Redirecting Assembly Bindings in .NET 3.5

I spent some time last week digging into assembly binding redirection.  This blog post provides working examples of assembly redirection through config files and publisher policies.

This example will use the following command line utilities and versions:

• csc - C# compiler (Version 3.5.30729.4926)
• sn - .Strong Name Utility (Version 3.5.30729.1)
• gacutil - Global Assembly Cache Tool (Version 3.5.30729.1)
• al - assembly linker (Version 8.00.50727.42)

Helpful utilities not covered in this article:

• fuslogvw - Assembly Binding Log Viewer for viewing the CLR binding process.
• Assembly Cache Viewer - For viewing assemblies installed in the GAC.

To get started lets create a program calling a method in another assembly which outputs the callee's assembly version.  These steps also create assemblies for use in the later examples.

1. Open a Visual Studio Command Prompt and navigate to a working directory where some files can be created.
2. Create and save a file named demo.cs containing the following text:
   
   using System;
   using System.Reflection;
   
   
   [assembly: AssemblyVersionAttribute("2.0.0.0")]
   
   
   public static class Demo
   {
       public static void Main()
       {
           Console.WriteLine("Version " + Assembly.GetExecutingAssembly().GetName().Version.ToString());
       }
   }
3. from a command prompt type "sn -k key.snk"
4. type "csc /target:library /keyfile:key.snk demo.cs"
5. type "echo f | xcopy demo.dll .\bin\v2\demo.dll"
6. Change the AssemblyVersionAttribute in demo.cs replacing "2.0" with "3.0"
7. csc /target:library /keyfile:key.snk demo.cs
8. echo f | xcopy demo.dll .\bin\v3\demo.dll
9. Change the AssemblyVersionAttribute in demo.cs replacing "3.0" with "4.0"
10. csc /target:library /keyfile:key.snk demo.cs
11. echo f | xcopy demo.dll .\bin\v4\demo.dll
12. Change the AssemblyVersionAttribute in demo.cs replacing "4.0" with "1.0"
13. csc /target:libarary /keyfile:key.snk demo.cs
1. echo f | xcopy demo.dll .\bin\v1\demo.dll
14. Create and save a file named demo.cs containing the following text:
    
    public static class Program
    {
        public static void Main()
        {
            Demo.Main();
        }
    }
15. csc /reference:demo.dll program.cs
16. program.exe
17. The output should read: Version 1.0.0.0

Redirecting 1.0 to 2.0 version in the GAC.

1. gacutil /i .\bin\v2\demo.dll
2. sn -T demo.dll
3. Note the Public key token as we'll need that in the assemblyIdentity publicKeyToken attribute in the next step.
4. Create program.exe.config with the following text:
   
   <?xml version="1.0"?>
   <configuration>
    <runtime>
     <assemblyBinding xmlns="urn:schemas-microsoft-com:asm.v1">
      <dependentAssembly>
       <assemblyIdentity name="demo" publicKeyToken="keyFromSn-T" />
       <bindingRedirect oldVersion="1.0.0.0" newVersion="2.0.0.0" />
      </dependentAssembly>
     </assemblyBinding>
    </runtime>
   </configuration>
5. program.exe
6. The output should read:  Version 2.0.0.0

Redirecting 1.0 to Version 3.0 using a URL:

1. Replace the program.exe.config text with the following:
   
   <?xml version="1.0"?>
   <configuration>
    <runtime>
     <assemblyBinding xmlns="urn:schemas-microsoft-com:asm.v1">
      <dependentAssembly>
        <assemblyIdentity name="demo" publicKeyToken="keyFromSn-T" />
        <bindingRedirect oldVersion="1.0.0.0" newVersion="3.0.0.0"/>
           <!--to use an absolute filepath use href="file://D:/Projects/...." or it can point to a web resource: href="http://www.fake.com/demo.dll"-->
        <codeBase version="3.0.0.0" href="./bin/v3/demo.dll" />
       </dependentAssembly>
     </assemblyBinding>
    </runtime>
   </configuration>
2. program.exe
3. The output should read: Version 3.0.0.0

Redirecting 1.0 to Version 4.0 in the GAC using a publisher policy.

