hobolog :: Error Collection

I haven’t posted an error message in a long time, but this one is too good to pass up. I did nothing unusual to make this happen, it just popped up while I was setting up a new laptop with Windows XP at work:

I nearly sprayed the new laptop with a mouthful of hot coffee.

A recent change to Google Maps brought this gremlin to the hobomap and many other Google Maps API users also. I found the explanation at Mapki. The cardinal sin is putting the map-creating javascript inside a <div> or other tag. It has to be right before the closing body tag.

This is a bit painful for some users of my geo mashup plugin. I’ve tried to remove all requirements on the page template, but this imposes a requirement again: the <?php wp_footer() ?> tag must be right next to the </body> tag, and not enclosed by other tags. Thankfully many existing templates meet this requirement already.

This is a typical error from the software I’m working with these days, Sungard SCT’s Banner, a “collegiate administrative solution”. We’ve been living with this particular error, which pops up regularly on one of the most used forms in the system, at Colorado State University for months, with no end in sight.

This error includes a little bit of information – a number and a variable name. Googling the number turned up this:

Cause: Application design error. A trigger references a global variable that does not exist. (from lazydba.com)

If you have a client login on SCT’s web site, you can see their answer. They acknowledge the defect and promise a fix in version 7.0.0.2 (we’re on the bleeding-edge 7.0). Of course we just went live with this version and an upgrade would send everyone into upheaval. This an example of how many people grow so tolerant of software misbehavior – so often the malady causes less pain than the cure.

If your daily computer use does not provide you with enough error messages, this handy tool will help you create more.

I’m collecting errors on a new platform! This is from Mac OS X 10.3. I’ve gotten the debugger up and am far enough to see that the error happens when the program tries to create a big (512k) chunk of shared memory. Smaller chunks work.

In 10.2 I had upped the shared memory from 4MB to 256MB in /System/Library/StartupItems/SystemTuning/SystemTuning, but that no longer works.

ERROR: Relation 29384293 does not exist

This is a PostgreSQL error, caused by me dropping an index while queries were running. Ever wondered if you can delete an index while it’s in use? Now I know.

ERROR: deadlock detected

This error is popping up in my PostgreSQL database server logs. It’s supposed to be pretty hard to get a deadlock, so we must have some good code out there. We have a routine that’s been hanging up lately that is the prime suspect, but it would be a job to find exactly why and where it’s deadlocking. The author has taken a stab at making it better, so maybe the error will go away and we won’t have to. Will I ever truly solve an error?

(gxine:3834): GLib-CRITICAL **: 
file ghash.c: line 225 (g_hash_table_lookup): assertion `hash_table != NULL' failed

(gxine:3834): libgnomevfs-WARNING **: 
Internal error: the configuration system was not initialized. Did you call _gnome_vfs_configuration_init?

(gxine:3834): GLib-CRITICAL **: 
file ghash.c: line 225 (g_hash_table_lookup): assertion `hash_table != NULL' failed

(gxine:3834): libgnomevfs-WARNING **: 
Internal error: the configuration system was not initialized. 
Did you call _gnome_vfs_configuration_init?

I was trying to get a player that would play the KCRW simulcast radio station we like to listen to. The RealPlayer for Linux would work for a while, then hiccup and spew noise until the machine was powered off. This message is from the gxine player. The funny thing is that the error sounds pretty bad, but the thing works anyway.

Here’s a situation where I want an error message and can’t get it. My ’96 Subaru Legacy’s “check engine” light has come on, and some googling suggested that I might be able to retrieve an error code that would tell what is wrong. After reading some instructions and making several trips out to peer under the dashboard, I concluded I was missing a connector. Finally I found the bad news: 1995 was the last year you could retrieve error codes without a “scanner”.

So the modern solution is to leave you with no idea whether it is safe to drive your car to a dealer (77 miles away), or whether it must be towed. Nice.

Assertion failure in 
journal_commit_transaction() at commit.c:535: "buffer_jdirty 
(bh)" 

Came in to find this message on our trusty dual-processor database server this morning. It’s on RedHat’s errata page, the first fix for 7.3, a problem with the Symmetric Multiprocessor Kernel. The annoying thing is that the kernel with just that fix is no longer available, they want me to go to a kernel that requires lots of other updates in addition. It might be easier just to upgrade to RedHat 8, but I’ll try first to find a less demanding fix…