navigation
 Tuesday, July 31, 2007

The other day I was trying to figure out why the ItemCommand associated with an ImageButton in a DataGrid had stopped firing and was surprised by why...

Here is the grid in action:

When I clicked on any of the ImageButtons in the grid, the page did post back, but the ItemCommand didn't fire. I experimented with binding the grid at different points in the page lifecycle, but that didn't help. I also tried switching from a DataGrid to a GridView, thinking such a drastic change might get rid of whatever oddity was going on. No such luck.

Finally, I noticed that the grid attribute EnableViewState was false. Originally, this grid had contained some template items with editable content, including text boxes, drop down lists and checkboxes, and ViewState was turned on. But I replaced that functionality with some client side logic using JavaScript and JSON web service calls, so I figured no viewstate was needed any more, and had turned it off. It wasn't until a few weeks later when someone clicked on these image buttons that the problem showed up.

So, long story short, I changed EnableViewState to true, and voila, it started to work again. It beats me why viewstate is needed for a command event to fire when an image is clicked (I could see the image.x and image.y variables in the postback even when it was turned off), somehow the event bubbling from the image click to the containing DataGrid's ItemCommand handler (RowCommand in the GridView) is dependant on viewstate being turned on. If anyone knows why I am all ears!

posted on July 31, 2007  #    by John Waters  Comments [0] Trackback
 Monday, July 30, 2007
 | 
posted on July 30, 2007  #    by Adam Anderson  Comments [0] Trackback
 Sunday, July 29, 2007

Back at TechEd 2007 in Orlando, Florida, both teams celebrated the release of the Telerik radControls Training courseware authored by Falafel Software and the start of a great relationship.  Here at Falafel we have nothing but respect and admiration for Telerik's products and their quality.  At the same time, we all here enjoy the Bulgarian spirit, sense of humor and great team.

The picture below was taken at the Texas De Brazil steakhouse with both teams after 14 bottles of Wine, and a lot of other weird drinks, for some reason no one remembers this picture on both teams Beer

posted on July 29, 2007  #    by Lino Tadros  Comments [0] Trackback
 Tuesday, July 24, 2007

While having some fun with the Visual Studio SDK I ran across a great example of a Navigation window by Gaston Milano.  It's a great sample to start with if you would like to dive into Visual Studio Package development.  One item of note that I found while using the navigation window on a very large project is that Visual Studio 2005 kept crashing everytime the window would be brought up.  I noticed that before VS crashed it had a difficult time drawing things. I immediately thought of a GDI leak.  I opened up Task Manager and checked the GDI objects and sure enough every time I openned the window the GDI object count would skyrocket upwards, never to return to a sane GDI count again.  VS crashed after reaching a count of about 10,000 GDI objects and I don't blame it for doing that either :) 

The offending routine, the getter for the HierarchyItem::Icon property in HierarchyItem.cs, calls a method NativeMethods.ImageList_GetIcon() which merely vectors off into the win32 version of ImageList_GetIcon().  Looking at the MSDN documentation for this routine you'll notice that the remarks at the end state

"It is the responsibility of the calling application to destroy the icon returned from this function using the DestroyIcon function"

Looking at the code you'll notice that DestroyIcon is never called on the temporary HICON returned from this call so every time this routine is called I noted 2 HICONs being leaked.  And on my large project, this really started adding up.

The fix?

Change the following code in HierarchyItem.cs:

      int handleIcon = NativeMethods.ImageList_GetIcon((int)imageList, (int)index, hbmMask);
      try {
          this.icon = (Icon)Icon.FromHandle((IntPtr)handleIcon).Clone();
      }
      catch (ArgumentException) {
      }
To:
int handleIcon = NativeMethods.ImageList_GetIcon((int)imageList, (int)index, hbmMask);
try {
     this.icon = (Icon)Icon.FromHandle((IntPtr)handleIcon).Clone();
     NativeMethods.DestroyIcon(handleIcon); // <== added to destroy the temp HICON
    } catch (ArgumentException) {
    }

And add a reference to DestroyIcon in the NativeMethods class like so:

        [DllImport("USER32")]
        public static extern int DestroyIcon(int HICON);
Hope that helps any of you running into similar problems.
posted on July 24, 2007  #    by Adam Markowitz  Comments [0] Trackback

So you have a cool asp.net application and everything is running great. You're running it under NET 2.0 and you're using a report server on the same machine. It may run great for months, and then out of nowhere your web application starts displaying the “Service unavailable” page. It does this until you do an iisreset. Where did this message come from? I recently had to find out for myself. My search lead me to the Event Viewer on the server and five warnings about various system failures that read something like this:

A process serving application pool 'ASP.NET V2.0' suffered a fatal communication error with the World Wide Web Publishing Service.

After the fifth warning I would get the following error:

Application pool 'ASP.NET V2.0' is being automatically disabled due to a series of failures in the process(es) serving that application pool.

