IAmUnknown

Goodbye windows phone 7

2011-10-13T02:47:00.001+13:00

After having played with a WP7 device for about 8 months, I now have to say it's not for me. There are two reasons.

Firstly my current phone was lent to me, and now has to be returned. But this is not the show stopper. I'd buy a new one tomorrow if it wasn't for the second reason. The main reason I'm giving up on the Windows Phone 7 is that I cannot find a way (that works for me) to back up my telephony data. My text (SMS) messages and calling data are important to me and in order to continue using these devices I need to have a way to record and back up this information.

Because Microsoft has made it incredibly difficult to back up this data via their API's I cannot buy or even write an app that will let me back up this data easily or in fact at all. So finally I gave up trying.

It's a sad day because for all of it's flaws, the Windows Phone 7 is one of the most ergonomic phones I've ever used. It was mostly intuitive and in the main just works. And looking at the future of Windows Phone and Windows 8 - the metro paradigm (the one on the Phone) look like it's the future of Windows.

I'm now looking at either an iPhone (deep sucky in breath) or and Android Phone (slightly less sucky in breath).

:-(

The Suprising Truth about Motivation

2011-02-07T08:47:00.000+13:00

I found this this morning via reddit and I think it sums up what's being playing on my mind for years. What motivates us as people who make machines do 'stuff'.

For reference: original link to the Royal Society of Arts in the UK

SQL Server 2005/8 OUTPUT clause

2010-03-05T10:41:00.002+13:00

This just passed me by. What a great function to avoid writing some triggers. Which we all know aren’t guaranteed to always fire.

Here are a couple of examples :

Hard Disk Beat Boxing

2010-01-29T14:50:00.000+13:00

I’ve just returned from a month away and have discovered that my disk has gone north last week, making the sounds of a rather rubbish DeeJay.

Since I use virtual machines for a lot of the work I’ve been doing and these machines actually run from a separate disk – I‘ve been fairly lucky and have only lost my vs2008 environment – the vs2010 is in a VM. This environment is easily recovered as we have a hard disk images for most of the vs2008 environments, and I only have to reinstall beyond compare, configure notepad++ and some VS settings, restore a couple of test databases and I’m back in business.

However this is another timely reminder for

Using Source Control – I’ve lost no source code, including all my part changes which are in TFS as shelvesets
Use a second disk for VMs – not only is it faster but it can also save your bacon from time to time
Keep hard disk images of important setups of your dev environment. I have to thank Aderant’s great IT team for this. They really are the best IT team I’ve ever worked with.
Technorati Tags: Disk backup
back up important things before going away – and please please please automate this as its trivial on windows to set up scheduled task to do this for you.

Make coding easier ???

2009-09-28T10:47:00.001+13:00

The Resharper 4.5 splash screen during installation made me laugh today. I completely dig Resharper (though I prefer Refactor Pro) so don’t get me wrong. I just thought “Wow, that makes my code easier to understand”.

TRX File decoder script updated to include Console output

2009-08-06T12:09:00.001+12:00

Can’t image how I forgot this, but I did:

<xsl:stylesheet version="2.0"  
                xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
                xmlns:t="http://microsoft.com/schemas/VisualStudio/TeamTest/2006"
                >

<xsl:template match="/">
  <html>
  <head>
        <style type="text/css">
            h2 {color: sienna}
            p {margin-left: 20px}
            .resultsHdrRow { font-face: arial; padding: 5px }
            .resultsRow { font-face: arial; padding: 5px }
            </style>
    </head>
  <body>
    <h2>Test Results</h2>
    <h3>Summary</h3>
        <ul>
            <li>Tests found:    <xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@total"/></li>
            <li>Tests executed: <xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@executed"/></li>
            <li>Tests passed:   <xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@passed"/></li>
            <li>Tests Failed:   <xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@failed"/></li>
            
        </ul>
    <table border="1" width="80%" >
        <tr  class="resultsHdrRow">
          <th align="left">Test</th>
          <th align="left">Outcome</th>
        </tr>
        <xsl:for-each select="/t:TestRun/t:Results/t:UnitTestResult" >
        <tr valign="top" class="resultsRow">
            <td width='30%'><xsl:value-of select="@testName"/></td>
            <td width='70%'>
              <Div>Message: <xsl:value-of select="t:Output/t:ErrorInfo/t:Message"/></Div>
              <br/>
              <Div>Stack: <xsl:value-of select="t:Output/t:ErrorInfo/t:StackTrace"/></Div>
               <br/>
              <Div>Console: <xsl:value-of select="t:Output/t:StdOut"/></Div>
            </td>
        </tr>
        </xsl:for-each>
    </table>
  </body>
  </html>
</xsl:template>

</xsl:stylesheet>

Using Beyond Compare with Visual Studio

2009-08-06T11:21:00.001+12:00

This info is in other places on the internet but I’m putting it here for my future reference:

Under Tools|Options|Source Control|Visual Studio Team Foundation Server

Select Configure User Tools

Then add the following tools:

Compare:

Tool: C:\Program Files (x86)\Beyond Compare 3\BCompare.exe

Arguments: %1 %2 /title1=%6 /title2=%7 /solo

Merge:

Tool: C:\Program Files (x86)\Beyond Compare 3\BCompare.exe

Arguments: %1 %2 %3 %4 /title1=%6 /title2=%7 /title3=%8 /title4=%9 /solo

MS Test Error: Metadata files that belong to projects cannot be edited. If you wish to use this metadata file, please import it into the solution

2009-06-29T11:55:00.001+12:00

Visual studio (2008) sometimes gets into a bind. It loses the binding with a testadatafile (vsmdi) and try as you might you cannot reattach it and you get the message:

“Metadata files that belong to projects cannot be edited. If you wish to use this metadata file, please import it into the solution”.

Well there is one proposed solution on msdn but it didn’t work for me. To get round it I dropped the vsmdi from the solution. Added it back using add existing item in the solution designer and then went to the Test | Create a new Test list

This allowed me to recreate my test list. since I don’t really use many test lists it was a piece of cake but I don’t recommend this solution if you have legacy test lists and are afraid of losing them.

MS Test Error : Test skipped/Test host process exited unexpectedly

2009-06-09T17:37:00.001+12:00

I get this occasionally – esp. when I’m killing the tests part way through a debug session. The trick is to kill the VSPerfMon process that hangs around.

Or the way I do it:


c:\>tasklist | findstr /i VSPERFMON
VSPerfMon.exe    6500 Console                    1     23,192 K
c:\>kill -f 6500

Why Oh Why Oh Why….

2009-05-21T13:50:00.001+12:00

Has MSTest decided it needs me to have a DVD in the drive in order to run my tests.

I swear its got it in for me… Ye gads even if I kill the process it and restart it still refuses to proceed unless I replace the Chef Series 2 DVD I was listening. I suppose it must have like Lenny Henry!

Update: It’s now decided I cannot take “Romeo and Juliet” out of the DVD drive!!!!

Awesome intro regexp!!!

2009-05-12T15:17:00.001+12:00

Found this tutorial on digg today. It’s definitely worth reading.

