Present Perfect

So it's 6.30 in the morning and I wake up and can't get no sleep. What better time of day to try my hand at packaging a working avifile version than this ? Now that I have a clean Red Hat 7.2 system, I should be able to make some sense out it all.

Avifile is a mess, currently. Snapshots of the 0.6.0 version get released every so often, and everyone is using different ones. I want to get vcr working so for now I'll use the tarball snapshot on that page. I wish the vcr author would at least adhere to the semi-official CVS snapshots. Oh well ...

Also, the snapshots always contain symlinks to all of the autotools stuff, so I first have to fix that. Then autogen some files here and there, because avifile looks like it's a few libraries slapped on top of each other.

All the while we try to keep a clean "fixed" tarball to build the rpm from, while trying to incorporate an extra tag somewhere in the naming scheme to indicate this is a modified version of the 0.6.0 version of the 11/11 release of avifile. I hope you see why this is a mess.

Anyway, after an hour of mucking about, rpm's seem to build. Let's see if tonight's automatic recording of Dark Angel works.

You might ask yourself, Why is this guy trying to build avifile packages when there are RPM's for RedHat 7.2 available on the site ?. Well, simply because the RedHat packages on the site are built against SDL 1.1 when Redhat includes SDL 1.2. So the packages there are worthless in that regard. And forcing the install is no good either : I'm a firm disbeliever of allowing that. If rpm complains about this, then it means the package isn't well-made or isn't suitable. Simple as that !

Anyway, if you want either the modified source or the RPM's, which install on a clean Red Hat 7.2 system, download them.

So it's 6.30 in the morning and I wake up and can't get no sleep. What better time of day to try my hand at packaging a working avifile version...

What a weekend. I was going to devote three days to work on the Dave/Dina project, but stuff got in the way. Among "stuff" was trying to fix someone's cable internet connection on Windows ME. The thing was : pings to all of the sites worked on IP address, but no name lookups, no connections possible (telnet, ftp, ...), and all programs claimed to have socket errors. Re-installed Millenium in three different ways and still nothing worked. I told him to seek professional help ;)

Anyway, I finally managed to get everything converted to Red Hat 7.2. I also worked on the install disk which now works pretty smoothly. I managed to get more than one kickstart config file on the disk, and now that I understand the disk image concept more I see the logic in it. Now it's just a matter of further tuning the package selection.

What a weekend. I was going to devote three days to work on the Dave/Dina project, but stuff got in the way. Among "stuff" was trying to fix someone's cable...

I suppose everyone sometimes gets that little voice in the back of the head telling you not to try something you're about to try. Either because it's stupid or because it makes no sense or whatever ...

Sometimes it pays to listen, as I just found out. The problem was simple : while writing a kickstart file, I needed to get one file from the floppy disk which was mounted as / to the new file system. Normally you'd do this in the %post section in the kickstart configuration file.

Now, %post executes in a chroot'ed environment equivalent to your newly installed system. So there's no way to get at the file you want to install. But you can run %post without the chroot'ed environment by using --nochroot.

Only problem is, you cannot go back to the chroot'ed environment (I thought the chroot command might do the trick, but that either starts a new shell interactively, which is bad of course, or allows you to execute the one command you put behind it).

Ok, so you think of a few ways around this :

• Actually try to mount the floppy disk in the %post script and copy the file from the disk. Works, but the user needs to keep the floppy in. He probably will, but it doesn't feel right.
• Write the new file line by line from the kickstart config file. Works, but totally style-less.
• Do %post --nochroot, and execute each of the other commands (which should run in a chroot environment) prepended with chroot /mnt/sysimage. Works, but equally style-less and cumbersome.
• Do %post --nochroot, put all of the chroot'ed commands in a separate file on the disk, and do something like chroot /mnt/sysimage /post.sh. By far the best solution, but still not satisfying.

So what do you do when none of the solutions is aesthetically pleasing, and a little voice in the back of your head says : Wouldn't it be nice if you could just have two %post sections, one labeled "%post" and another labeled "%post --nochroot" and just put the right commands in the right section ?

And you think to yourself, yeah, that would be nice, but no way would that work. I'm sure the Red Hat engineers didn't code to allow that, and otherwise I would have come across that on Usenet and in the docs.

But the voice says : Try it. You know you want to.

So you do.

And what do you know ? It works.

So to all of those out there who have the voice : make sure you listen.

I suppose everyone sometimes gets that little voice in the back of the head telling you not to try something you're about to try. Either because it's stupid or because...

Ok, so the router I had to configure seems to do it's work. I should adjust some of the threshold, because this morning at work I got twenty e-mails from a one hour period telling me the router had switched to ISDN and back.

In any case, it works, so I can be happy, the boss can be happy, and I won't get calls at 5.15 AM. All in all, not bad.

This weekend I started working on converting the Dave/Dina box to Red Hat 7.2. One of the goals would be to allow for re-installs of selected partitions. Right now, the box has two "production" partitions, alpha and omega, and two development partitions, dev1 and dev2.

The idea here is that there will always be at least one working partition which is not touched. New stuff is installed on the other production partition. The first dev partition is used to actually create the RPM's with all of the stuff you need to compile them. The second dev partition is used to try out new stuff all of the time, on top of the current base install.

One of the hassles in the past was that I could not figure out how to put multiple KickStart configuration files in the boot disk. Now that I've figured that out (you have to get the kickstart files inside of the initrd.img file system and give an extra option to the kernel in syslinux.cfg) I've been able to create a boot disk which allows you to do a fresh install of the whole disk, or only re-install one of the partitions.

Now I just have to weed out the package selection (Red Hat 7.2 installs so much these days) and get the post-configuration stuff right. I've also allowed myself the time to make some sort of ASCII "art" as well ;-)

Now I have to reconfigure and recompile the kernel. Can anyone tell me if the sources installed in the kernel-sources package are the redhat-patched versions, or if they get installed as clean kernel sources ? Or is there an easy way to recompile the kernel through the RPM, but with my .config file ?

Ok, so the router I had to configure seems to do it's work. I should adjust some of the threshold, because this morning at work I got twenty e-mails from...

We've been having major network problems at work lately : our ADSL went bonkers with outside ping times of over 13 seconds, our leased line connection couldn't be used, and the Windows 2000 server started acting up, including the first blue screen on this machine in about a year and a half.

So I finally convinced my boss to let me install a dedicated linux router to handle all of this stuff, and put an ISDN modem and regular modem in that to be brought up when the standard ADSL fails.

I had no idea that the basics would be this easy : my boss gave me an old pentium he still had lying around. I plunked in the Red Hat 7.2 disc, booted, installed packages, gave it some IP info, and put it on the net. Then I just did echo 1 > /proc/sys/net/ipv4/ip_forward, gave my own machine this new machine as a gateway, and hey presto ! It worked.

Right now I'm going to a few routing HOWTO's, to learn some more on how to control bandwidth and stuff. With good reason : one of the main reasons the ADSL connection went fully saturated is because someone here thought it was a good idea to send a 2.5 MB attachment to 70 different people. Some people should be shot for sending more mail MB's in one mail than I do in a whole lifetime !

And I should be shot for having allowed this behaviour in the first place, but one can go insane listening to whining of marketing people who really insist on sending 5MB+ powerpoint presentationts through e-mail. And they make me feel like a dinosaur trying to explain them why e-mail shouldn't be used for this ...

We've been having major network problems at work lately : our ADSL went bonkers with outside ping times of over 13 seconds, our leased line connection couldn't be used, and...