Jay Harris is Cpt. LoadTest

"We have a great facility and a big chunk of time, we just need to fill the slots with all the brilliant speakers in the area. So if you are willing to come and share your expertise with your fellow developers in the area, please lend a hand." ~ Jeff McWherter

Scott Hanselman posted an entry yesterday about Managing Multiple Configuration File Environments with Pre-build Events. His design uses pre-build events in Visual Studio to copy specific configuration files to the default file name, such as having "web.config.debug" and a "web.config.release" configuration files and the pre-build copying the appropriate file to "web.config" based on which build configuration you are in. This is a great idea, but large web.config files can get tedious to maintain, and there is a lot of repeated code. Even using include files as Scott suggests would help, but major blocks, such as Application Settings, may have to be repeated even though only the values change. This exposes human error, since an app setting may be forgotten or misspelled in one of your web.config versions.

At Latitude, we manage this problem through NAnt. One of our former developers, Erik Nelsestuen–brilliant guy–authored the original version of what we call "ConfigMerge". Essentially, our projects have no web.config under source control. Instead, we have a web.format.config. The format config is nearly identical to the web.config, except all of the application settings and connection strings have been replaced with NAnt property strings. Rather than have a seperate web.config for each environment and build configuration, we simply have NAnt property files. Our build events (as well as our automated build scripts) pass the location of the format file and the location of the property file and the output is a valid web.config, with the NAnt property strings replaced with their values from the environment property file.

It's simple. It only takes one NAnt COPY command.

default.build

<project default="configMerge">
  <property name="destinationfile"
    value="web.config" overwrite="false" />
  <property name="propertyfile"
    value="invalid.file" overwrite="false" />
  <property name="sourcefile"
    value="web.format.config" overwrite="false" />
 
  <include buildfile="${propertyfile}" failonerror="false"
    unless="${string::contains(propertyfile, 'invalid.file')}" />
 
  <target name="configMerge">
    <copy file="${sourcefile}"
        tofile="${destinationfile}" overwrite="true">
      <filterchain>
        <expandproperties />
      </filterchain>
    </copy>
  </target>
</project>

For an example, lets start with a partial web.config, just so you get the idea. I've stripped out most of the goo from a basic web.config, and am left with this:

web.confg

<configuration>
  <system.web>
    <compilation defaultLanguage="c#" debug="true" />
    <customErrors mode="RemoteOnly" /> 
  </system.web>
</configuration>

In a debug environment, we may want to enable debugging and turn off custom errors, but in release mode disable debugging and turn on RemoteOnly custom errors. The first thing we will need to do is create a format file, and then convert the values that we want to make dynamic into NAnt property strings.

web.format.config

<configuration>
  <system.web>
    <compilation defaultLanguage="c#" debug="${debugValue}" />
    <customErrors mode="${customErrorsValue}" /> 
  </system.web>
</configuration>

Next, we need to make NAnt property files, and add in values for each NAnt property that we've created. These property files will include the values to be injected into the web.config output. For the sake of simplicity, I always give my property files a '.property' file extension, but nant will accept any file name; you can use '.foo' if you like.

debugBuild.property

<project>
   <property name="debugValue" value="true" />
   <property name="configMergeValue" value="Off" />
</project>

releaseBuild.property

<project>
   <property name="debugValue" value="false" />
   <property name="configMergeValue" value="RemoteOnly" />
</project>

Finally, we just execute the NAnt script, passing in the appropriate source, destination and property file locations to produce our environment-specific web.config.

nant configMerge -D:sourcefile=web.format.config -D:propertyfile=debugBuild.property -D:destinationfile=web.config
web.config output

<configuration>
  <system.web>
    <compilation defaultLanguage="c#" debug="true" />
    <customErrors mode="Off" /> 
  </system.web>
</configuration>

And that's all there is to it. This is extendable further using the powers of NAnt. Using NAnt includes, you can put all of your base or default values in one property file that's referenced in your debugBuild.property or releaseBuild.property, further minimizing code. You can use Scott's pre-build event idea to have each build configuration have its own NAnt command to make mode-specific configuration files.

Feel free to use the above NAnt script however you like; but, as always YMMV. Use it at your own risk.