1. Create a policy.config file in the working directory with this text:
   
   <?xml version="1.0"?>
   <configuration>
    <runtime>
     <assemblyBinding xmlns="urn:schemas-microsoft-com:asm.v1">
      <dependentAssembly>
       <assemblyIdentity name="demo" publicKeyToken="keyFromSn-T" />
       <bindingRedirect oldVersion="1.0.0.0" newVersion="4.0.0.0" />
      </dependentAssembly>
     </assemblyBinding>
    </runtime>
   </configuration>
2. al /link:policy.config /out:policy.1.0.demo.dll /keyfile:key.snk
3. gacutil /i policy.1.0.demo.dll
4. gacutil /i .\bin\v4\demo.dll
5. delete the program.exe.config in the working directory or erase the bindingredirect element.  The bindingredirects seem to be applied first and will break this example.
6. program.exe
7. The output should read: Version 4.0.0.0

A new life with Microsoft Dynamics CRM...

Friday January 15th was my last day at Urban Science where I learned a lot about software development and made a lot of great friends.

On Monday January 18th I started a new job at Sonoma Partners.  We specialize in customizing and implementing Microsoft Customer Relationship Management software for our customers.

If you are a technologist, programmer, personal contact, or professional contact in the Chicago area give me a call and lets get together.

If you are a potential Microsoft CRM customer, give me a call and lets make a deal ;)

NDepend

Patrick Smacchia kindly gave me the opportunity to use and evaluate his NDepend product.  NDepend is a static analysis tool meaning it can provides insight and guidance about one's work by inspecting assemblies and source code.  It identifies things like classes which may be better as structures by looking at the size of an instance of the class, the number of children, and the depth of inheritance.  It also graphically displays assembly dependencies and can do comparisons of an application between releases.  All in all a very cool tool!


First I started up NDepend to run it over a WPF desktop application I worked on in 2008.  I started by creating a new project.  It was very slick the new project open a .sln file to determine the assemblies to analyze.  NDepend understands there are application assemblies (assemblies compiled from the user's code) and tier assemblies (generally compiled references).  This picture shows what I mean.  After taking just 15 seconds to analyze 353,899 IL instructions NDepend showed an interactive view displaying many different views of the codebase at once which were all linked together, so if I drilled into something in one view or hovered over it there would be an effect in the other windows.  Being a noob this was a bit much for me and I opted to view a static HTML report.  The report didn't look so great in Chrome or FireFox but looked fine in IE8.  I read through the 50 page report awash in information.  I wish I had used the tool in 2008 to focus some of our refactoring efforts.  There are a lot of fantastic queries out of the box showing things that are blatantly wrong with the code we wrote.  Speaking of queries, NDepend has a very cool Code Query Language (CQL) one can use to write queries which throw errors or warnings during a build process.  NDepend comes with many queries out of the box which you can run or use as examples.  


Here is a query provided in the NDepend install showing types that could potentially be marked sealed:
// <Name>Class with no descendant should be sealed if possible</Name>
WARN IF Count > 0 IN SELECT TOP 10 TYPES WHERE
  IsClass AND
  NbChildren ==0 AND
  !IsSealed AND
  !IsStatic


Here is another query provided in the NDepend install showing classes that might be better as structures:

// <Name>Classes that are candidate to be turned into Structures</Name>
WARN IF Count > 0 IN SELECT TOP 10 TYPES WHERE 
  IsClass AND
  SizeOfInst <= 16 AND    // Structure instance must not be too big, else it degrades performance.
  NbChildren == 0 AND     // Must not have children
  DepthOfInheritance == 1 // Must derive directly from System.Object
  // AND IsSealed    : You might want to add this condition to restreint the set.
  // AND IsImmutable : Structures should be immutable type.
  // AND !IsPublic   : You might want to add this condition if you are developping a framework
  //                   with classes that are intended to be sub-classed by your clients.

These are two examples of things I would like to know but wouldn't like to take the effort to figure out.  NDepend makes time consuming code analysis tasks simple and shareable by providing the Code Query Language.

