Just dealt with another Visual Studio 2008 “feature”.

You can specify all the necessary “Project Dependencies” in Visual Studio, but will get “CSC : error CS0006: Metadata file FooBar.dll could not be found“. Even if your csproj files have correct references to other solution projects, msbuild will fail.
Maybe it appears only if project output path is outside of project directory.

It appears that Visual Studio keeps the dependencies in two ways, only one of which is read by MSBuild. I see that because I still can specify dependencies in GUI, copy solution to other machine and build it with VS in correct order.

Not with MSBuild.

The data needed by MSBuild is a “ProjectSection(ProjectDependencies) = postProject” section of SLN file. Like this:

Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "make_versioninfo", "make_versioninfo.vcproj", "{F0E0541E-F17D-430B-97C4-93ADF0DD284E}"
EndProject
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "pythoncore", "pythoncore.vcproj", "{CF7AC3D1-E2DF-41D2-BEA6-1E2556CDEA26}"
	ProjectSection(ProjectDependencies) = postProject
		{F0E0541E-F17D-430B-97C4-93ADF0DD284E} = {F0E0541E-F17D-430B-97C4-93ADF0DD284E}
	EndProjectSection
EndProject

Note where to take the GUID of a referenced project.
If you create this section by hand, your project will build.

Not too much fun, if you have over 20 projects, where you can have up to 190 project dependencies… Have fun.

I just won a bet for $10 by solving a quiz:

Check the correctness of a braces, brackets and parentheses sequence. The solution should be of linear complexity (to say more, it’s 1-pass).

For instance, these expressions are correct: “()”, “()()[]“, “([][][]{})”, “([])()[]“, and these are not: “][", "())(", "(()".

It's basically solved in 15 lines, all others being auxiliary. Can you reproduce the algorithm?

The solution follows.

import java.security.InvalidParameterException;
import java.util.Stack;

public class Braces
{
    private static boolean isClosing(char c)
    {
        return c == ')' || c == '}' || c == ']';
    }

    private static boolean isOpening(char c)
    {
        return c == '(' || c == '{' || c == '[';
    }

    private static char matching(char c)
    {
        switch (c)
        {
        case '(': return ')';
        case '[': return ']';
        case '{': return '}';
        case ')': return '(';
        case ']': return '[';
        case '}': return '{';
        default:
            throw new InvalidParameterException("Not a brace");
        }
    }

    public static boolean isCorrect(String expression)
    {
        Stack<Character> lastBrace = new Stack<Character>();

        for (int i=0; i<expression.length(); i++)
        {
            char c = expression.charAt(i);

            if (isClosing(c))
            {
                if (lastBrace.size() == 0) return false;
                if (lastBrace.peek().charValue() != matching(c)) return false;
                lastBrace.pop();
            }
            else if (isOpening(c))
            {
                lastBrace.push(Character.valueOf(c));
            }
            else
            {
                throw new InvalidParameterException(expression + "[" + i + "]=" + c);
            }
        }

        return lastBrace.size() == 0;
    }
}

import junit.framework.TestCase;

public class BracesTest extends TestCase {

    public void testIsCorrect()
    {
        assertTrue(Braces.isCorrect("()"));
        assertTrue(Braces.isCorrect("[]"));
        assertTrue(Braces.isCorrect("{}"));
        assertTrue(Braces.isCorrect("([])"));
        assertTrue(Braces.isCorrect("()()[]"));

        assertTrue(Braces.isCorrect("([][][]{})"));
        assertTrue(Braces.isCorrect("(()[()]())"));
        assertTrue(Braces.isCorrect("([])()[]"));

        assertFalse(Braces.isCorrect("([]"));
        assertFalse(Braces.isCorrect("[]]"));

        assertFalse(Braces.isCorrect("]["));
        assertFalse(Braces.isCorrect("(][)"));
        assertFalse(Braces.isCorrect("(][)"));

        assertFalse(Braces.isCorrect("(][)"));
        assertFalse(Braces.isCorrect("[(){}}"));

        assertFalse(Braces.isCorrect("]["));
        assertFalse(Braces.isCorrect("(("));

        assertFalse(Braces.isCorrect("(()))"));

        assertFalse(Braces.isCorrect("{}{}}{"));
    }
}

Whenever you implement own datasource for .NET GUI binding, you’lll have the choice - whether to implement IList or IListSource.

IListSource is a simplistic interface with two members: IList<T> GetList() and bool ContainsListCollection;

MSDN help about IListSource.ContainsListCollection states it’s “indicating whether the collection is a collection of IList objects”.
MSDN is not true here.

If IListSource.ContainsListCollection is false, GetList() just returns your IList.

If IListSource.ContainsListCollection is true, GetList() is expected to return ITypedList, which needs to provide a collection of PropertyDescriptor-s for every field of your collection.
Its purpose is to provide field names (including by-name field access) in runtime.

Quoted below is a code piece from a Microsoft newsgroup that resolves both cases to a data list (field values list, in second case).