After the iisreset, things would go OK, but eventually the above warning/error pattern would repeat. What was going on? Why five warnings, the single error, and then the “Service unavailable” page? Well, when you search for "Application pool 'ASP.NET V2.0' is being automatically disabled due to a series of failures", you get plenty of hits. There are plenty of ways to get these errors. What I ended up finding out was that my report server and my web application were using the same application pool. I would not think this would be a big deal, until I found this tab in the properties of the application pool within the IIS Manager.

Mystery solved. It was doing exactly what it was supposed to do. The Enable rapid-fail protection was enabled and I had gotten five failures within 5 minutes to cause the application pool to be disabled. Let's say your web application calls for a report from your reports server. If your reports server fails for some reason, you can also get a failure with your web application. Since they are both in the same application pool this is two strikes with one failure. A couple more of these and your application pool gets shut down. Lesson Learned. It's good practice to move the web application and reports servers into their own application pools.

posted on July 24, 2007  #    by Bary Nusz  Comments [0] Trackback
 Monday, July 23, 2007

I am working on a web application, and some of the client side logic uses JavaScript to get an integer value from a text box. A user logged a bug claiming that when he entered a value that had a leading 0, for instance 0556, the result was in his words "a random number".

I looked up the parseInt global function here, and lo and behold! parseInt interprets a leading 0 as denoting that an octal number follows! It also has special processing for 0x (hexadecimal). To prevent this interpretation, you can specify the radix yourself in an overload of parseInt, for instance:

parseInt( tb.value, 10 );

For base (radix) 10. 

That's all I have to say about that!

posted on July 23, 2007  #    by John Waters  Comments [0] Trackback
 Thursday, July 19, 2007

While setting up a CruiseControl.NET server for a project here at Falafel I ran into this error:

svn: Working copy '.' locked
svn: run 'svn cleanup' to remove locks (type 'svn help cleanup' for details)

After executing svn cleanup from the command line numerous times to no avail I finally realized the problem was a missing <workingDirectory> node for the project in the ccnet.config file. Well, I won't be making that mistake again.

posted on July 19, 2007  #    by Steve Trefethen  Comments [0] Trackback
 Tuesday, July 17, 2007

There is a pattern to these errors.  Same asp.net source, same database, same IIS version, same Visual Studio version, same components version, but still some chunk of JavaScript is absent without leave. The actual error messages tend to be new each time, but at least the pattern is recognizable. The reason is that the missing JavaScript functions live in webresource.axd, webresource.axd is handled by an aspnet_isapi.dll, and chances are you are missing a mapping for "axd" in your IIS configuration for your web site. To fix:

  • In the IIS Microsoft Management Console snap-in, right-click the virtual directory and select Properties.
  • On the Virtual Directory tab click the Configuration button. 
  • On the Mappings tab of the Application Configuration dialog click the Add button. This brings you to the Add/Edit Application Extension Mapping dialog. Enter for the Executable the full path to aspnet_isapi.dll located in .NET framework directory.  In this case the path was C:\WINDOWS\Microsoft.NET\Framework\v2.0.50727\aspnet_isapi.dll. Set the Extension to be ".axd".  Don't forget the dot or the dialog will not let you proceed. Uncheck the "Check that file exists" checkbox.
  • If the OK button doesn't become enabled, there's a weirdity here (in the tech support business we used to call these "issues", in the real world we call it a "bug").  Click the Executable textbox entry and the OK button should become enabled.

Cycle the website and try again.  Different flavors of this problem occur periodically, so tag this page for later and save yourself some pain.

posted on July 17, 2007  #    by Noel Rice  Comments [0] Trackback
 Sunday, July 15, 2007

A while back I was trying to figure out how to make F1 bring up a custom help window in our web application ActiveFocus. It is easy enough to trap a keypress and open a window showing the help, but to my frustration, after popping up the new window, the built in Internet Explorer help window popped up too!

I tried various variations of cancelling the kepress event, but nothing seems to work. This is for instance how I cancel Ctrl+F (which normally pops up IEs built in Find dialog, but I wanted it to show a custom search dialog instead) :

<script type="text/javascript">

function onKeyDownH(e)
{
  e = window.event;

  var ctrl = (e.ctrlKey) ? true : false;
  if (ctrl == true )
  {
    if ( e.keyCode == 70)
    {
      e.returnValue = false;
      e.keyCode = 0;
      ShowSearchPage();
    }
  }
}

function onloadH(e)
{
  document.onkeydown = onKeyDownH;
  return true;
}

window.onload = onloadH;

</script>

This cancelling approach by the way does work for other keypresses, just not F1, which has some kind of special internal handing. After much head scratching I finally stumbled across this simple solution:

<body onhelp="ShowHelp(); return false;">

Here, ShowHelp does the actual showing of the help window.

Simple, huh? If anyone knows how to make this cross browser compatible I am all ears!

posted on July 15, 2007  #    by John Waters  Comments [0] Trackback