A couple months later I decided to use NDepend to analyze the ASP.NET solution I worked on in 2009.  I precompiled the application as recommended in the NDepend FAQ and browsed to that output instead of using the ASP.NET solution file to begin my analysis.  Unfortunately this project's code quality wasn't as high as the other project and there were 491,298 IL instructions for NDepend to analyze, although it only took 13 seconds. :)  Again I mostly skipped the UI and dove into the 492 page report.  Being a web app, some of the assemblies have funky names like App_Web_c55buezi which makes some items in the report harder to consume than the human friendly names from the desktop project.  Unfortunately I will not be working on this codebase anymore and did not have significant ownership of it so I didn't diligently read up all the goodness NDepend was giving me access to.

My lesson learned is that I need to run this fabulous tool over code I'm actively working on so I can take advantage of it and learn like I've learned from FxCop.  This tool is great because it has something for everyone.  From rules with specific recommendations on what to fix to strategic insights into your dependency structure.

Ensuring lambdas don't leak when passed to EventHandlers

I'm writing a wrapper around a class which outputs results of commands executed into a stream buffer.  I wanted to capture the command text and return that to the caller of my wrapper.  I wanted to use a lambda and clean up my delegate after the command returned.  This contrived code sample shows how to do this improperly and properly:

static void Main()

{

    ExecuteCommand();

    Debug.Assert(CommandExecuted == null);

    ExecuteCommandLeaks();

    //This assert will fail

    Debug.Assert(CommandExecuted == null);

}

static void ExecuteCommandLeaks()

{

    CommandExecuted += new EventHandler<EventArgs>((o, e) => Console.WriteLine("Event called"));

    CommandExecuted.Invoke(new object(), new EventArgs());

}

static void ExecuteCommand()

{

    var commandCapture = new EventHandler<EventArgs>((o, e) => Console.WriteLine("Event called"));

    CommandExecuted += commandCapture;

    CommandExecuted.Invoke(new object(), new EventArgs());

    CommandExecuted -= commandCapture;

}

static EventHandler<EventArgs> CommandExecuted;

SQL Server Management Studio Generating Change Scripts

For those of you who need to lookup SQL syntax for things like alter table, this will let SQL Server Management Studio (SSMS) do the work for you giving you a baseline script to modify or check right into source control.

Generating an Alter Table Statement:

1. In SSMS use the object explorer to navigate to a table.
2. Right click and select design.
3. Make your changes.
4. From the menu select Table Designer --> "Generate Change Script..."
5. Copy and paste the script text or click the yes button to save it to a text file.

Adding a Foreign Key:

1. In SSMS use the object explorer to navigate to a table.
2. Expand the table node.
3. Right click on Keys  and select "New Foreign Key..."  Notice this brings up the table designer if it isn't already open.
4. Fill out the new FK and click Close
5. You should now be in the table designer.  Follow the steps from above.

Delete a Blogger Comment

To delete a blogger comment, like spam, the easiest way I have found to do it is by

1. Opening the post I want to delete in Chrome (the technique should work in any browser, but Chrome is my favorite because it is much faster than IE and because I like the Google brand better than FireFox)
2. Right Clicking on the page and selecting "View page source"
3. Press ctrl+f to bring up the search box and enter "delete".
4. Look for the delete link.  It was the second search result on the page for me and looked like this: <a href='http://www.blogger.com/delete-comment.g?blogID=5094545033872573936&postID=6517665474722511356' title='Delete Comment'>
5. Click the link.
6. Follow the directions on the next page confirming the deletion.

There is probably a way to make the delete button show up on the page but I don't know how.

TEDxDetroit recap

My cousin, Gabriel Mott, turned me on to TED in 2005 by linking to what is still my favorite TED talk. I was excited to be accepted as an attendee of the inaugural TEDxDetroit event on 10/21/2009 at Lawrence Tech University.

The most beneficial part for me was speaking with other attendees such as Alex Fisher, Henry Balanon, Bret Kopf, Betsy Weber and meeting new people such as Lisamarie Babik, David Benjamin, Andrew Brown, Jamie Favreau, Lee Andros (bus buddy!), Janak Mehta, Rich Sheridan, and Dug Song.

