Thursday, December 01, 2005

Today I ran into an issue that I had not encountered (at least not in this manifestation) while compiling an assembly so I thought I'd share the solution.  The error was simply:

Cryptographic failure while signing assembly '....' Access is denied.

As it turns out I had just rebuilt my laptop from the ground up.  I don't run my computer in an administrator capacity and my process of installing was, this time, a little different that how I usually rebuild my machines; I made sure I wasn't an administrator immediately upon having installed Windows.  However, rather than temporarily promoting myself to Admin via the MakeMeAdmin utilities to install my development environment, I installed VS.NET while logged in as the Administrator user directly.  The byproduct of this was that my non-administrator user account was never granted permissions to a required directory for signing assemblies.

To remedy the problem I simply opened an Explorer window as an administrator and altered the ACLs on the necessary directory.  I granted my user Full Control on %ALLUSERSPROFILE%\Application Data\Microsoft\Crypto\RSA\MachineKeys.  Simply restart Visual Studio (if it was open) and you should be off and rolling.

Thursday, December 01, 2005 3:57:00 PM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [26]  |  Trackback

Today I had the pleasure of attending the VS.NET 2005 / SQL Server 2005 Launch event here in Salt Lake City, Utah.  The event was held at Jordan Commons and it's a good thing too.  We had one of the theatres completely packed.  About 700 people attended the afternoon session on VS.NET and almost that many in the morning for the SQL Server event.

It was a lot of fun to once again make contact with so many friends and associates and partake in the geekdom.  The opportunity was mine to work on the event staff, as I am inclined to do for these events, participating in the various community events around the valley and neighboring cities..  It was great to greet people and then catch periodic glimpses of the sessions.  What a follow-up to the Geek Dinner that we had last night.

There are days where I love what I do.

Thursday, December 01, 2005 9:09:00 AM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [0]  |  Trackback
 Wednesday, November 30, 2005

We had an awesome Geek Dinner tonight organized by DevUtah and sponsored by DevUtah and ProvoLabs.  We had about 65 people show up for a fantastic dinner at Los Hermanos.  After a few logistical issues we all got settled in and had a great presentation by Dr. Phil Windley on microformats.  It was a lot of fun.  We then each had the opportunity to share information about the various user groups around the valleys (.NET, PHP, Linux, Unix, Ruby, etc).

When Phil and I started DevUtah we were hoping for some community enthusiasm and I think we got it...what fun!  I just wish that more of the .NET group had shown up - virtually no one from the .NET User Group was there which was very disappointing.  The vast majority of attendees were BYU students and open source enthusiasts (but there's nothing wrong with that).

Our plan is to be able to put on these Geek Dinners once a month, and I hope this enthusiasm continues.

Wednesday, November 30, 2005 4:03:00 PM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [3]  |  Trackback

Today, in less than 6 hrs we're having a Geek Dinner at Los Hermanos Restaurant in Lindon, Utah.  The event is open to everyone and all are invited.  If you'd like to come, please RSVP on the wiki.  Thus far 57 people have registered, which is awesome!

Don't forget, there will be giveaways, a great speaker, Dr. Phil Windley from BYU, great food (everyone buys their own plate), wireless internet, etc.  The even will be filmed and podcasted as well.

So sign up and be there tonight @ 6:00 PM.

Wednesday, November 30, 2005 5:42:00 AM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [0]  |  Trackback
 Tuesday, November 29, 2005

About a month ago I posted an article on how I like to use SqlDataReaders.  In said article I discussed how, rather than actually retrieving a data reader from my command object and walking the results, I like to walk the data reader in a callback method.  I propose that this strategy helps constrain how data readers are used and better manages their lifetimes.

I want to take it a step further now and talk about some additional enhancements that can be made to enhance the experience of using data readers all the while leveraging some cool .NET 2.0 functionality.  Last month's post was intended to be informative without dwelling on and nitpicking little details.  For example, sure I could have used anonymous delegates as the callback methods, but that would have muddied the discussion.  Today, however, I'm going down that path :-)

I really like the generic QueryReader method that I proposed last month, but there are some issues with it:

  • The generic method uses a generic delegate callback.  However, using this mechanism you're required to have the callback method return a value of type T.  This isn't always desirous, especially if the callback method simply populates a control and doesn't want to burden itself with building a result set to return it only to have it iterated immediately to populate the control.  Sure, if the list is going to be used elsewhere and/or more generically, then it's great, but not 100% of the time.
  • Along with the previous point, you cannot call an instance of the method with T of type void.
  • If the callback method needs some information (e.g. state) on which to operate other than the data reader the signature needs to be enhanced.