Scratch That

2009-04-13T00:00:00.010+12:00

After realising that my son is producing his own videos using ms movie maker, hypercam and the awesome pivot stick figure animator and seeing the quality of this work I decided to introduce him to the really cool Scratch tool from MIT.

This is his first attempt at event driven programming.

This is his second

and its a great but simple game. The diver swims from left to right and back again. The left right cursor keys control the cannon (the pointy wizard's hat) and the space bar fires a bullet at the diver. Each time you hit the diver you get one point.

My son is only 10 and the game's official tester only just turned 7 this week. But they've managed to build and have a global release of their first game in the space of one sunday afternoon - amazing what two easter eggs can achieve in terms of productivity. I wonder if Joel Spolsky is listening....

Beer and Bytes inaugural meeting

2009-04-08T13:29:00.002+12:00

Last night was the first meeting of the Auckland North Shore Beer and Bytes group. I consider is a great success not because I wasn't the only person there (I ensured that by dragging a couple of willing cohorts from work - cheers Joe and Steve), but because I had not anticipated how the evening would progress. Would everyone be bored? What would we talk about?

Well as it happened there was some very interesting things discussed. Joe tells me there were tape (actually digital but due to my age they'll always be tape) recorders in action. If anyone has any snippets they'd a like to add please feel free to add a comment.

Mark and Greg from Tracado gave us a run down on all the cool new things in Silverlight 3 that make this the best release yet. I had discussion about NHibernate and magic. Peter recommends this project as a great example of why WPF is great. But I think my favourite of all was Peter (please could you, and everyone else who wants to, send me links to your web pages?) encouraging us to have a .net (real guns and bullets) shooting night!

Actually the other main thing we discovered was how many of us have DBase or Clipper in our backgrounds.... Thanks to Steve, Guy, Patrick, Peter, Allan, Joe, David, Mark, Greg and Bruce!

I think the general consensus was that we should do it again so I'll organise one for a couple weeks after Easter. Next time I'll bring my laptop and maybe we'll look at some code too :-)

MSTest - Specified cast is invalid

2009-03-18T17:35:00.001+13:00

Just remembered this one. When running a VSMDI from the command line (using the mstest command switch /testmetadata) sometimes this error just happens - no help just 'specified cast is invalid'.

I found out a while ago that if you have test lists that are empty this error sometimes happens. What you do is go to the VSMDI editor in visual studio and in the VSMDI in question ensure that you either have tests in each Test List, or remove the test lists that are empty.

One more hassle from MSTest!!!!

Generating SQL from the NHibernate mapping file

2009-03-17T13:58:00.001+13:00

I spent 10 minutes searching for a simple program that would dump out the SQL that NHibernate generates from a mapping file if requested. In the end I just wrote my own. This is the relevant snippet.

                NHibernate.Cfg.Configuration config = new NHibernate.Cfg.Configuration();

                config.Configure();

                config.AddFile("hbm_mapping.xml");

                var export = new NHibernate.Tool.hbm2ddl.SchemaExport(config);

                export.Execute(true, false, false, true);

For future use

2009-03-16T08:38:00.001+13:00

For those idle moments I like to visit http://stackoverflow.com/questions/tagged/command-line - there's always a gem or two there. This morning I learned about quux's Excellent WMIC cheat sheet!

Finally it's out!

2009-03-13T08:59:00.001+13:00

It's been a while coming and taken its toll in terms of long hours and hard work but the ADERANT team have finally released our new software - ADERANT Expert "Golden Gate" Release. I leave you read the press release but I must say it's an exciting release - workflow, DSLs, WCF, WPF, new User Experiences, - you name it - it's in there! Well done everyone!

Awww what a shame ....

2009-03-12T23:46:00.002+13:00

Sorry but I couldn't help but feel a tinge of sadness this.... Only joking!

Taking ownership of a file without explorer

2009-03-03T10:01:00.001+13:00

Use the following command on vista (elevated) command prompt:

icacls full_path_and_filename.ext /setowner name_of_user_or_group

Notepad++ on Vista 64 - replacing Notepad.exe

2009-03-03T09:28:00.001+13:00

There are links here and here (among many other places) that shows how to replace notepad editor with the excellent Notepad++ on a vista or xp machine as the system default text editor. If you use vista 64 you'll need to add another directory to the location where you replace the original notepad.exe

(%SystemRoot%\SysWOW64).

So the set is :

%SystemRoot%

%SystemRoot%\System32

%SystemRoot%\SysWOW64

As always on Vista you'll need to jump through a few hoops to allow you to do this. The first is to replace the file owner with someone other than the default 'TrueInstaller' and secondly to give the new owner full control over the file.

The above links give more info if you need it.

Don't forget - you mess with windows at your own risk :-)

Heaven

2009-02-23T13:31:00.001+13:00

There are other nerds out there!!! Nuff said!

Chainsaw (log4net viewer) won't refresh properly

2009-02-23T13:17:00.001+13:00

We use Apache project's Chainsaw utility to view our log4net based logging messages. Being a Java Swing/AWT app it has a loads of funnies when running on windows such as commands not being where a windows user typically expects them but as a log viewer I find it outstanding - well, once you've got it to work that is!

Anyway recently we had a case where on one and only one machine it just refused to refresh the display properly. Though messages could be read one by one, using the cursor key down method, it was slow and painful, as reading the general flow and history of messages usually helps debugging more than an individual message. We spent ages over the past few weeks looking at different setting - java, log4net, networking/firewalls etc (we use chainsaw over UDP), we even tried running different versions of chainsaw but to no avail.

Finally a flash of inspiration - where are user settings stored? Yes! In your c:\users\blah\.chainsaw directory!!! Just delete this and low and behold like any decent app it recreates the info with sensible defaults.

Sometimes it helps not to be too clever!

NB. I always forget how to format a application configuration url for chainsaw on windows as there is small funny in chainsaw setting it via the UI. The correct format for us on vista64 is:

file:///c|/users/blah/some-dir/somefile.xml

You can find this under the application preferences:

In the Automatic Configuration URL

Stopping command line building from signing assemblies

2009-02-13T16:23:00.001+13:00

we all know that signing assemblies is a good thing, and I'm sure that most people know that the visual studio built in profilers (in instrumentation mode) need to work with unsigned code. So to stop the team build from signing I used the property SignAssembly and set it false before running msbuild (via the environment). I don't know if it works yet as it's running right now but I hope its as simple as that.

Just managed to GET (groan!!!) my first WCF REST Service running from my own host and not IIS

2008-12-20T22:12:00.001+13:00

More info later. Its 10pm on a Saturday night and I'm off home :-)

During a code review

2008-12-11T08:47:00.001+13:00

My code may not be the best in the world, but yesterday I came across the following snippet:

            DrawingImage image = null;
            IDictionaryEnumerator enumerator = dictionary.GetEnumerator();
            if (enumerator != null) {
                enumerator.Reset();

                while (enumerator.MoveNext()) {
                    DictionaryEntry key = (DictionaryEntry)enumerator.Current;
                    image = key.Value as DrawingImage;
                    if (image != null) break;
                }
            }

            return image;

Which after thinking about reduced to this:

            return dictionary.Values.OfType<DrawingImage>().FirstOrDefault();

Interesting? No not very except for the following thought, as we move to a world where we write less and less imperative code and more and more declarative code, some of us (esp. me) really have to give up control to other more talented programmers out there a just use their code 'declaratively'.