Here are my take aways:

• Ask high quality questions.  High quality questions provoke thought.  An example could be "How can we generate y revenue through x technology?",  "How can social media help solve world hunger?"
• What can we do now given the current technology and environment?  If our plans depend on a change (new technology, etc) what if that change doesn't manifest?  If our plans don't consider the current environment (ie. culture) how will the implementation go?
• There are two ways to make money, by nickel and diming the competition or by taking big leaps over them.  Giants take big leaps.  Who are your giants?
• Users are not stupid. Designs are stupid.  Observer your users in their environment and reduce their pain.
• Some people are economic hurricanes.  Everyone has great ideas.  Those who execute on those ideas are the hurricanes.
• Research the problem before providing the solution.  When we have ideas we think of our solution and present that.  Instead of presenting that, present the problem inspiring your solution and see what others have to say.
• At a certain level of wealth problems are problems of perception.  This was a particularly interesting talk for me because I tend to think in functional terms instead of presentation.
• All value is relative.
• Persuasion is better than mandates.
• Contextual timely and immediate.

P.S. If you would like to attend a regional TED event, TEDxLansing is seeking volunteers.

Using gmail from C# and inspecting HttpRequests

This post can help you with two things

1. Sending mail via gmail or google apps through C#.
2. Inspecting the contents of an HttpRequest as a string similar to HttpRequest.SaveAs.

Today while working on a code for fun project at lunchtime we wanted to see the HttpContext.Request.  Initially we were calling HttpRequest.SaveAs(context.Server.MapPath("Request.txt"),true) and life was good.  When we deployed to our test server hosted by Verio, an awesome company, we got an access denied exception when trying to write.  Our corporate firewall prevented us from changing the folder permissions, and we couldn't figure out how to do it through our ftp client, FileZilla.  We decided to email the request to ourselves with the following code:

SmtpClient mailer = new SmtpClient("smtp.gmail.com", 587);

//mailer.EnableSsl = true;

mailer.Credentials = new NetworkCredential("emailAddress@yourDomain.com", "YourPassword");

mailer.Send("sender@sender.com", "recipient@recipient.com", "snazzySubject", Request.RequestAsString());

Running that code we get: System.Net.Sockets.SocketException: No connection could be made because the target machine actively refused it 216.239.59.109:587.  This is our firewall blocking us again.

At home I got "The SMTP server requires a secure connection or the client was not authenticated. The server response was: 5.7.0 Must issue a STARTTLS command first."

Uncommenting the mailer.EnableSsl line above and rerunning fixed the problem and my email arrived!

//End point #1 and begin point #2
Here is the code duplicating the request.saveAs functionality:  RequestAsString is an extension method.  In my code I put it in the System.Web namespace so it shows up for me when using System.Web.

/// <summary>

/// Turns an HttpRequest into a string for inspection.  Similar to request.SaveAs.

/// </summary>

/// <param name="request">The request to turn to a string.</param>

/// <returns>A string to inspect the HttpRequest.</returns>

public static string RequestAsString(this HttpRequest request)

{

    StringBuilder returnValue = new StringBuilder()

        .Append(request.HttpMethod + " " + request.RawUrl);

    if (returnValue.ToString().EndsWith("?"))

    {

        returnValue.Length = returnValue.Length - 1;

    }

    returnValue.Append(" " + request.ServerVariables["SERVER_PROTOCOL"])

        .Append(Environment.NewLine)

        .Append(TurnNamedValuesToLists(request.Headers, ": ", Environment.NewLine))

        .Append(Environment.NewLine)

        .Append(request.ContentEncoding.GetString(request.BinaryRead(request.ContentLength)));

    return returnValue.ToString();

}

/// <summary>

/// Splits out a NameValueCollection.

/// </summary>

/// <param name="target">The collection to be split.</param>

/// <param name="seperator">Seperates the key from the value in the output.</param>

/// <param name="terminator">Put at the end of the key and the value.</param>

/// <returns>A string of all the key value pairs.</returns>

private static string TurnNamedValuesToLists(NameValueCollection target, string seperator, string terminator)

{

    StringBuilder returnValue = new StringBuilder();

    string[] keys = target.AllKeys;

    foreach (var key in keys)

    {

        returnValue.Append(key + seperator + (string)target[key] + terminator);

    }

    return returnValue.ToString();

}

Useful Visual Studio 2008 Settings

Here are some Visual Studio 2008 settings I have turned on to speed up my development time.