To address these issues, I've made some changes and ended up with the following:

internal delegate T DataReaderCallback<T>(object state, SqlDataReader dr);
internal delegate VoidDataReaderCallback(object state, SqlDataReader dr);

internal static class DataAccess {
   internal static T QueryReader<T>(SqlCommand cm, object state, DataReaderCallback<T> callback) {
      SqlConnection cn = getConnection();
      SqlDataReader dr = null;
      try {
         cm.Connection = cn
         dr = cm.ExecuteReader(CommandBehavior.CloseConnection);
         return ( null != callback ) ? callback(state, dr) : default(T);
      }
      finally {
         if ( null != dr ) dr.Dispose();
         if ( null != cn ) cn.Dispose();
         cm.Connection = null;
      }
   }


   internal static void QueryReader(SqlCommand cm, object state, VoidDataReaderCallback callback) {
      SqlConnection cn = getConnection();
      SqlDataReader dr = null;
      try {
         cm.Connection = cn
         dr = cm.ExecuteReader(CommandBehavior.CloseConnection);
         if ( null != callback ) callback(state, dr);
      }
      finally {
         if ( null != dr ) dr.Dispose();
         if ( null != cn ) cn.Dispose();
         cm.Connection = null;
      }
   }
}

By implementing it as such, you have two overloads: one similar to before where the callback method returns a value and the second simply operates on the reader and is done - no return value is set.  Purists might identify that the 'state' parameter is typed as 'object' rather than generic or typed.  Ok, ok, if you want it, go ahead and make that change.

This now gives us the ability to do the following:

using ( SqlCommand cm = new SqlCommand("SELECT Name FROM Account ORDER BY Name") ) {
   VoidDataReaderCallback callback = delegate(object state, SqlDataReader dr) {
      while ( dr.Read() )
         cboAccounts.Items.Add(dr.GetString(0));
   };
   DataAccess.QueryReader(cm, null, callback);
}

Ok, so it's nothing earth shattering, but I like its simple elegance.

Tuesday, November 29, 2005 10:36:00 AM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [8]  |  Trackback
 Saturday, November 26, 2005

Well, it's finally happening.  DevUtah and Provo Labs (Infobase Ventures) are helping coordinate and put on a Geek Dinner.  The first of hopefully many of these Geek Dinners will be held on Wednesday, Nov 30th, 2005 from 6:00 PM to 8:00 PM at Los Hermanos Restaurantin Lindon.

For full details, please visit the DevUtah geek dinner announcement page.

Dr. Phil Windley will be the keynote speaker and will present on micro-formats and the writable web.

Please visit the website and RSVP - details are on the site.  See you there!

Saturday, November 26, 2005 3:35:00 AM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [0]  |  Trackback
 Wednesday, November 23, 2005

Alright, before railing on me about this and telling me how stupid it is to create this because a simple MSN/Google search would reveal tons of already-available solutions, I admit to having not found any that I really like.  I therefore took the liberty of creating two little .REG files that give you VS.NET 200x Command Prompt Here context menu shortcuts to the folder/drive/directory that you select.  There is a version for VS.NET 2003 and another for VS.NET 2005.

Feel free to download the .ZIP file here.  Simply run the appropriate .REG file by double-clicking it (you must have Administrator privileges) and wallah! (er, voila!) you have your shortcuts.  NOTE: these files will alter your system registry (though just slightly).  No guarantee is implied.  Just because it works great on my machine doesn't imply it will work universally, though I don't not see it working.

Note, both of these rely on their respective environment variables, notably %VS71COMNTOOLS% and %VS80COMNTOOLS%.

Very simple, I know, but effective and without all the other hassle and fluff...just simple, clean shortcuts.

Wednesday, November 23, 2005 4:21:00 AM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [0]  |  Trackback
 Monday, November 21, 2005

One thing that I failed to mention in my previous post with regard to troubleshooting applications in .NET 2.0 in which you have the potential for erroneous cross-thread calls (e.g. calls from a worker thread to a UI thread) is a new diagnostic method that can greatly simplify a developer's life.

You were kind of on your own in the 1.0 and 1.1 days for finding cross-thread method invocations and it was sometimes difficult to manage.  However, new to the world of .NET 2.0 is a static method on the Control class called  CheckForIllegalCrossThreadCalls.  This property happens to default to true in a DEBUG configuration, but it can be manually overridden.  Effectively, when set to true, the debugger will be able to identify if a control's handle is referenced by a non-UI thread and will immediately throw an exception in order to diagnose the problem.

Can save lots of digging, for sure!