Quick Mention For Ex Colleagues New Venture - tracado.com

2008-11-22T20:55:00.001+13:00

A couple of colleagues have just left Aderant to set up their own venture (www.tracado.com) in Auckland, New Zealand. Mark Greene and and Greg Smith are two of the most talented software people I've had the privilege of working with. Between them they have an immense and detailed understanding of building large software systems using the very latest Microsoft technologies. It's been a great pleasure working these guys and I'm sure that they'll make huge success of their new consultancy. So if you want properly architected and built software see these guys!

SQL Server Management Studio Madness - why do I have to have scripts as strings?

2008-08-28T09:14:00.001+12:00

What do you think the following option in the SQL Server Management Studio does?

Now one side of my brain says that I'll get a some preamble on my script looking a little bit like

IF NOT EXISTS (SELECT * FROM sys.views WHERE object_id = OBJECT_ID(N'[dbo].[SomeTable]'))

Well you'd be right! Well in the main.

Without the option:

/****** Object:  Table [dbo].[SomeTable]    Script Date: 08/28/2008 09:03:29 ******/
CREATE TABLE [dbo].[SomeTable](
    [Column1] [nchar](10) NULL,
    [Column2] [nchar](10) NULL
) ON [PRIMARY]

And with the option:

/****** Object:  Table [dbo].[SomeTable]    Script Date: 08/28/2008 09:03:49 ******/
IF NOT EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[SomeTable]') AND type in (N'U'))
BEGIN
CREATE TABLE [dbo].[SomeTable](
    [Column1] [nchar](10) NULL,
    [Column2] [nchar](10) NULL
) ON [PRIMARY]
END

But for stored procedures?

Without the option:

/****** Object:  StoredProcedure [dbo].[DoIt]    Script Date: 08/28/2008 09:10:53 ******/
create procedure [dbo].[DoIt]
as 
    select * 
    from SomeTable
    where Column1 = 'blah'

With the option:

/****** Object:  StoredProcedure [dbo].[DoIt]    Script Date: 08/28/2008 09:10:35 ******/
IF NOT EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[DoIt]') AND type in (N'P', N'PC'))
BEGIN
EXEC dbo.sp_executesql @statement = N'create procedure [dbo].[DoIt]
as 
    select * 
    from SomeTable
    where Column1 = ''blah''
    ' 
END

What I want to know is why does the 'IF NOT EXISITS' check require me to generate a script as a string invocation of sp_executesql?

RDP over two monitors

2008-08-26T20:43:00.001+12:00

I'd not even thought about doing this until today but at work we have dual monitors and we can work via remote desktop but this usually doesn't expand correctly over my home laptop and spare lcd setup. Today I did some hunting around and came up with the /span option on MSTSC (the rdp client). Cannot do full screen but it expand just about perfectly!

Sweet

TFS Command line (un)locking a file

2008-08-26T15:04:00.001+12:00

I locked a file to prevent other users being able to checkin as I was doing radical surgery on it. This is something I hardly ever have to do with TFS but I was justified in this case. Since I use the command line I found that the easiest to issue a

tf lock /lock:checkin somefile.xml

However when unlocking it I discovered no unlock command. Rather you have to

tf lock /lock:none somefile.xml

This is (IMO) a real programmer choosing an API that is sensible for them (lock a file with no lock) but opaque to the new user. As I write lots of little tools/scripts etc, it's made me think about API discoverablility/design for the future.

Playing with Extension Methods

2008-08-21T14:47:00.001+12:00

I've had a issue with option Xml such as in our configuration stores. Basically I wanted to grab data from an XAttribute (see System.Xml.Linq) but wanted to avoid having to constantly check for null nodes. To ease the pain I wrote the following:

public static class Extensions {
    /// <summary>
    /// The extension method allows you to grab data from an XAttribute of if its null, supply a default
    /// </summary>
    /// <typeparam name="T"></typeparam>
    /// <param name="node">The node.</param>
    /// <param name="defaultValue">The default value.</param>
    /// <returns></returns>
    public static T ValueOrDefault<T>(this XAttribute node, T defaultValue) where T : IConvertible {
        if (node == null) {
            return defaultValue;
        }
        return (T)Convert.ChangeType(node.Value, typeof(T));
    }
}

now I can write

var results = from element in doc.Descendants("SomeTag").Where(e => e != null)
       select new {
           Prop1 = element.Attribute("attribute1").ValueOrDefault(String.Empty),
           Prop2 = element.Attribute("attribute1").ValueOrDefault(String.Empty),
       };

instead of

var results = from element in doc.Descendants("SomeTag").Where(e => e != null)
       select new {
           Prop1 = element.Attribute("attribute1") != null ? element.Attribute("attribute1").Value :  String.Empty,
           Prop2 = element.Attribute("attribute2") != null ? element.Attribute("attribute2").Value :  String.Empty,
       };

Or some multi line version of it. Its was just as easy to create one for XElement.

How Did I Get Started In Software Development?

2008-08-13T20:46:00.001+12:00

I just been tagged by my friend Mike Hadlow to join with this meme. The question is a very popular topic with us nerds, and for me especially considering my passion for technology.

How old were you when you first started in programming?

This is a favourite story of mine. I was 11 years old (1978) and in Mr Beardsley's maths class. At the back of the room was a partition, behind which only the 6th formers (16 year and older) kids were allowed. In the fateful particular lesson a series of beeps and other noises were heard. Asking Mr Beardsley about them was the probably the most life changing thing (a part from having kids) ever to happen to me - "That's the 6th formers playing with the computers. Would you like to find out more about that?" Well we started youngsters computer club that week and the rest is history.

What was your first programming language

Sorry I'd like to say machine code but no it was BASIC (PET ISTR then TRS80 then Acorn).

What was the first real program that you wrote

The first program I entered into a computer was from my (then) best friend Mandeep:

10 print "********************* Nuclear Alert ******************************"
20 goto 10

However the first program I actually wrote myself was called fatty, and it was a game to teach my younger brother maths. It had a some simple maths questions like 'what is 15 + 4'. If you selected the right answer a fat man would move along the screen all way across, but if you got it wrong then the man would randomly fall down the screen on his journey across the screen :-) It was called fatty because the man sort of looked, well, fat!

What languages have you used since you started programming

Ok, apart from BASIC and some simple machine code on my home computer I've used Pascal, Modula2, C, C++, Assembler (many varieties), PL/1, PLS, SQL, and of course C# and VB, not to mention PERL, and various shell languages and now iron python. But without a doubt the most interesting language I've ever written in was micro code at university. Here we had to do such things like loading buses with addresses, issuing memory refresh and read instructions, unloading buses etc. Together these instruction primitives would end up forming a single machine code instruction like load register with contents of memory address.

What was your first professional programming gig

I was hired by IBM for a year during my degree to to testing of network apps, but after graduation I started work at NatWest bank in the UK as a systems analyst but left that after 9 months to return to programming at small company in Surrey, UK. Here I learned C then C++ and became firmly entrenched in the MS world with the first visual studio. I have to say working here was was another life changing experience - I learned the difference between hacking code and engineering software systems (which I'll expand in another post).