1. Default Browser - I use Chrome as my default browser, but our application is targeting IE.  To have Visual Studio launch IE instead of chrome I found out one can right click a .aspx page select 'Browse ' With' and then from the next dialog set Internet Explorer as default.
2. Autorecover - Sometimes over bad VPN connections AutoRecover will freeze the VS User Interface.  I used to change the autorecover path but now I go to settings --> Environment --> Autorecover and uncheck the save checkbox.
3. Source Control - Settings --> Source Control --> Environment  I check get everything when a solution or project is opened.  VS still prompts for get latest on open so you are in control, but it is a nice reminder to sync up.  For saving I have 'Prompt for check out' selected to avoid silent checkouts.  Some files like .vsmdi and some dependant files need to be checked out prior to editing for this to work.  For Editing I have 'Do nothing' so I can edit and continue without checking out files.
4. Outlining - Settings -->Text Editor --> C# --> Advanced I uncheck 'Enter outlining mode when files open.'  This prevents regions from hiding code from me when I open the file.  I also uncheck 'Surround generated code with #region' because I don't like regions.  Ever.
5. Debugging - Settings --> Debugging --> General I uncheck 'Require source files to match the original version'  This is another one for edit and continue debugging.  Sometimes I'll notice something I want to change, like changing a string literal to reference a constant from a helper class, but don't want to stop my debugging.
6. Symbols - Settings --> Debugging --> Symbols I cache my symbols to c:\debugsymbols so they can be shared between VS and Windbg.

Catching inlining in the act from managed code.

This blog post shows ilining in action all without leaving the comfort of managed code.

When you compile C# code it is compiled to Intermediate Language (IL).  When the Common Language Runtime executes IL it uses a Just In Time (JIT) compiler to compile the IL to machine instructions.
Here is some C# sample code:

using System;

using System.Diagnostics;

static class Program

{

    static void Main() { Person me = new Person("Dave"); Console.ReadKey(); }

}

class Person

{

    public Person(string name) { Name = name; }

    private void SetName(string name)

    {

        Console.WriteLine(new StackTrace().GetFrame(1).GetMethod().Name);

    }

    public string Name

    {

        set { SetName(value); }

    }

}

Copy that code into notepad and save it as Program.cs.

From a Visual Studio command prompt type these lines.

csc /out:d.exe Program.cs /optimize- /debug+ 

csc /out:u.exe Program.cs /optimize- /debug- 

csc /out:o.exe Program.cs /optimize+ /debug- 

Then execute the programs:

d outputs set_None

u outputs .ctor

o outputs Main

Kinda crazy huh?

PS. if you run this program from Visual Studio it will output set_None because you run with a debugger attached and the JIT Compiler won't inline.  If you compile in Release mode and run without the debugger (ctrl+F5) you can see Main if you are fast enough ;)

Lets all fight spam

I sent myself an email last night to forward out to the team.  A while after getting to work I wondered where was my email?  It had been put in my spam folder!  Email is more critical to me than postal mail and I would love for it to be reliable.  I've had problems this year where important emails where put into my spam folder.

Sender Policy Framework (SPF) exists to help reduce spam and while I'm not sure if this will reduce the odds of emails from my domain being marked as spam I figured it won't hurt.

This blog post will describe how I setup SPF for my website which uses Google Apps for email and web content and GoDaddy's Total DNS product for Domain Name System configuration.  If you use the same infrastructure it will be easy copy and pasting.  If not you might have to do some additional reading, or switch your infrastructure :)

I ran across this page with SPF Tools including a wizard to generate an SPF record and some record testers.