Monday, November 21, 2005 9:41:00 PM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [0]  |  Trackback

Historically, creating multithreaded applications had been challenging.  There's a lot to consider - resource contention, performance, cross-thread invokations, and much more.  Granted it is not 'difficult' to create an application that spins off secondary, tertiary, quaternary threads (or background threads for that matter) to accomplish 'work'.  Resource contention, mutexes, semaphores, and locks aside, it's trivial to have code that spins up a thread.  The implications, however, are farther reaching that this seemingless inocuous code.

Thread th = new Thread(new ThreadStart(workerMethod));
th.IsBackground = true;
th.Start();

The stakes are raised, however, when you have to deal with cross-thread invocations.  User Interface (UI) programming is fraught with perils in this regard.  Windows Forms are, for example, affinitized to their thread.  That is, the only thread that can legally call into UI code and perform UI updates must be the thread on which the UI was created.  (NOTE: a non-UI thread update may be successful from time to time.  If you're doing this and are successful, I wouldn't be sitting too comfortably.  At best it can lead to some very tricky bugs to track down.  The UI might even become completely non-responsive).  Usually, this isn't much of an issue.  Unless, of course, you have an application that performs many operations on non-UI threads but you need to update the UI by providing some form of status from the non-UI thread.

Most (perhaps all - I haven't checked) of the Windows Forms controls are thread safe in their events.  Even though a control might be multi-threaded, its events occur on the UI thread, so you're free to update the UI.  If you write your own controls, however, this is something of which you must be cognizant.  To help us in this battle, the .NET framework makes it pretty easy to make cross-thread calls, marshalling a call back onto the proper thread.  It's easy, but if you don't know what you're looking for it might not be the easiest thing to find.

Let's talk about a couple of things.  First, let's look at how to perform cross-thread calls to Windows Forms controls that is compatible with .NET 1.0, 1.1, and 2.0.  Then we'll take a peek at some really cool 2.0 stuff that might be of some assistance.

Since the dawn of the .NET Framework, we've had a special interface called ISynchronizeInvoke.  This interface, implemented by Windows Forms controls (such as the Form), provides a mechanism through which we can marshal a call onto the UI thread.  There is a performance hit associated with such a call, so we really only want to make the call when we really have to.  To that end, the designers of the interface saw fit to include a method called InvokeRequired which returns true if we're calling it from a non-UI thread, false otherwise.

Suppose we have a method named workerMethod() which may be running on a non-UI thread.  Then again, it might not be, so we need to make the determination before updating the UI or raising an event to the caller.  Note, events are not automatically marshalled to the UI thread - the call is made on the thread that raised the event in the first place.  Suppose further that the method is running in-place on a Windows Form for simplicity.

internal sealed class SomeForm : Form {
   private delegate void ReportProgress(string status);
   private ReportProgress _reportProgress;

   internal SomeForm() {
      InitializeComponent();
      _reportProgress = new ReportProgress(showStatus);

      Thread th = new Thread(new ThreadStart(workerMethod));
      th.IsBackground = true;
      th.Start();
   }

   private void workerMethod() {
      // do some work and then report progress
      _reportProgress(“I'm doing something very important.“);
      // do more work
      _reportProgress(“Finished.“);
   }

   private void showStatus(string status) {
      ISynchronizeInvoke sync = this as ISynchronizeInvoke;
      if ( sync.InvokeRequired ) {
         sync.Invoke(_reportProgress, new object[] { status });
         return;
      }
      lblStatus.Text = status;
   }
}

This simple example illustrates a few things.  I took a shortcut (for performance reasons) of having a form-level delegate instance on which all invocations would occur.  This instance (_reportProgress) provides a means to call the showStatus() method (rather than calling it directly) and allowing us to call back to (from itself) without having to create a new instance for each call.

Note that the showStatus() method casts the form to an ISynchronizeInvoke.  It then determines whether or not the call was made on a non-UI thread.  If it was (where InvokeRequired returns true), it then calls the Invoke() method (which transfers control to the UI thread synchronously) passing the parameter.  This go around, the InvokeRequired property would return false and control would fall through to update the label appropriately.

This is a pretty slick strategy that I developed for a game I was writing a couple of years ago in which I had a lot of TCP/IP traffic on a special socket implementation that I had written.  I needed a mechanism to update the UI with the various messages and this is an adaptation of my solution.

Tricky note: An interesting and quite difficult situation to track down when using this technique is that if you are trying to always marshal to a UI thread, but the control to which you are marshalling the call is not yet parented to a container form, InvokeRequired will ALWAYS return false - giving you a sense that you can go ahead an update the UI when in reality you cannot do so in this same manner.  If anyone is interested, I have a great solution for this situation.  With affermative feedback, I'd be happy to post my resolution.