What's the most fun you've every had programming

For me its all about the paradigms and abstractions, and how they affect cognition. It sounds lofty but really all I mean is that I think civilisation is really just the application of one technology after another that leads us from one way of thinking to another. Software is a great medium as we can create new models, use and break them, learn from them and then start over much more easily than say working with, say, hardware. I have fun with programming nearly every day. But I get the most fun from getting into a tough debug session where you dig deep, learn some, try again , dig deeper and so on, every time building new models and breaking through abstractions to try and understand what was intended and what was achieved.

Who am I calling out?

Peter Goodman and Simon Shearston....

MS Test is hurting

2008-07-22T20:55:00.001+12:00

Normally I try not 'me-too' post but I had to link to this. A friend of mine has just written an interesting post about how difficult he finds MS Test. Mentioning this to a colleague resulted in a link to Jeff Atwood's Coding Horror 'Why Can't Microsoft Ship Open Source Software?' post basically going further and trying to understand the reasons behind this.

I'm actually fed up with littering my tests with [DeploymentItem] attributes, having to hack around virtually opaque XML files, and bending over to assume the god forsaken position that MS Test wants me in. Yes I've tried to get change implemented - but some arguments I lose and some I win!!!

Sorry - rant over.

Access Denied Error when attaching to process

2008-07-01T16:28:00.001+12:00

I've had this problem the that I've been trying to solve over the last few days and its stopped me from debugging into services (remoting and WCF) in other processes. Every time I went into (in VS2K8) Tools|Attach to Process, and I selected a one of our processes, I got an Access Denied error. There was nothing in the event log, so I really didn't know what to try - except perhaps to reinstall VS, which hasn't worked.

The solution is actually very mundane and is related to the fact the I was attempting, not to do managed code debugging, but something called 'workflow debugging'. I'm not sure what this is but VS tells me it incompatible with managed debugging. I suspect that its TFS related and I was basically attempting to attach the wrong kind of debugger, and since I'm no longer a TFS admin, I suspect this is why I'm getting access denied messages.

To fix this all I had to do was to specify the correct debugger in the attach dialog, by clicking on the Select button and and changing the debugger type.

Before:

After:

MSBuild Tasks - DLL or EXE

2008-06-27T11:17:00.001+12:00

Quite often I find that I need the reuse the functionality that is in a custom msbuild task within small tool elsewhere. Usually the way round that issue is to refactor the nuts an bolts of the task into another dll, and then referencing this where needed. This is still the best option in my opinion. However this morning we discovered that you put your tasks assembly in exe wrappers also. This allows you to have a static main etc etc to allow to you call the nuts and bolts of the task from, say, the command line. Thanks to colleague Martin for verifying my ad hoc musings on this subject.

MSTest fails with UTA058: Trust problem

2008-06-25T14:25:00.001+12:00

I kept getting this error for all my unit test dlls even though I was absolutely certain that their locations were 100% fully trusted on my local cdrive. I even checked using the .net config tool in control panel. Finally I had a go at specifying the full paths to the dlls and 'viola' it worked! You'll see that in the previous posting we use /testcontainer:%%~ff . The necessary bit is ~f. To find out more about this you can type 'for /?' in the a command window.

In addition, substitution of FOR variable references has been enhanced.
You can now use the following optional syntax:

    %~I         - expands %I removing any surrounding quotes (")
    %~fI        - expands %I to a fully qualified path name
    %~dI        - expands %I to a drive letter only
    %~pI        - expands %I to a path only
    %~nI        - expands %I to a file name only
    %~xI        - expands %I to a file extension only
    %~sI        - expanded path contains short names only
    %~aI        - expands %I to file attributes of file
    %~tI        - expands %I to date/time of file
    %~zI        - expands %I to size of file
    %~$PATH:I   - searches the directories listed in the PATH
                   environment variable and expands %I to the
                   fully qualified name of the first one found.
                   If the environment variable name is not
                   defined or the file is not found by the
                   search, then this modifier expands to the
                   empty string

The modifiers can be combined to get compound results:

    %~dpI       - expands %I to a drive letter and path only
    %~nxI       - expands %I to a file name and extension only
    %~fsI       - expands %I to a full path name with short names only
    %~dp$PATH:I - searches the directories listed in the PATH
                   environment variable for %I and expands to the
                   drive letter and path of the first one found.
    %~ftzaI     - expands %I to a DIR like output line

MSTest.exe on the command line

2008-06-25T14:17:00.001+12:00

Heres a little batch file to run your tests locally (neatly) from test containers (courtesy of Paul)

rmdir TestResults /S /Q
for %%f in ("..\bin\debug\*UnitTest*.dll") do del "%%~nf".trx &  mstest /testcontainer:%%~ff /noisolation /resultsfile:"%%~nf".trx
for /f "Tokens=*" %%f in ('dir *.trx /b') do @nxslt2.exe  "%%f" trxsummary.xsl

Works for me!!!

Debugger.IsAttached

2008-06-24T09:46:00.001+12:00

Have just discovered System.Diagnostics.Debugger.IsAttached after years of using using #if. This way I can check if my unit tests are being debugged and thus alter timeouts as needed! As they say down here - Sweet As!

Visual Studio 2005 to 2008 Migration - Collating the upgrade logs

2008-06-22T20:42:00.001+12:00

I've been involved in the migration of a fairly large code base from VS2005 to VS2008 this past few weeks. Well kudos to MS for making this a simple process, however one little thing really got on my nerves. I wanted to run the devenv /upgrade xxxxx.sln across the entire code base and not a solution at a time. Next I wanted all the upgrade reports to be amalgamated into one report that I could search easily for errors/issues/warnings etc.

The first bit was solved by my colleague Larry in Tallahassee building a batch file that ran through all the solution files in a DOS for loop and running down across the whole code base looking for solution files and then running the above command over them.

for /f "tokens=*" %s in ('dir *.sln /s/b') do devenv /upgrade "%s"

Next I wanted to collect all the upgrade xml reports (all called upgradelog.xml), but I didn't want to generate a different format to the one that MS creates. The ms report is actually just a xsl transformation that is run on the upgradelog.xml. It takes the contents of log file (an xml file with the root node <Log />, and generates some nice html with some expand/collapse JavaScript.

So I figured my best bet would to extend this template making is work over a series of logs rather than just one. This left the next issue, how to amalgamate all the XMLs. My first choice was hindered by the fact that XSLT to me a write only language. Try as I might I cannot seem to get the syntax to stick :-)

Anyway my next option was to create use python to create a simple XML file that listed all the logs. This was trivial but again fun as I don't know much python either.

dir UpgradeLog.XML /s/b | \Python25\python.exe generatemasterlog.py > upgrade.xml.list

The python is trivial

import sys
print '<?xml version="1.0" encoding="utf-8"?>'
print '<Logs>'
while True: 
    logfile = sys.stdin.readline().strip()
    if not logfile: break        
    print '\t<UpgradeLog report="%s" />' % logfile
print '</Logs>'

this upgrade.xml.list has the form

<?xml version="1.0" encoding="utf-8"?>
<Logs>
 <UpgradeLog report="c:\development\SomeSln\UpgradeLog.XML" />
            .
            .
            .
  <UpgradeLog report="c:\development\AnOtherSln\UpgradeLog.XML" />
</Logs>

The using my old favourite nxslt2.exe I transformed this into a master upgrade log xml file.