Using their wizard I got this string: "v=spf1 mx ~all" spf1 means spf version 1, mx means accept mail from mail servers, and ~all means softfail all mail. So basically this would softfail all mail not sent from google's mail servers. This made sense. Being a cautious type I decided to see what Google Apps Admin Help has to say on the matter. Hmmmmmmmm they recommend "v=spf1 include:aspmx.googlemail.com ~all".  This seems to do a DNS lookup at aspmx.googlemail.com which redirects over to spf.google.com which has a list of acceptable IP addresses.  I decided to go with Google's recommended method.  I also decided to change the ~ which is a softfail to a - which is a fail.  Using the tilde seems to be a precautionary measure which reduces the usefulness of the spec.  It is interesting to note that the SPF Tool wizard generates SPF records with a ~ and so does GoDaddy's generator.......  But oh well I'll live on the wild side :)


To implement my record I login to GoDaddy's Total DNS Control and click add new text record.  For the TXT name I put "davidsilvasmith.com." and for the txt value I put "v=spf1 include:aspmx.googlemail.com -all" and a Time To Live (TTL) of 1 hour.

Then I validate my results by emailing check-auth@verifier.port25.com.

Here are my results before implementing the changes:
  SPF check:          neutral
  DomainKeys check:   neutral
  DKIM check:         neutral
  Sender-ID check:    neutral
  SpamAssassin check: ham

Here are the results after:
  SPF check:          pass
  DomainKeys check:   neutral
  DKIM check:         neutral
  Sender-ID check:    pass
  SpamAssassin check: ham

Then I emailed my work account and got my email sent to the spam folder still.  Rats.

Search all columns of all tables for a pattern

Today at work someone wanted to know how to search all columns of all tables for a pattern.  The code below uses a recursive select to generate dynamic SQL which will search all varchar columns for the specified pattern.  Not all columns are able to be cast to varchar.
declare @searchString varchar(max)
set @searchString = '%dealer sites%' --the pattern to search for


declare @queryToRun varchar(max) --internal variable.

--this variable needs to be non null otherwise it will make the recursive select return null.

set @queryToRun = ''

--create a statement to search all columns of all tables for the text.

SELECT @queryToRun = @queryToRun + 'select '''+t.table_name +'.'+c.column_name+''' from ' + QUOTENAME(t.TABLE_SCHEMA) + '.' + QUOTENAME(t.TABLE_NAME) + ' where ' + QUOTENAME(c.column_name) + ' like '''+ @searchString + ''' union '

                  FROM INFORMATION_SCHEMA.TABLES t JOIN

                  INFORMATION_SCHEMA.COLUMNS c on t.table_name = c.table_name

                  WHERE             TABLE_TYPE = 'BASE TABLE' and data_type='varchar'

--trim off the last union statement.

select @queryToRun = left(@queryToRun, datalength(@queryToRun) - datalength('union '))

--run the query.

exec (@queryToRun)

T4 - Text Template Transformation Toolkit

At the Lansing Day of .NET I saw a presentation about Microsoft's T4 (Text Templating Transformation Toolkit).  At the time I couldn't think of where I could use this tool to be more DRY because dry folks go over big at parties :)  Some of the examples given were in generating documentation and in generating a data layer.  There are tools specifically for those tasks so that didn't sit well with me.


Our current product is a web product and we sometimes use the session to pass data between our pages.  It was a lot of redundant typing and I figured T4 could help.


The template takes a list of strings and makes a static class exposing the strings as constants.  These strings are used as keys to pull data out of the session.  Some of the keys need post processing when the data is pulled out.  In this implementation I'm doing that via reflection but after writing it I figure a dictionary could do the trick and be faster.


Put this text in a .tt file and save it.  It will generate a .cs file shown below:




<#@ template language="C#v3.5" debug="true" #>
<#@ output extension="cs" #>

namespace tt
{
    /// <summary>
    /// Used to store keys to lookup session data.  This should ensure we don't have collisions and can easily refactor.
    /// </summary>
    /// <remarks>
    /// This class is autogenerated by SessionKeys.tt.  Do not modify this file!
    /// </remarks>
    public static class SessionKeys
    {
<# foreach ( string constant in GetConstants())
{
string[] inputs = constant.Split(',');

for (int i = 1; i < inputs.Length; i++)
                {
string attribute = inputs[i].Trim();
if(attribute.Length > 0)
{
WriteLine("[" + inputs[i].Trim() + "]");
}
                }

WriteLine("public const string " + inputs[0].ToUpper() + " = \"" + inputs[0] + "\";");
}
#>