Enjoy.

I need to just make time to blog. I’m finding that I am getting behind, further and further, as the bleeding edge of technology pulls out ahead of me. [sarcasm]Keeping up to date is hard.[/sarcasm].

I am working on a few things to try to get myself back up to speed, beyond just the on-going procrastination over Microsoft .Net 2.0 certification. Last month, I attended Jay Wren’s talk on IoC with Castle at the Ann Arbor .Net Developers Group. Though we’ve been using Castle within our LMS for a few months now, patterns is definitely one of my weak points in developer-land, so I did pick up a few concepts on Dependency Injection. It was a great talk, and I got some swag!

Last week it was back to AADND for Michael Wood’s talk on Windows Workflow Foundation. I haven’t had much any exposure to Windows Workflow Foundation other than to not call it WWF. This stuff blows me away, and makes me even more jealous that we haven’t converted the LMS out of .Net 1.1. I wish I could have made it to the GANG meeting tonight on WF, just to learn more.

Tomorrow is another meeting, and another “Geek Fest” as my wife so lovingly refers to it as. Heading west to MSU for GLUG’s meeting on C# 3.0 by Bill Wagner. There’s going to be a bit on Code Rush, too (I love CodeRush!). Maybe I’ll win more swag!

CruiseControl.Net 1.3 was released this morning. To me, most important was the new Build Queue functionality, stopping multiple projects from building at the same time. If ProjectB depends on ProjectA, and they both get code changes committed at the same time, they will fail from contention. Either ProjectA will fail because it can’t delete its old assemblies (because ProjectB has a lock on them) or ProjectB will fail because it can’t find the ProjectA assemblies (because ProjectA deleted them in its rebuild).

No more.

I’m so excited that I am already upgrading our servers!

All right. As mentioned before, here it is: I found a bug in mx.managers.DepthManager in Flash 8, though I’m sure it affects all ActionScript 2.0 versions of the DepthManager to date. The bug involves a lack of error handling when DepthManager encounters an object without a depth, or rather, without a getDepth() method. It shows up through any use of the DepthManager’s methods, the PopUpManager (mx.managers.PopUpManager), the Alert class (mx.controls.Alert), etc., when an object on the stage does not have a getDepth function.

What Happens?
The issue manifests itself in many variations, usually by objects placed with the DepthManager inexplicably replacing previous objects placed with the DepthManager. Most notably, it manifests itself as modal windows (via PopUpManager) and Alerts not rendering to the screen, even though the modality seems to have worked and all other objects on the screen no longer can receive input.

Why does this happen?
The DepthManager works off a rather inefficient array to maintain its records. An object within the DepthManager, the DepthTable, is an indexed array containing record of every object on the stage and its depth. The DepthManager stores a reference to the object within the array at the position corresponding to the depth of the object. If the stage has one object at a depth of 32,000, depthTable[32000] = myObject, and we have a 32,000-length array with one object stored in it [an inefficiency]. Values in the manager’s depth table are based on available positions in the array, such as kTopmost being greater than the last movieclip or object in the array [it iterates one-by-one through each position in our 32000-length array; another inefficiency].

Excerpt from mx.managers.DepthManager.as

var t:String = typeof(i);
if (t == “movieclip” || (t == “object” && i.__getTextFormat != undefined))
   if (i._parent == this) {
      depthTable[i.getDepth()] = i;
   }

However, the DepthManager code is assuming that the object has a getDepth function. If the object does not have the function, references to that function return undefined, setting depthTable[undefined] = i, which blows chunks all over the place, and any reference to any of DepthManager’s depth-returning methods will always return zero.

That zero is why nothing works anymore; the Alert class creates the new Window-class-like object, then creates a transparent movieclip below that object to trap all inputs, creating a sense of modality. The Alert class uses the DepthManager to create itself at kTopmost, which DepthManager says is zero. The Alert class then uses the DepthManager to create the modality movieclip above itself, kTopmost, which it then hopes to swapDepth with to get itself back above the modality clip. However, the Alert class, with itself at depth zero, creates the modality clip at depth zero because DepthManager said it was the kTopmost depth. Creating an object at depth zero toasts any object already at depth zero, thus the modality clip blows away the Alert clip and we get a SWF that no longer responds to anything, and no Alert ‘OK’ button to restore peace in the world. Likewise happens when a modal window is created through the PopUpManager.