<?xml version="1.0" encoding="utf-8" ?>
<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform" xmlns:msxsl='urn:schemas-microsoft-com:xslt'>
    <xsl:output method='xml' indent='yes'/>
    <xsl:template match='/*'>
        <xsl:processing-instruction name="xml-stylesheet">
            <xsl:text>href="UpgradeReport.xslt" type="text/xsl"</xsl:text>
        </xsl:processing-instruction>

        <xsl:element name="Logs">
            <xsl:for-each select='/Logs/UpgradeLog'>
                <xsl:copy>
                    <xsl:copy-of select='document(@report)//UpgradeLog/*'/>
                </xsl:copy>
            </xsl:for-each>
        </xsl:element>
    </xsl:template>
</xsl:stylesheet>

What this does is run through all the report="" attributes in the above upgrade.xml.list and pull the out the contents of that xmlfile and output it in with in xml file within the root node of <Logs />. Hence I now have a master XML file with all the logs in the form

<Logs>
   <UpgradeLog> ....
   </UpgradeLog>
   <UpgradeLog> ....
   </UpgradeLog>
</Logs>

I also output a processing statement (href="UpgradeReport.xslt" type="text/xsl") that allows a browser automatically transform this xml into the nice html view.

The final step was to update the Microsoft XSLT UpgradeReport.xslt to support the new amalgamted xml. Thats too big to put here, but it basically change the master match from Log to Logs (<xsl:template match="Logs">) and then change the existing <body> to apply template (<xsl:apply-templates select="/Logs/UpgradeLog"/>). This meant that is applies to all the <UpgradeLog /> elements rather than just one.

I'll email it to anyone who wants it (preetsangha gmail etc), but remember is mostly Microsoft's work.

Getting round the Visual Studio 2008 MSTest Appbase issue/bug

2008-06-19T15:45:00.001+12:00

The Problem

There is a known bug in visual studio 2008 (and here) Well actually its in how MSTest runs the unit tests. The problem as I see it is that the AppDomain that your unit tests run in is not the same one that is initially created in the VSTestHost.exe. Its this initially created domain cannot find your types/assemblies. This is caused by VSTestHost.exe having an APPBASE of "C:\Program Files (x86)\Microsoft Visual Studio 9.0\Common7\IDE (or the equivalent on 32 bits "C:\Program Files\Microsoft Visual Studio 9.0\Common7\IDE).

Debugging through with the .net framework reference source, led me to the following explanation. When the tests run there is an AppDomain transition from your testing AppDomain. This can be caused by may things, I my case it was a System.Configuration call that was resulting a GetEvidence() call deep in framework. Once this transition had occurred, the assembly resolver delegate chain only had one resolver delegate within

Microsoft.VisualStudio.TestTools.TestTypes.Unit.AssemblyResolver.

This is the one that cannot find the unit test dlls. Our resolver on the other hand was in our AppDomain and was therefore never called.

My simple workaround

I couldn't find a complete workaround, but I have what I think is one that is fairly unobtrusive. However this one relies on a known locatation for the Unit test dlls to reside (which is only really true when you done a complete build).

What you do is create a Vista symbolic link to the required appbase inside the the problematic appbase, and then tell VSTestHost.exe to probe this path as well. This fools the resolver into following this link to your required appbase.

The simplest way to do this use the mklink, on the command line, and then edit the VSTestHost.exe.config in the above directory to include this symlink. For example if my unit test dll was in directory c:\UnitTests. I'd use (mind the wrap)

mklink /D mklink /d "C:\Program Files (x86)\Microsoft Visual Studio 9.0\Common7\IDE\MyUnitTestsDir" C:\UnitTests

and I'd add edit the C:\Program Files (x86)\Microsoft Visual Studio 9.0\Common7\IDE\VSTestHost.exe.config and replace

<configuration>
    <runtime>
        <assemblyBinding xmlns="urn:schemas-microsoft-com:asm.v1">
            <probing privatePath="PrivateAssemblies;PublicAssemblies"/>
        </assemblyBinding>
    </runtime>
</configuration>

with

<configuration>
    <runtime>
        <assemblyBinding xmlns="urn:schemas-microsoft-com:asm.v1">
            <probing privatePath="PrivateAssemblies;PublicAssemblies;MyUnitTestsDir"/>
        </assemblyBinding>
    </runtime>
</configuration>

You can see in both cases the symbolic link is called MyUnitTestsDir.

My better workaround

In the case of running the tests from the command line, or via the IDE I picked a similar method, but actually create and delete the symlink in the [AssemblyInitialize]/[AssemblyCleanup] methods. I created a small helper class (thanks to Creating symlinks in C# by Bart De Smet), that can be used to create and delete the symbolic links.

public class SymLink : IDisposable{

    [System.Runtime.InteropServices.DllImport("kernel32.dll")]
    static extern bool CreateSymbolicLink(string lpSymlinkFileName, string lpTargetFileName, int dwFlags);
    
    
    private int SYMLINK_FLAG_DIRECTORY = 1;
    private bool linkCreated;

    public bool LinkCreated {
        get { return linkCreated; }
    }

    string symbolicLinkPath;

    /// <summary>
    /// Create a symbolic link of the given name to in a given directory to an actual location
    /// </summary>
    /// <param name="linkName">name of the link. </param>
    /// <param name="root">this is the directory within which the symlink should be created</param>
    /// <param name="actual">this is the actual location where the symlink points</param>
    public SymLink(string linkName, string root, string actual) {
        symbolicLinkPath = System.IO.Path.Combine(root, linkName);
        linkCreated = CreateSymbolicLink(symbolicLinkPath, actual, SYMLINK_FLAG_DIRECTORY);
    }

    /// <summary>
    /// Create a symbolic link of the given name to in a given directory to an actual location
    /// </summary>
    /// <param name="linkName">name of the link. </param>
    /// <param name="root">this is the directory within which the symlink should be created</param>
    /// <param name="actual">this is the actual location where the symlink points</param>
    /// <param name="force">Where the link should be deleted before use</param>
    public SymLink(string linkName, string root, string actual, bool force) {
        symbolicLinkPath = System.IO.Path.Combine(root, linkName);
        if (force) {
            Dispose();
        }
        linkCreated = CreateSymbolicLink(symbolicLinkPath, actual, SYMLINK_FLAG_DIRECTORY);
        int getLastError = System.Runtime.InteropServices.Marshal.GetLastWin32Error();
    }


    #region IDisposable Members

    /// <summary>
    /// Dispose
    /// </summary>
    public void Dispose() {
        if (System.IO.Directory.Exists(symbolicLinkPath)) {
            System.IO.Directory.Delete(symbolicLinkPath);
        }
    }

    #endregion
}

I then added the appropriate calls with in the [AssemblyInitialize]/[AssemblyCleanup] methods.

private static SymLink link;

[AssemblyInitialize()]
public static void AssemblyInitialize(TestContext testContext) {

    // I suggest that symLinkName should be a guid or uniqueness without } as this messes up string.format unless you think about it.
    string symLinkName = "SymLink-CF06174B-AA2A-4890-9209-ECCD0F9D172E";
    string oldAppBase = (string)AppDomain.CurrentDomain.GetData("APPBASE");
    string requiredAppBase = (new System.IO.FileInfo(new Uri(System.Reflection.Assembly.GetExecutingAssembly().CodeBase).LocalPath)).DirectoryName;
    link = new SymLink(symLinkName, oldAppBase, requiredAppBase, true);

    Assert.IsTrue(link.LinkCreated, "MS Test directory symlnk creation failed");

}

[AssemblyCleanup()]
public static void AssemblyCleanup() {
    link.Dispose();
}

You still have to add a probe path entry to 'SymLink-CF06174B-AA2A-4890-9209-ECCD0F9D172E' like in the first workaround, but atleast this version will take into account the funny deployment directories

...\TestResults\SomeDymanicName[xxx]\Out

And there you go.