    }
}

<#+
///<summary>
/// provides constants for SessionKeys.cs.  Some constants need attributes to be processed.
/// This is done via Attributes in the namespace UrbanScience.Si2.Module.LeadReporting.Utilities.
/// The format of the strings in this list should be IDENTIFIER, Attribute, Attribute, Attribute....
///</summary>
private string[] GetConstants()
{

return new string[]
{
"Name,InputAllowsWildCard",
"StartDate",
"EndDate",
"StateList, CleanMultipleInput",
};
}
#>

Here is the generated class:

namespace tt

{

    /// <summary>

    /// Used to store keys to lookup session data.  This should ensure we don't have collisions and can easily refactor.

    /// </summary>

    /// <remarks>

    /// This class is autogenerated by SessionKeys.tt.  Do not modify this file!

    /// </remarks>

    public static class SessionKeys

    {

            [InputAllowsWildCard]

public const string NAME = "Name";

public const string STARTDATE = "StartDate";

public const string ENDDATE = "EndDate";

[CleanMultipleInput]

public const string STATELIST = "StateList";

           

    }

}

Debugging win32 API errors

This blog post documents the methodology I used to diagnose a win32 API exception with a generic error message.  If you are in the same situation hopefully you'll find this helpful..

In April Tess Ferrandez blogged about Visualizing virtual memory usage and GC Heap.  I thought it would be cool if her code were modified to call Windbg programmatically and use a WPF frontend.  I found out Windbg uses dbgeng.dll!  Awesome!  I'll just write a managed wrapper around that.  While researching writing the wrapper I found CodePlex already had such a project called: mdbglib.  Great!  So I grabbed the code compiled and......Unhandled Exception!  No....... "First-chance exception at 0x7c9666c6 (ntdll.dll) in ASDumpAnalyzer.exe: 0xC0000139: Entry Point Not Found.".

My call stack looked like this: (bolded entries will be referenced later)

  ntdll.dll!_RtlRaiseStatus@4()  + 0x26 bytes
  ntdll.dll!_LdrpSnapThunk@32()  + 0x2a2b2 bytes
  ntdll.dll!_LdrpSnapIAT@16()  + 0xd9 bytes
  ntdll.dll!_LdrpHandleOneOldFormatImportDescriptor@16()  + 0x7a bytes
  ntdll.dll!_LdrpHandleOldFormatImportDescriptors@16()  + 0x2e bytes
  ntdll.dll!_LdrpWalkImportDescriptor@8()  + 0x11d bytes
  ntdll.dll!_LdrpLoadDll@24()  - 0x26c bytes
  ntdll.dll!_LdrLoadDll@16()  + 0x110 bytes
  kernel32.dll!_LoadLibraryExW@12()  + 0xc8 bytes
  mscorjit.dll!Compiler::impImportBlockCode()  + 0x5661 bytes
  mscorjit.dll!Compiler::impImportBlock()  + 0x59 bytes
  mscorjit.dll!Compiler::impImport()  + 0xb2 bytes
  mscorjit.dll!Compiler::fgImport()  + 0x20 bytes
  mscorjit.dll!Compiler::compCompile()  + 0xc bytes
  mscorjit.dll!Compiler::compCompile()  + 0x270 bytes
  mscorjit.dll!jitNativeCode()  + 0xa0 bytes
  mscorjit.dll!CILJit::compileMethod()  + 0x25 bytes
> ASDumpAnalyzer.exe!ASDumpAnalyzer.MainForm.MainForm_Load(object sender = {ASDumpAnalyzer.MainForm}, System.EventArgs e = {System.EventArgs}) Line 66 + 0x8 bytes C#

Based on the call stack I figured the code was throwing during JIT.  Looking at the method being jitted I was able to determine the Debuggee class was causing the Exception.  I figured the project was written on Windows Vista and was using a function that didn't exist in XP.  Looking around the CLI / C++ project nothing jumped out at me.  I decided to buy a C++ / CLI book, read about the win32 API calling convention: stdcall, and read about getting the parameters to LoadLibraryExW.  I also found out Visual Studio has  memory windows and a registers window.  I also found out that the ntdll.dll function names with a p in them are private, so I wouldn't be able to get their signature from MSDN like I could for Kernel32 LoadLibraryExW.