In the .NET 2.0 days, now, Microsoft created a new class which can really help us out here.  (Disclaimer: I have not investigated how it reacts when the recipient control is not yet parented as mentioned in the previous notice)  The new class is called BackgroundWorker.  It resides in the System.ComponentModel namespace.  This class provides a pretty slick mechanism for running an operation on a non-UI thread and have it report updates on the proper UI thread (in a very similar manner to how I created a TCP/IP socket control).  Essentially, you simply need to create the instance, set up the event handlers and invoke it.

internal sealed class SplashScreen : Form {
   internal SplashScreen() {
      InitializeComponent();
   }

   protected override void OnLoad(EventArgs e) {
      base.OnLoad(e);
      this.Show();

      BackgroundWorker worker = new BackgroundWorker();
      worker.DoWork += new DoWorkEventHandler(initializeApplication);
      worker.RunWorkerCompleted += new RunWorkerCompletedEventHandler(doneInitializing);
      worker.ProgressChanged += new ProgressChangedEventHandler(reportInitializationProgress);
      worker.WorkerReportsProgress = true;
      worker.RunWorkerAsync();
   }

   private void reportInitializationProgress(object sender, ProgressChangedEventArgs e) {
      lblStatus.Text = (string)e.UserState;
   }

   private void doneInitializing(object sender, RunWorkerCompletedEventArgs e) {
      // determine if an error was thrown.
      // it would not be a good idea to display the messagebox directly within the worker method
      // as it is not running on a UI thread.

      if ( null != e.Error )
         MessageBox.Show(this,
            “Unable to initialize application.\n\n“ + e.Error.Message,
            “Error Initializing“, MessageBoxButtons.OK, MessageBoxIcon.Information);
      this.Close();
   }

   private void initializeApplication(object sender, DoWorkEventArgs e) {
      BackgroundWorker worker = sender as BackgroundWorker;
      // do something amazing but time consuming
      worker.ReportProgress(0, “Doing some amazing work...“);
      // do something more
      worker.ReportProgress(50, “Continuing with our general amazingness...“);
      // wrap up
      worker.ReportProgress(100, “Done.“);
   }
}

This simple example illustrates the sublime.  With this simple class you can easily inject cross-thread programming into your Windows Forms applications with minimal pain.

Monday, November 21, 2005 3:32:00 PM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [2]  |  Trackback

This is trivial, I know, but it might help someone out there that might have this issue.

I have rebuilt my web server box finally (having lost the domain controller several weeks ago due to a hard drive crash).  I had backed up all of my databases and now needed to restore them onto my newly reinstalled box.  This time, I decided to use my SQL Server 2005 Standard rather than my previous SQL Server 2000 (but that doesn't really have anything to do with this).

I don't like to put my databases in the standard SQL Server location (e.g. DRIVE:\Program Files\Microsoft SQL Server\MSSQL\Data).  Instead, I place them in a completely different file tree mostly for consolidation purposes, not to mention that it makes management easier to not have files scattered all over.

Additionally, as I install SQL Server, I don't like to have the various support services run under LocalSystem or Administrator, but rather create a low-privilege account.

When you go to restore a database .BAK file, you are presented with a directory tree of the server.  You then need to navigate to the location of the .BAK file to select it.  Despite the fact that you might be logged into SQL Server as Administrator or anyone high-level account, it will browse the directory tree under the user credentials that you specify for the MSSQLSERVER service - NOT the account with which you're logged in.  Therefore, you need to grant that user appropriate permissions in the database directories (both the destination folder as well as the backup folder).

Once you do that, you should be all set!

Monday, November 21, 2005 9:06:00 AM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [0]  |  Trackback
 Thursday, November 17, 2005

I had the opportunity last night (technically, though we're in the same span of darkness that has since transitioned to Friday) with a buddy of mine, Phil Burns, to attend the Utah County .NET User Group (UCNUG - Uck-Nug).  I try to make a point to visit each of the groups throughout the extended valley as frequently as possible.  This was, however, my first opportunity to experience the UCNUG first hand.  It was fun.  It looks like they have a good group of (mostly) students that are pretty interested in .NET technologies - which is great.

Scott Golightly, a colleague and friend of mine, was speaking tonight, focusing on SQLCLR.  Unfortunately a few demos went south, but Scott did a great job, as usual, at dissiminating the information to the group, and there were definitely a few nuggets of data which were great.

Thursday, November 17, 2005 5:58:00 PM (Mountain Standard Time, UTC-07:00)  #    Disclaimer  |  Comments [0]  |  Trackback