Some thoughts

This is only a hack, as it requires you to edit your C:\Program Files (x86)\Microsoft Visual Studio 9.0\Common7\IDE\Vstesthost.exe.config file with known identifiers. But got me working until we see the real solution from MS.

Decoding the VS2008 Trx file

2008-06-17T12:23:00.001+12:00

I've just updated my script to support 2008 and the of course if you look at the internals you're going to have to upgrade on new release! Here's the new xslt. I've updated it to include the errors too (mind the wrap):

<xsl:stylesheet version="2.0"  
                xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
                xmlns:t="http://microsoft.com/schemas/VisualStudio/TeamTest/2006"
                >
<xsl:output method="text" />
<xsl:template match="/">
    <xsl:text>&#xA;&#x9;</xsl:text><xsl:value-of select="substring(t:TestRun/t:TestRunConfiguration/t:NamingScheme/@baseName, 1, 20)"/>: <xsl:value-of select="t:TestRun/t:TestRunConfiguration/t:Deployment/@runDeploymentRoot"/>.trx
    <xsl:choose>
        <xsl:when test="(t:TestRun/t:ResultSummary/@outcome) = 'Failed'">
        Failed <xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@failed"/> (Executed [<xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@executed"/>], Found [<xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@total"/>], Passed [<xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@passed"/>])
        <xsl:for-each select="/t:TestRun/t:Results/t:UnitTestResult" >
            <xsl:if test="@outcome = 'Failed'" >
                <xsl:text>&#xA;&#x9;&#x9;</xsl:text><xsl:value-of select="@testName"/>
                <xsl:text>&#xA;&#x9;&#x9;&#x9;</xsl:text>
                <xsl:value-of select="substring(t:Output/t:ErrorInfo/t:Message, 1, 170)"/>
                <xsl:text>&#xA;&#x9;&#x9;&#x9;</xsl:text>
                <xsl:value-of select="substring(t:Output/t:ErrorInfo/t:Message, 171, 340)"/>
                <xsl:text>&#xA;</xsl:text>
            </xsl:if>
        </xsl:for-each>
        </xsl:when>
        <xsl:otherwise>
        Passed (Executed [<xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@executed"/>], Found [<xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@total"/>], Passed [<xsl:value-of select="t:TestRun/t:ResultSummary/t:Counters/@passed"/>])
    </xsl:otherwise>
    </xsl:choose>
    
</xsl:template>
</xsl:stylesheet>

Decoding the MS Test TRX file

2008-05-13T11:51:00.001+12:00

I got fed up with loading a command line MS Test output data (.trx file) into visual studio in order to see the the failed tests and their messages so with the help of a colleague we came up with a simple XSL transformation that turns the output of the MS test session into a piece of text or html. We use the text version to print out a summary, and the html version to show the errors (if any :-)

In order to incorporate it into the command line desktop build I used nxslt2 from xmllab.net. Which takes the trx file, the xsl stylesheet and the output on its command line e.g.

nxslt2.exe  testrun.trx trxformatter.xsl -o testrun.results.htm

The summary stylesheet just outputs how many test found, passed and failed, and the name of the trx file if any did fail (so you can interrogate easily in vs if you want).

           1: <xsl:stylesheet version="2.0"  xmlns:xsl="http://www.w3.org/1999/XSL/Transform" >

           2:     <xsl:output method="text" />

           3:     <xsl:template match="/">

           4:         Test Results Summary : Codebase: <xsl:value-of select="Tests/TestRun/tests/value/codeBase"/>

           5:         <xsl:if test="(Tests/TestRun/result/executedTestCount - Tests/TestRun/result/passedTestCount) > 0">

           6:             Summary : Failed <xsl:value-of select="Tests/TestRun/result/executedTestCount - Tests/TestRun/result/passedTestCount"/> (Executed [<xsl:value-of select="Tests/TestRun/result/executedTestCount"/>], Found [<xsl:value-of select="Tests/TestRun/result/totalTestCount"/>], Passed [<xsl:value-of select="Tests/TestRun/result/passedTestCount"/>])

           7:             Results : <xsl:value-of select="Tests/TestRun/runConfig/runDeploymentRoot"/>.trx

           8:         </xsl:if> <xsl:if test="(Tests/TestRun/result/executedTestCount - Tests/TestRun/result/passedTestCount) = 0">

           9:             Summary : Passed (Executed [<xsl:value-of select="Tests/TestRun/result/executedTestCount"/>], Found [<xsl:value-of select="Tests/TestRun/result/totalTestCount"/>], Passed [<xsl:value-of select="Tests/TestRun/result/passedTestCount"/>])

          10:         </xsl:if>

          11:  

          12:     </xsl:template>

          13: </xsl:stylesheet>

and outputs a summary similar to :

The error details stylesheet :

<?xml version="1.0" ?>
<xsl:stylesheet version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform">

    <xsl:template match="/">
        <html>
            <head>
                <style type="text/css">
                    h2 {color: sienna}
                    p {margin-left: 20px}
                    .resultsHdrRow { font-face: arial; padding: 5px }
                    .resultsRow { font-face: arial; padding: 5px }
                </style>
            </head>
            <body>
                <h2>Test Results</h2>
                <xsl:for-each select="Tests/TestRun/result">
                    <h3>Summary</h3>
                    <ul>
                        <li>
                            Tests found: <xsl:value-of select="totalTestCount"/>
                        </li>
                        <li>
                            Tests executed: <xsl:value-of select="executedTestCount"/>
                        </li>
                        <li>
                            Tests passed: <xsl:value-of select="passedTestCount"/>
                        </li>
                    </ul>
                </xsl:for-each>
                <table border="1" width="80%" >
                    <tr  class="resultsHdrRow">
                        <th align="left">Test</th>
                        <th align="left">Outcome</th>
                    </tr>
                    <xsl:for-each select="Tests/UnitTestResult">
                        <tr valign="top" class="resultsRow">
                            <td width='30%'>
                                <xsl:value-of select="testName"/>
                            </td>
                            <td width='70%'>
                                <Div>
                                    Message: <xsl:value-of select="errorInfo/message"/>
                                </Div>
                                <br/>
                                <Div>
                                    Stack: <xsl:value-of select="errorInfo/stackTrace"/>
                                </Div>
                            </td>
                        </tr>
                    </xsl:for-each>
                </table>
            </body>
        </html>
    </xsl:template>

</xsl:stylesheet>

creates the following HTML output

Note that the test name will of course be yours, and any messages and stack trace will appear in the

I'm aware that you can optimise the xsl but it works for me as is.