So what can I do about it?
How do you get around this? Simple. Don’t put anything on the stage that extends from Object. If you put it on the stage, extend it from MovieClip. Even though the DepthManager code (above) explicitly calls getDepth on MovieClip or Object, it does not check to see if the object in question actually has a getDepth method. Object does not have getDepth by default, but MovieClip does. I suppose you could write your own getDepth function for your Object-inheriting class, or you could rewrite the DepthManager (which would mean also rewriting PopUpManager, Alert, ComboBox, and a half-dozen other controls), but it is much simpler to extend from MovieClip.

Why it happened to me
I have a movie that is 170KB. There’s a lot of crazy mojo in there, and I want my prel0ader to load within the first 1KB, not after 170KB. So, I created my preloader, removed “Export in First Frame” from everything, and set my classes to export in frame 5 (after the end of my preloader loop). However, this made a few of my Object-extending components no longer work anymore since they were not referenced in the “Assets” layer of any of the other components (no reference=no load, much like the DataBindingClasses), to I had to drag them to the stage to ensure that they loaded.

*Poof*

My world exploded.

I lost two days of work tracking this little punk down. And you already know the rest of the story.

The day is here: Internet Explorer 7 released this morning, though it is only available to users running Windows XP SP2 or Windows Server 2003. Though it does not yet appear on the list of available updates through Windows Update, you can download it from Microsoft.com. [ News: ZDNet | CNet ] In a CNet review, the reviewing editor notes how the browser is still not compliant with standards set by W3C and recommends switching to Firefox.

“IE 7 was Microsoft’s one chance to leapfrog ahead of the competition, but the company has only barely caught sight of the current front-runners. For more features and greater security, switch to Mozilla Firefox.” ~ CNet [ article ]

There is already a version branded by Yahoo! that includes the Yahoo! toolbar, links to Yahoo’s tools (like Yahoo! Mail), and the default homepage set to Yahoo. Of course, Yahoo! is catching some flak, since it released its “optimized” version prior to Microsoft’s own official release. [ ZDNet ]

Mozilla plans to release Firefox 2.0 in the coming weeks. You can download the current beta, Firefox 2.0 RC3, or the current public release, Firefox 1.5.0.7, from Mozilla.com. [ v2.0 RC3: download | release notes. v1.5: download | release notes ]

Both NAnt and NAntContrib released version 0.85 on Sunday. The changes to NAnt from 0.85 rc4 only include a few bug fixes. NAntContrib has added the ability to specify the encoding on SQL files. All-in-all, not much has changed since 0.85 rc4, but that is a good thing, since it indicates the version is finally ready for release. The first release candidate was made available nearly two years ago.

Despite the minimal changes in the final package, consider upgrading just to get rid of the ‘release candidate’ tag.

NAnt v0.85 [ homepage | download | release notes ]
NAntContrib v0.85 [ homepage | download | release notes ]

I’ve been working on creating a Flash component for within our LMS that allows a student to view their completion statuses in a curriculum, as well as allows managers to view the curriculum status of all of their students. The component relies on Web Services to volley data with the server. I was having some trouble binding data from my WebServiceConnector to my DataSets because, as it turned out, Flash was holding on to a cached version of the WSDL–a cached version that was ultimately obsolete.

I was on the hunt to find the source, and kill the evil cache file. After clearing out every Temp folder I could think of, I started playing around in my Local Settings folder and it turns out that Flash will forever-store cached copies of your WSDL deep within your ‘Documents and settings’.

C:\Documents and Settings\[user name]\Local Settings\Application Data\Macromedia\[your flash version]\en\Configuration\WebServices\

I just booted every file in the WebServices directory, but I am assuming that you could get away with just deleting ‘WSDLCacheMap.xml’, or if you want to keep the remainder of your cache, just delete the appropriate WSDL cache file (ex: WSDLkgzmcu.wsdl). In any event, once you have exercised your delete button, simply restart Flash and reload the WSC instance, and you will be good to go.