I Make Things - Work

mod_rewrite errors on Mac OS X Server

nospam@example.com (Latcarf) — Tue, 27 May 2008 20:14:05 -0400

I was a little concerned with the web service on my new Intel Xserve running Mac OS X Server 10.5 (Leopard) when I started seeing the following log entry in my Apache error log:

[crit] (2)No such file or directory: mod_rewrite: could not init rewrite log lock in child

I wasn't so much concerned that there was a critical error. I could see my Apache 2 based Subversion server (DAV svn) was performing just fine. My larger concern was that there was an error about mod_rewrite when I wasn't the one that turned it on in the first place. You see, I could believe it was my human error that might have caused the problem but I'd have a harder time believing it was a problem with the default install.

So, I looked at my config file: /etc/apache2/sites/0000_any_80_.conf and compared it to the default config file: /etc/apache2/sites/0000_any_80_.conf.default and found the <IfModule mod_rewrite.c> containers to be exactly the same. Annoying.

Iquickly gave up trying to point the finger and lay blame and decided to take the initiative to fix the problem. I fixed it by giving mod_rewrite an actual log to write to (make it happy - if for only a moment) and then told it to not to log anything to the file (mwahahaha). The change was simple.

From this:

<IfModule mod_rewrite.c>
RewriteEngine On
RewriteCond %{REQUEST_METHOD} ^TRACE
RewriteRule .* - [F]
</IfModule>

To this:

<IfModule mod_rewrite.c>
RewriteEngine On
RewriteCond %{REQUEST_METHOD} ^TRACE
RewriteRule .* - [F]
RewriteLog /var/log/apache2/rewrite.log
RewriteLogLevel 0
</IfModule>

Problem solved it seems. Now why did I have to do this!?

New Xserve Build Machines

nospam@example.com (Latcarf) — Thu, 13 Dec 2007 18:47:30 -0400

I just got two brand new Xserve build machines and the first thing I did was wipe out the default Tiger OSX Server install and install Leopard OSX Server instead. That wasn't too bad except for the fact that I had to go through the setup assistant first in Tiger before being able to boot off the Leopard disk, wipe the drive and install.

After I got Leopard OS X Server up and running on the Xserve, I tried to ssh out from a user account to another machine (my source repository server) that's running 10.4. While establishing communication, ssh had this huge delay (more than a minute - I think). I had to figure out what was going on so I ran ssh -v to get the verbose output from ssh while it was connecting.

The verbose output showed this.

debug1: Unspecified GSS failure.  Minor code may provide more information
Server not found in Kerberos database

debug1: Unspecified GSS failure.  Minor code may provide more information
Server not found in Kerberos database

debug1: Unspecified GSS failure.  Minor code may provide more information

So, I went spelunking in and ended up comparing the /etc/sshd_config files on both the local (Leopard Xserve) and remote (Tiger source repository) machines. I found on the remote machine (Tiger source repos) that Kerberos authentication was turned off by default but commented out.

So, I changed this:

# Kerberos options

# GSSAPI options
#GSSAPIAuthentication no

to this:

# Kerberos options
KerberosAuthentication no

# GSSAPI options
GSSAPIAuthentication no

in the /etc/sshd_config file on the remote (Tiger OS X server) and reconnected via ssh without a problem. Then it was on to more work setting up my new xserve build systems.

Manager/Employee One-to-Ones

nospam@example.com (Latcarf) — Mon, 29 Jan 2007 17:58:17 -0400

My manager today decided to cancel all of her current one-on-ones including mine (I prefer to call them one-to-ones - so they don't sound like a wrestling match). She noted that she'd setup "quarterly goal review 1-1 focused exclusively on career development" and that she would "work with each individual to determine whether a weekly 1-1 is appropriate."

It's amazing that managers, whether new or seasoned, don't fully grasp the importance of one-to-one time. In here defense, she does have around 17 direct reports now - but that's a failing of her managers. I thought about sending her the Manager Tools link to their podcast on "the single most effective management tool - the one-on-one" but I figured I'd look like I was telling her how to do her job. Obviously, that wouldn't go over too well. Maybe I could leave an anonymous note with the link to Manager Tools on it...

Instead, I think I'll listen to the podcast again myself and the next time I talk to her, (probably at my quarterly one-to-one) I'll try to bring up some of the points Mark and Mike make. Also, I think I'll privately review (and possibly fill out) the Manager Tools: 1-on-1 Key Points and Prep Form (PDF). After I've done that, I'll start looking for a new position. I don't have time for inattentive and overburdened managers anymore.

