I just spent way too much time troubleshooting what should be a very simple SVN-related problem. Using svnant (1.1.0-RC1 and 1.1.0-RC2) I kept getting the same error
[svn] svn: authentication cancelled
It’s only been two years, 11 days, and change since my last post. Look for a *lot* more to be coming from me and TransitionPoint in the coming weeks. Assuming anyone still reads this.
I like to organize my Eclipse plugins in multiple extension locations — that way when I have Eclipse 3.0.x, 3.1.x, and some very specific install like Eclipse with the Business Intelligence Reporting Tools (BIRT) all on the same machine, I don’t have to install plugins and update existing ones on multiple versions. This process is discussed in article at Javalobby.
I organize the Eclipse installs like this
\eclipse-3.0.2
\eclipse-3.1.0
\eclipse-someReleaseMx
I then have a directory that contains all of the collections of Eclipse plugins organized by theme
d:\eclipse-extensions\common\eclipse
d:\eclipse-extensions\lamp\eclipse
d:\eclipse-extensions\cfeclipse\eclipse
d:\eclipse-extensions\test\eclipse
d:\eclipse-extensions\v3only\eclipse
d:\eclipse-extensions\v31only\eclipse
One problem though, is moving the plugins, from test to common for example. It’s easy enough to manually do the move, but it’s crucial to also restart eclipse with the -clean flag
D:\eclipse\eclipse-3.1\eclipse.exe -clean
Without the -clean flag, Eclipse uses cached information about the plugin(s) that were moved which can cause problems.
As an aside, I also found out that some plugins (AnyEdit, for one example) must be installed in the main \plugins directory for the currently running Eclipse instance to work properly. This is the first one I’ve noticed like that — moving it to one of the additional Extension locations caused part of it become disabled.
After two consecutive Sundays evenings with a locked CFMX process on a clients MS-SQL 2000 box, I went digging through the db logs and found a problem with the midnight integrity check in their database. When I tried to modify the plan to remove the problematic job, I got a surprising error — “Error 14274: Cannot delete a job that originated from an MSX server”. Turns out that I couldn’t edit/delete any of the jobs on the server. That seemed vaguely familiar from a long-ago MSSQL DBA gig I had, but I couldn’t place it so of course it was off to google…
Long story short, that message is common when the name of the database server has been changed after the scheduled job has aready been set up. And fortunately, it’s easy to fix. In the msdb database, there is a table named sysjobs — verify that the originating_server column contains a name different from the current server name. If the name is incorrect, simply run this SQL
UPDATE sysjobs SET originating_server='yourNewServerName'
This assumes, of course, that you’re dealing with standalone MS-SQL servers — in a cluster or some other enterprise configurations, it may well be that another server originated a job on the local server, in which case the error message is 100% appropriate.
I spent a few minutes yesterday between conference calls installing the new Firefox 1.1 alpha release, codenamed Deer Park. Wow — the browsing speed improvement is noticable, especially if you use the back/forward keys a lot. And all of my extensions with the exception of FireFTP and DownloadThemALL are working fine. I’d still back up the Firefox 1.0.5 profile, just in case, but I’ve been running both of them without issue. I’m on my 2nd day and have yet to run into a major issue.
One minor, but cool new feature is the ability to drag tabs around to reorder them — I seem to recall extensions that provided that capability, but now it’s built-in.
One minor quirk is that I haven’t been able to get both of them to run simultaneously — once one of them is open, starting the other one creates a new instance of the version that’s already open — which is weird since the executables are installed in different directories. So if I’ve got DeerPark open and use my Firefox shortcut, I get another DeerPark browser instead of Firefox 1.0 — same if I open Firefox and then click on DeerPark.
Sitting in my ersatz office (aka the local B&N) I finally noticed the CD playing in the background…. sounds Sinatra-y, so I’m thinking it’s got to be the Starbuck’s latest boy toy, Michael Buble, belting out a standard. Then I start listening:
Blackhole sun, won’t you come? And wash away the pain?