Stupid Stupid Stupid

2007-11-05T12:50:00.001+13:00

Installing live writer on Vista 64 - hahahahahahhahahahahhahahahhahahahahahhaha.

Save yourself a quizzical 10 minutes Installing Windows Live Suite on Vista x64.

Only one issue the directory on my XP machine was "C:\WINDOWS\system32\config\systemprofile\Local Settings\Application Data\WindowsLiveInstaller\MsiSources"

How hard is it to detect the OS and say - please download this instead - BTW there's no support.

Problems with the formatting

2007-10-31T09:37:00.001+13:00

I'm using windows live writer to do the postings to this blog but I'm finding that the Insert Code plugin is not formatting correctly on the finished blog. I need to get to the bottom of this as my postings are looking rubbish. Bear with me as I try to find out what's going on.

Sorry

Signing a third party assembly

2007-10-31T09:34:00.001+13:00

The other day I had to add a library to our third party tools folder (nmock2 actually) but in order to use it in our code I had to ensure that it had a strong name. Now one way I could have done this is to have downloaded the source code for the assembly and then recompiled adding a strong name. However I found a simpler solution on Adrian Hara's blog that showed how I could use the ILMerge tool from MS Research to achieve the same effect but without the hassle (and there would have been) of recompiling the unsigned module.

Essentially what you do is sign your own assembly and merge in the unsigned assembly, creating a single signed assembly package. This does requires some code compilation but not the original code from the unsigned third party assembly, just some of your own. Being a lazy so an so (:-) I decided to automated the process:

Create temporary assembly (ILMerge.dummy.dll) with a correct strong name. This is my favourite bit as the code I'm using to create this assembly doesn't exist (e.g. echo.>tmp.cs) .
Run ILMerge on this assembly together with the third party one, also specifying the strong name.
Delete the temporary assembly

with a command file and that's it.

Here's a batch file snippet that I used to do all of this:

 1: @echo off

 2: setlocal

 3: echo.>tmp.cs

 4: set thekeyfile="MyStrongNameKeyFile.snk"

 5: csc /out:"ilmerge.dummy.dll" /target:library /keyfile:%thekeyfile% tmp.cs

 6: if exist "%%~nf.XXXXX%%~xf" del "%%~nf.XXXXX%%~xf"

 7: for %%f in (%1) do "ILMerge.exe" "ilmerge.dummy.%%~nf%%~xf" %%f /ndebug /keyfile:%thekeyfile% /out:%%~nf.XXXXX%%~xf

 8: if exist "ilmerge.dummy.dll" del "ilmerge.dummy.dll"

 9: del tmp.cs

 10: endlocal

By the way the XXXXX is an identifier I've used to distinguish the output from the original assembly (e.g. preet.dll becomes a signed preet.XXXXX.dll). To use it just save the code snippet above use like this:

       1:  

       2: C:\temp\thirdparty>dir /b

       3: MyStrongNameKeyFile.snk

       4: NMock2.dll

       5: signasm.cmd

       6:  

       7: C:\temp\thirdparty>signasm.cmd NMoc*

       8: Microsoft (R) Visual C# 2005 Compiler version 8.00.50727.42

       9: for Microsoft (R) Windows (R) 2005 Framework version 2.0.50727

       10: Copyright (C) Microsoft Corporation 2001-2005. All rights reserved.

       11:  

       12:  

       13: C:\temp\thirdparty>dir /b

       14: MyStrongNameKeyFile.snk

       15: NMock2.dll

       16: NMock2.XXXXX.dll

       17: signasm.cmd

       18:  

       19: C:\temp\thirdparty>

SDC TeamBuild Task Clash

2007-10-10T10:08:00.001+13:00

Just ran into this and put a Codeplex discussion posting on the Microsoft Codeplex site for Microsoft SDC Tasks for MsBuild, indicating this fault we got with the latest version.

Essentially the team build has Get Source Code target and this calls the Get task in Team Foundation, unless you use SDC in which case the Get task used is from SDC - and boom!

C:\Program Files\MSBuild\Microsoft\VisualStudio\v8.0\TeamBuild\Microsoft.TeamFoundation.Build.targets(331,11): error MSB4064: The "Workspace" parameter is not supported by the "Get" task. Verify the parameter exists on the task, and it is a settable public instance property.

 C:\Program Files\MSBuild\Microsoft\VisualStudio\v8.0\TeamBuild\Microsoft.TeamFoundation.Build.targets(329,5): error MSB4063: The "Get" task could not be initialized with its input parameters.

Heres the changed lines in Microsoft.Sdc.Common.tasks file:

 1: <!-- Not needed so excluded -->

 2: <!-- <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.Changes" />

 3:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.ChangesInInterval" />

 4:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.CreateBranch" />

 5:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.CreateClientFromTemplate" />

 6:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.CreateChangelist" />

 7:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.DeleteBranch" />

 8:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.GetChangelistFromDateTime" />

 9:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.GetChangelistFromLabel" />

 10:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.GetUsersForChangelists" />

 11:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.Integrate" />

 12:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.LabelSync" />

 13:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.ReverseIntegrate" />

 14:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceDepot.Sync" />

 15:  -->

 16:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.Changes" />

 17:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.Checkin" />

 18:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.Checkout" />

 19:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.CreateBranch" />

 20:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.CreateClientFromTemplate" />

 21:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.DeleteBranch" />

 22:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.GetChangelistFromDateTime" />

 23:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.GetChangelistFromLabel" />

 24:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.GetUsersForChangelists" />

 25:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.Integrate" />

 26:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.LabelSync" />

 27:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceTfs.Sync" />

 28: <!-- Not needed so excluded -->

 29: <!-- <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceSafe.Get" />

 30:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceSafe.LabelGet" />

 31:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceSafe.LabelLatest" />

 32:  <UsingTask AssemblyFile="$(TasksPath)Microsoft.Sdc.Tasks.dll" TaskName="Microsoft.Sdc.Tasks.SourceSafe.Changes" />

 33: -->

Silly little MsBuild gotcha

2007-10-09T08:41:00.001+13:00

I've inherited some custom MsBuild tasks that just worked - until I moved their deployment location from C:\WINDOWS\Microsoft.NET\Framework\v2.0.50727 to another directory (e.g. under program files\msbuild).

Once moved they started giving out the message

MSB4062: The "Blah.BlahTask" task could not be loaded from the assembly C:\Program Files\MSBuild\Blah\Blah.BlahTasks.dll or one of its dependencies. The system cannot find the file specified. Confirm that the <UsingTask> declaration is correct, and that the assembly and all its dependencies are available.

Thinking I'd missed a hidden assembly I searched until I was blue in the face. Moving the assembly back made it work again when I got message MSB4036 saying that the task assembly could be found, but that now my new tasks (in another new assembly) were not implementing ITask, or not public.