Robust Shell Scripting

nospam@example.com (Latcarf) — Tue, 23 Jan 2007 20:46:22 -0400

I started working for a new group around mid-September and inherited an old (5 years or so) shell-script build system. Most of it is in KSH (I really prefer BASH) but alot of it is so poorly written and thought out that I have a hell of a time trying to fix/enhance it. Every time I need to do something to the scripts I oscillate for an hour on whether I should try to work my change into the existing script or just take the time to re-write the whole thing!

I was Googling today for some thoughts about writing shell scripts and found a great article about writing robust bash shell scripts. While I don't use everything mentioned in the article, I was happily surprised to find that I do use more than a couple of them.

Of course, one reason for my irritation with the system I've inherited is that it doesn't use any. Before working in this group, all SCM infrastructure I'd done was built ground up by myself. I feel for those poor developers that get stuck trying to enhance old code now that I've really gotten a taste of how hard it can be. I imagine there is alot more nasty build system scripts out there than there are good ones but I'll certainly try to even the odds a bit by making my scripts as robust, readable and commented as possible.

Hopefully someone who comes three or four years after I'm one will be pleasantly surprised when they have to add something to my scripts!

Marketing, Schmarketing

nospam@example.com (Latcarf) — Mon, 31 Jul 2006 13:03:48 -0400

<rant>
I've always held a general disdain for marketing people. Not all marketing people mind you, just most of them. I used to think it was because they are so often self-important, arrogant and close-minded. Many of them hold themselves in such high regards that they think the software industry (or other industries) just can't go on without them. When I think of marketing, I think of slimy snake oil salesman, so enamored with the fake little world they've created that they rarely ever pitch the actual thing we are all making.

Anyway, I realized this morning the real reason I'm so annoyed by them. It's just that for all they think they're worth, they generally never make anything lasting. There are exceptions to this rule of course. Some marketing campaigns we remember for a few years and actually leave a lasting impression on us, but those are the exceptions. In general, we forget marketing campaigns shortly after we've been subjected to them. The have absolutely no longevity. Now I know that everything in life is impermanent but it's a noble goal to create things which benefit society and last long enough for people to actually remember the impact those things had on their lives. Marketing is in the business of making trash, and trash has limited use, limited life-span, and limited (if any) benefit to society.

This is why I don't really have any respect for marketing people.
</rant>

Change Subversion Commit Messages Post Commit

nospam@example.com (Latcarf) — Fri, 12 May 2006 14:56:02 -0400

Today I had a developer who wanted to change his commit message after he'd already made the commit to the Subversion repository. Initially, I told him it wasn't possible, but then I decided I should check if my assertion was correct. It wasn't...

To change/edit a Subversion commit message after the commit, do the following on the Subversion server:

$ svnadmin setlog /path/to/repository \
> -r 10662 --bypass-hooks message_file.txt ¬

The --bypass-hooks argument is necessary because you are editing a non-versioned property of the revision. As such, you lose the history of the original commit message so Subversion requires the pre-revprop-change hook to be enabled. Passing --bypass-hooks bypasses this requirement and allows the change.

Collective Mental Acuity

nospam@example.com (Latcarf) — Fri, 06 May 2005 17:58:54 -0400

It's an interesting phenomenon in the software industry that no matter how well you've done your job, you inevitably find more and more issues the closer you come to a release date. My group is near a beta release and it is peculiar how everyone gets "hypersensitive" as we near the date.

By hypersensitive, I don't mean everyone is at each other's throats (although I have seen that happen in other groups). I mean people become very sensitive to all aspects of what is being delivered. The most minute details that have been overlooked until now, are seen with surprisingly clarity as we inch toward the day of release.

Somehow, I believe our collective team consciousness plays off of one another's anticipation, and causes each of us to scrutinize things even more than we did prior to the impending deadline. In this way, our "hive-brain" becomes far more acute and sensitive and allows us to individually "see" much better.

It would be nice to learn the manner of how this happens and exercise control over the faculties in charge. The power to engage this super sensitivity whenever needed would be a beneficial one indeed. I don't think it could ever be an always-on characteristic though, as I think our bodies and minds cannot sustain it for too long. The group is always exhausted after a release - no matter how minor it may be.

Configuration Management Resources

nospam@example.com (Latcarf) — Fri, 04 Mar 2005 22:21:16 -0400

Lately I've been working very hard to increase my knowledge and understanding of my new role at work (Software Configuration Management Engineer - SCM). I am the only SCM for two products and a total of ten developers (though I support around 20 people between the two teams in total).

Both development teams loosely base their process in the Agile style of Extreme Programming. I have found a few resources that seem to be quite good for information on my position and role.

CM Crossroads is an online community and resource center for configuration management. The have a wealth of information at their site and they do a good job of recommending books, best practices, and software to make my job easier and the development teams more productive.

I'll be detailing things I learn here in a new category: "SCM" under the category "Work".