Huh? That’s, umm, BlackHole Sun by SoundGarden. Then it’s Van Halen’s Jump. And to mix things up with a torch song, Lionel Ritchie’s Hello. Interestingly they work really well. I had heard one of his albums that my wife received for her birthday from a friend and thought it was perfectly adequate background music. Now I’m intrigued.
UPDATE: Actually, it’s Paul Anka (!), not Micheal Buble, from the album Rock Swings
Today I had my first stumper as far as using SVN — I was pulling from the SVN repository of a new client and kept getting this weird error. The error was
svn: In directory 'dia/src'
svn: Can't copy 'dia/src/.svn/tmp/text-base/TAGS.svn-base' to 'dia/src/TAGS.tmp'
: The system cannot find the file specified.
So what the heck is going on? Tried Tortoise and svn command line. Tried looking through the repo browser in Tortoise and didn’t see anything useful. I’ve used http:// and file:/// protocols a lot more than svn:// but didn’t think that was it. I upgraded to the latest svn (1.1.4) and tortoise (1.1.7) but still the same thing.
Then it hit me — there was already a TAGS file in the dia/src directory…. it just happened to be named “tags”. No problem on the FreeBSD server or the Linux clients, but definitely a problem on my WinXP box. I guess I couldn’t see it through the Tortoise repo-browser because it was a file with the “same name” as another file?
Turns out one developer is using emacs and one is using vi and both TAGS and tags got inadvertently checked into the repository. Removed those files and we’re now in business.
So I’m installing Windows 2003 Server on a development machine and as I sit in front of the screen happily erasing partitions, I realize that my USB backup drive is plugged in. Of course I don’t realize this until after I erase the partition. Crap!
Now intellectually, I know that since it’s not a bootable partition that I haven’t done anything unfixable — booting a Linux OS on that box (Knoppix since that’s sitting here handy) shows me the files are still there. But the freakin’ USB drive is basically full of data and I don’t have a big enough drive handy. I could run out and buy a new one and install it, but I should simply be able to restore that partition information somehow, right? I mean it should just be default partition information since there’s nothing but data on this USB drive…
So I dig around a bit, in the process finding quite a few panicked users posting to various forums about doing exactly the same thing (deleting the partion on a USB drive during an OS install) — so while I’m feeling like an idiot, at least I know I’m not alone.
Turns out that there’s a slick little GNU tool for handling just this situation — TestDisk. This tiny app can handle partition problems on more partition types than you can shake a stick at (BeOS, CramFS for example). Using it I wrote a new, default master boot record to the USB drive, unplugged it from the laptop, plugged it back in and was back in business.
For the record, I also came across a Windows-oriented tool for pulling the files off of deleted partitions, reformatted drives, etc. It seemed to work ok (it immediately found the partition and I could save files from the original partition to a new drive, but I couldn’t figure out how to simple write the partition information to reset the drive back to normal using it. It’s from PCInspector for those that are interested.
As promised, the slides for my Data Warehousing with MySQL tutorial are now online.
The are slightly updated to fix a couple of typos. I also was gently chided by the MySQL folks for using “Table Type” instead of “Storage Engine”, which is a perfectly valid point so that’s now been changed.
This tutorial was written using the open source S5 presentation tool, which is an open-source, standards-compliant web-based presentation tool from CSS guru Eric Meyer. The zip file contains an HTML file containing all of the content and a /ui directory that contains some stylesheets, graphics, and javascript. It also contains a 16-page PDF printout of the slides. It is licensed under the Creative Commons Attribution-NonCommercial-NoDerivs License, which means you’re free to use it for non-commercial purposes with no modifications.
Now that the past month is behind me - featuring not only the previously mentioned family flu but also a week vacation featuring a four-year-old with a week of gastroenteritis - I,m shifting into conference season. With my tutorial on Datawarehousing with MySQL tomorrow at the MySQL users conf and the preliminary talk for CFUNITED due this past week it’s been busier than normal. That said, the vlog backlog is slowly catching up - thanks in no small part to Plogit for my new Treo 650. Blogging from a bar in San Francisco waiting for my table may be geeky, but it’s catching me up