This was really interesting as I'd written the new tasks from the ref docs and they were indeed public and implemented ITask. So I pulled up the source to the legacy tasks and low and behold they were private classes derived from Task and not explicitly labeled as implemented ITask. Quick recompile sorted not only this problem but also (I figured) the location issue and yep it did.

So the question is why did my new task give a message about other tasks? Maybe it was a glitch as I only saw it once.

TFS Cloaking as a way of avoiding the "get all the stuff I don't need for my build"

2007-10-03T09:02:00.001+13:00

The source tree has things that you don’t always want. So use the TFS command line program tf with the workfold command to cloak the stuff you don’t want. This way a “get specific” or “get latest” will ignore those directories

e.g.

to ignore one directory

tf workfold /server:tfs-server /workspace:"my workspace name" /cloak $/Project1/Prototypes/Directory1

to ignore the directories in the in the prototypes directory:

for /f %f in ('dir C:\Development\Project1\Prototypes /ad/b') do tf workfold /server:tfs-server /workspace:"my workspace name" /cloak $/Project1/Prototypes/%f

Notes:

You can only seem to cloak a directory not individual file

You cannot decloak a directory that has an ancestor cloaked. So you have to use.

if you’re an UI guy then you use the TFS Manage Workspaces to achieve the same effect. However it does take longer as you have navigate the entire TFS project hierarchy for each cloak. Which is mucho boring after the second time!

Little TFS Team Build Customisation

2007-09-11T21:57:00.001+12:00

A colleague (James) asked if I knew of a way to stop TFS builds for creating a work item if the build in question fails. The answer took a few minutes to dig up but in a nutshell the answer is that you have to create a CreateWorkItem target (see update at bottom for better answer) in your MSBuild file.

Ok why is that?

MSBuild scripts , like NAnt, are defined using a set of targets. Each target (defined with <Target /> tag) and is an outcome which MSBuild must assert as true for the build to succeed (or conversely not assert as true for the build to fail). This is not meant as a tutorial on MSBuild, as there's hundred's out there, but it does help to know the basics of what MSBuild is doing before you start customising the build.

If you think of the build process as of essentially asserting one target (e.g. <Target Name='MakeItSoTarget' />). This in itself may not do very much, but what a target can do it assert other targets, so MakeItSo may assert that it will be true if PowerEngines and SetCourse also assert true.

In this way a hierarchy of dependencies can be created to define the entire build process from start to finish. In fact VS2005 .csproj files are essentially MSBuild build files that Visual Studio passes to the MSBuild to create your lovingly crafted code masterpiece. But I think a real strength of MSBuild is the pre-existing targets (and thus build processes) that MS has defined, which can be used to create a standard build process such as for Visual Studio or say TFS.

Anyway back to the question, why do I need to override CreateWorkItem. Well for team build there is a targets file which MS uses to control the build process. This does number of things including, preparing for a build, getting the source code from TFS, compiling, testing, and packaging (I'll put a more detailed description of this in another post) . You can see the structure yourself in the file but essentially the main target of concern is OnBuildBreak. This is called (surprisingly) when a build breaks (i.e. some other target is not asserted).

   1:       <PropertyGroup>

   2:                <OnBuildBreakDependsOn>

   3:                       BeforeOnBuildBreak;

   4:                       GetChangeSetsOnBuildBreak;

   5:                       DropBuild;

   6:                       CreateWorkItem;

   7:                       AfterOnBuildBreak;

   8:                </OnBuildBreakDependsOn>

   9:         </PropertyGroup>

  10:         <Target Name="OnBuildBreak"

  11:                       Condition=" '$(IsDesktopBuild)'!='true' "

  12:                       DependsOnTargets="$(OnBuildBreakDependsOn)" />

Here we can see that the OnBuildBreak asserts true if the targets defined in property group OnBuildBreakDependsOn also assert (line 12). This means that we have a number of options for customisation. We can

Change OnBuildBreak to not depend on this precise property group
Alter the property group
Create a custom version CreateWorkItem

Since MS recommends that we don't actually alter the supplied .targets files, we can do each of these in our build file as any targets we define with the same name as the MS ones will be used by MSBuild instead. The path of least resistance suggests that we should endeavour to keep true the MSBuild as possible as this leads to a smaller ongoing maintenance cost (thanks for that Howard!) so to me the simplest thing is to define a new target:

<Target Name='CreateWorkItem' >
     <Message Text='Not creating a work item !!!.' />
</Target>

Now a new work item will not be created!

UPDATE: looking again I see an even simpler way - define the property SkipWorkItemCreation=true in your script, then team build will not do the task. I hope the exercise in understanding was still of value.

Setting up slick run shortcuts quickly

2007-08-27T09:06:00.000+12:00

I use slickrun for lots of things, but what I find it most useful for is opening up visual studio with the correct solutions. The reason its so useful is that our project is based on about 80+ vs solution files, and numerous more project files, carefully organised in to appropriate directories and subdirectories. After getting bored setting up 10 odd shortcuts in slickrun (even using drag and drop) I decided that code would be a quicker solution. Anyway 5 minutes in notepad++ (qrs.cs) and using c# from command line I had:

using System;
namespace qrs {
public class p {
[STAThread] public static void Main(string [] args) {
Console.WriteLine("[{0}]", args[1] );
Console.WriteLine("Filename=\"{0}\"", args[0] );
Console.WriteLine("Path=\"\"");
Console.WriteLine("Params=\"\"");
Console.WriteLine("Notes=\"\"");
Console.WriteLine("GUID={{{0}}}", Guid.NewGuid().ToString());
Console.WriteLine("StartMode=5");
Console.WriteLine("UseRunAs=0");
Console.WriteLine("");
}
}
}

(yes thats a rubbish name :-)

This only requires you to run the exe with the first parameter as the sln file, and the second parameter as the shorcut name. To generate the list I did a simple command line "for".

for /f "tokens=*" %f in ('dir c:\project_root\*.sln /s/b') do @echo qrs.exe "%f" "%~nf" >> myqrs.txt

This generated a load of entries in a file called myqrs.txt ,that I was able to cut and paste into the slickrun ini file.

[Solution2]
Filename="C:\project_root\Subdir\Solution2.sln"
Path=""
Params=""
Notes=""
GUID={29647410-3995-45bf-9e46-cd8540f94c35}
StartMode=5
UseRunAs=0
[Solution1]
Filename="C:\project_root\Solution1.sln"
Path=""
Params=""
Notes=""
GUID={443ba4c0-339a-45d7-bae4-97a2d7bc511f}
StartMode=5
UseRunAs=0

Finding the .net framework path programattically

2007-08-23T10:17:00.000+12:00

This seemed to be an easy question to answer when my colleague Jamie first asked. But it took a while to track it down.

System.Runtime.InteropServices.RuntimeEnvironment.GetRuntimeDirectory()

Thanks to Stephen Toub

Decided to move my technical musings here

2006-12-30T11:34:00.000+13:00

After having kept a few technical postings on another site (http://preetsangha.spaces.live.com/blog) I'm now convinced that this is a better place - so I've finally created a diary here. Hope its not too boring.

Apologies to ISerializable - Roy Osherove's Blog and IUnknown John Lam on software for not being as orignal as them in the name chosen - hope you don't mind.