **Z-Man** » Sun Mar 26, 2006 2:50 pm

Responded here:
http://forums.armagetronad.net/viewtopi ... 5104#45104

Website · Post by **Z-Man** » Mon Mar 27, 2006 6:18 pm

Ok, I merged the peculiar OpenBSD init script method. It's not yet commited, but will be in a couple of hours.
Is there a canonical way to shut down daemons? The startup problems with renice you were experiencing are caused by stale PID files from the last boot still being around. No problem, I let the start script clean them up if they're invalid.
Uninstalling the init script also works

belenus · Post by **belenus** » Mon Mar 27, 2006 7:21 pm

z-man wrote:Ok, I merged the peculiar OpenBSD init script method. It's not yet commited, but will be in a couple of hours.

Sounds nice...

z-man wrote:Is there a canonical way to shut down daemons? The startup problems with renice you were experiencing are caused by stale PID files from the last boot still being around. No problem, I let the start script clean them up if they're invalid.
Uninstalling the init script also works

The problems with renice were only that -p is missing in front of the PID output... it should be:

Code: Select all

renice value -p PID

but unless you changed it the way I did in my scripts its just

Code: Select all

renice value PID

The "-p" is very important...

About shutdown, /etc/rc.shutdown is called at shutdown, could make an entry there to stop AA... never got to do that actually

Btw, it would be nice to include a little script at /usr/local/bin or just a link like my aactl to point to rcd_server so AA can be controlled from everywhere with just "aactl start|stop|status|whatever", just like "apachectl"...

If you need help with anything, just say so and I will do my best... =)

Website · Post by **Z-Man** » Mon Mar 27, 2006 9:40 pm

Then I was getting different errors than you

The -p has already been added in the last round.

I'll add the aactl link.

The shutdown question rather was: Do OpenBSD deamons usually get shut down with a script or are they just collectively killed? Our script wouldn't to much else than that, it'll just clean up the PID files afterwards

And thanks, you already were a great help.

belenus · Post by **belenus** » Mon Mar 27, 2006 11:28 pm

z-man wrote:are they just collectively killed?

Actually, yes... with some exceptions like apcupsd...

z-man wrote:And thanks, you already were a great help.

Hey, how about adding this to rcd_startstop?

Code: Select all

    log)
                tail -f $LOGFILE
        ;;

Automaticly displays the last 23 lines (common shell height-1) of the log and refreshes every 2 seconds...

Another thing, could you please fix the error from Step 3 of my HOWTO? =)

Edit: HOWTO changed slightly...

belenus · Post by **belenus** » Thu Jan 18, 2007 4:10 pm

Little update...

Version 0.2.8.2 works out of the box.

Version 0.3.0 needs some changes mentioned in the following thread:
http://forums.armagetronad.net/viewtopic.php?t=11231

Post by **Lucifer** » Thu Jan 18, 2007 4:50 pm

Hey belenus, would you be willing to run an openbsd buildslave for us? When my fingers have thawed out, I'll tell you more about it.

Post by **Lucifer** » Thu Jan 18, 2007 5:45 pm

Ok, fingers thawed. Basically you just need python, twisted, and buildbot. Twisted'd just a dependency for buildbot. Then I give you a user and password, and you checkout a certain svn module to get a client. Then you figure out how to make it so your build client starts and stops when you want it. I recently learned that the buildmaster queues jobs when a buildslave is offline, so you can pick when you want it to run and use a cronjob or something to trigger it.

That's it. You install buildbot, do the checkout, modify the files (directions on the wiki, actually), and the rest is automatic.

belenus · Post by **belenus** » Thu Jan 18, 2007 7:49 pm

Lucifer wrote:Hey belenus, would you be willing to run an openbsd buildslave for us? When my fingers have thawed out, I'll tell you more about it.

I would do that, BUT I would first need to figure out how to create an OpenBSD port from armagetronad-dedicated (never done that before, ever) and the package would be genereated from the port.

Ok, I will try to learn how to make a port... shouldn't be that hard.

If the port works and is good it MIGHT actually make its way into the official OpenBSD ports/packages.

Post by **Lucifer** » Thu Jan 18, 2007 7:54 pm

Well, the reason for the buildslave is so we can see the build problems for ourselves and work interactively with the builder to fix them. I'm basically offering for us to adopt your problem for you, if you can give us a buildslave.

At least, to the extent that we can get it building, with whatever we can do from our end. And you offer to be a tester, of course. The only requirements you have to satisfy are a subversion client (forgot to mention that one), and whatever the README says you need. For the trunk you'll need more, but basically just "install the distro-provided packages for this stuff" as much as possible.

belenus · Post by **belenus** » Fri Jan 19, 2007 11:41 am

Well, let me say it with simple words...

I only understand about 25% you say about that buildbot thingy.

Right now I would guess it would build an installable package (a snapshot) every day or so and thats about it, maybe you could explain further.

Anyway, I managed to create a port of the dedicated server for OpenBSD. package generates and installs on 4.0-current without issues. Server runs and you can play on it. What is left to do is adding some LIBS as dependencies and some other modifications because some stuff a make install usually does needs to do otherwise here, ie. create a user etc.