Armed with this knowledge I dug back in.

I put a breakpoint (bp) at {,,ntdll.dll}_LdrpSnapThunk@32.  This ended up getting hit all the time, so I put a hit count bp on {,,kernel32}_LoadLibraryExW@12 with a count of 59 because that is right before the app explodes.  I disabled my SnapThunk bp and enabled it after the LoadLibrary bp was hit.  Looking at the 32 bits on the stack using offsets from the esp register (esp, esp+4, esp+8........). and resolving the four 8 bit values addresses to memory locations I didn't see a function name like I was hoping.

So I started debugging assembly line by line.  After a few minutes I see 

7C917EE5  push        edi  

7C917EE6  push        eax  

7C917EE7  push        dword ptr [ebp+8] 

7C917EEA  push        dword ptr [esi+18h] 

7C917EED  push        ebx  

7C917EEE  call        _LdrpNameToOrdinal@20 (7C917EFDh)

I get very excited because there is a function name in esp when 7C917EEE executes!  This loop gets called a lot so I decide to put a bp on 7C917EEE and print the function name when it is called instead of breaking by putting this: {*(char**)esp}.  I hit F5 and watch function names fill up my output.  Right about the time I get my exception message I see DebugConnectWide in output.  Searching through the mdbglib project I see Debuggee.cpp line 56 has this call: Tools::CheckHR(DebugConnectWide(pwRemoteOptions, __uuidof(DbgClient), (void **)&debugClient ))  Comment that line rebuild, and no exception!

Recursive Selects

Last week at work I created some SQL Server functions to take a list of rows and turn them into a comma delimited string.  It feels like data formatting which should be done on the client side, but that was the method we went with.

It does show recursive selects though, which can also be used for dynamic cross tab queries.  At first this method didn't make any sense to me, but thinking about a select statement as a loop makes it make sense.

declare @siblings table (id int, firstName varchar(max))

insert into @siblings

      select 0, 'Jordan' UNION

      select 1, 'Mia'

declare @siblingString varchar(max)

set @siblingString = ''

select @siblingString = @siblingString + firstName + ', '

      from @siblings

if (datalength(@siblingString) > 2)

      select @siblingString = left(@siblingString, datalength(@siblingString) -2)

select @siblingString

Outputs 'Jordan, Mia'

Lansing Day of .NET 2009 Review

I had a great time at the Lansing Day of .NET Saturday.  I really enjoyed seeing everyone there, and we got fancy cool LDODN 2k9 beer tumblers.  There was a great after party at Jeff McWherter's that Eric Vogel helped me get to.  My favorite sessions were T4: Code Generation with Visual Studio 2008 by Steve Andrews and Improving our Craft: A Discussion on Software Estimation by Mike Eaton.

The T4 Code Generation facilities allow you to leverage the power of C# or VB.NET to write code generation scripts.  It also seemed straightforward to extend the code generators and also code generation hosts. The code generation kicks off by default when you save the file in Visual Studio, but one could write a custom host to execute it from a command line environment or during the build process..  The session showed some interesting scenarios for generating help documentation and a data layer using T4.  I don't plan on using T4 anytime soon, but I'm glad to know the tool exists and have had an overview of it's capabilities.  I've seen Steve Andrews present before and I like his style.  He mixes in keyboard shortcut tips and shows Visual Studio configuration options during the talks I've seen.

Mike Eaton's presentation on Software Estimation initiated some interesting discussion topics and thoughts to ponder.  He is an engaging speaker and had some funny cartoons.  Mike was especially adept at involving the audience in the presentation.

Foundations of C++ /CLI, Book Review

I just finished reading Foundations of C++/CLI: The Visual C++ Language for .NET 3.5 .  I read this book hoping to get a grasp of how to wrap native libraries with C++/CLI.  Unfortunately I didn't read the Amazon Editorial review and only chapter 13 covered this topic.  The book assumes the reader knows C++ and wants to learn .NET.  This made the majority of the book review for me.

This book would probably be good for someone with a C++ background wanting to learn .NET, although if one isn't going to do interop the book recommends using C# instead.