And yes, if you’re planning to mutate the UI-bound collection, use IBindingList<> instead of IList<>.

static IEnumerable GetResolvedDataSource(object dataSource, string dataMember)
{
    if (dataSource != null)

    {
        IListSource source1 = dataSource as IListSource;
        if (source1 != null)

        {
            IList list1 = source1.GetList();
            if (!source1.ContainsListCollection)

            {
                return list1;
            }
            if ((list1 != null) && (list1 is ITypedList))

            {
                ITypedList list2 = (ITypedList) list1;
                PropertyDescriptorCollection collection1 = list2.GetItemProperties(new PropertyDescriptor[0]);

                if ((collection1 == null) || (collection1.Count == 0))

                {
                    throw new HttpException(HttpRuntime.FormatResourceString("ListSource_Without_DataMembers"));

                }
                PropertyDescriptor descriptor1 = null;
                if ((dataMember == null) || (dataMember.Length == 0))

                {
                    descriptor1 = collection1[0];
                }
                else

                {
                    descriptor1 = collection1.Find(dataMember, true);
                }

                if (descriptor1 != null)
                {
                    object obj1 = list1[0];

                    object obj2 = descriptor1.GetValue(obj1);
                    if ((obj2 != null) && (obj2 is IEnumerable))

                    {
                        return (IEnumerable) obj2;
                    }
                }

                throw new HttpException(HttpRuntime.FormatResourceString("ListSource_Missing_DataMember", dataMember));

            }
        }
        if (dataSource is IEnumerable)
        {
            return (IEnumerable) dataSource;

        }
    }
    return null;
}

A friend asked me for it, so probably someone might need a jumpstart in unit tests.

Here’s a short list of short (though very deep, if not best) intros.

• Except for JUnit cookbook (~2 pages) (which has a lot of “ports”, like CppUnit cookbook), I’d name:
• “Endo-Testing: Unit Testing with Mock Objects” at mockobjects.com (Tim Mackinnon, Steve Freeman, Philip Craig) (7 pages);
• “Mocks aren’t stubs” by Martin Fowler (about 7 pages);

For deeper dive, take a “xUnit patterns” book (this one is longer).

For Kent Beck’s book on TDD, shame on me - I never read it. Though after “XP explained” I don’t strive for reading his books.

// If you’re interested in XP and “XP explained” criticism, take a look at “Extreme Programming Considered Harmful for Reliable Software Development” by Gerald Keefer.

My not-so-humble evaluation of VS tests distinctions from NUnit is (in points on -10 to 10 scale).

• (-4) To run tests, VS launches an entire process which takes no less then 3 seconds to start. If you’re in “F5 hit breakpoint - Shift-F5″ loop, that’s disgusting. NUnit tests can be run in-process by Resharper or TestDriver.NET;
• (-2) GUI that shows where test failed is ugly. You don’t get to failed line on double-click, you first get to test log page. Status and call stack are necessary, but I’d like not to trash my document tabs, I already have enough of them open. You can’t make that kind of windows docked/floating;
• (-1) You got a “public TestContext TestContext;” sticking in your code. You’ll need it only when it comes to testing on data, which doesn’t always happen, even in a data-driven application;
• (-1) It’s a “not invented here” technology, while NUnit was around for years;

Edit: In a moment of madness, I mixed together coverage profiling and unit testing tools. May God and readers forgive my aberration.

Edit2: If you need a half-page kickstart in unit testing techniques, the next post can help.

Lua recently became very popular. I encountered Lua scripting in several applications including popular games, I believe it’s used in even more then Wikipedia says.

It’s extremely simple:

• 6 data types (that’s counting “nul” and “function”),
• no C++-style OOP out-of-box (but you can program one, he-he),
• 400K of pure-C code

but is loaded with in-fashion features like

• first-class functions,
• closures,
• coroutines…
• again, Wikipedia is the best.

The only metaprogramming tutorial (with a ready code to implement virtual methods) is “Programming in Lua” book. Still, I don’t see a code to implement ad-hoc polymorphism… Maybe it’s a reason for another post

Redistributed book removed: it violated copyright.

Did you ever think, what happens if you take a member pointer to a virtual function of a class, and then call it on another instance of derived class (that overrides that member)?

What if you do the same for nonvirtual function?

Here’s a test:

#include <iostream>

class A
{
  public:
  virtual void f() { std::cout << "A::f()\n"; };
  void g() { std::cout << "A::g()\n"; };
};

class B : public A
{
  public:
  virtual void f() { std::cout << "B::f()\n"; };
  void g() { std::cout << "B::g()\n"; };
};

typedef void (A::*VirtualFunctionPointer) ();

int main()
{
  A a;
  B b;
  B* pb = &b;

  VirtualFunctionPointer p = &A::f;
  (pb->*p) ();

  p = &A::g;
  (pb->*p) ();
  return 0;
}

Answer is in comments.

One of my friends once encountered a C++ compilation issue.
A program is given:

#include <iostream>
#include <sstream>
using namespace std;

#define SFORMAT(e) ((dynamic_cast<const ostringstream&>(ostringstream() << e)).str())

int main(int argc, char* argv[])
{
cout << SFORMAT("2 x " << " 2 = " << 2*2);
return 0;
}

No one asks you what it will print, unless you’re a god.

Question is: why is the first const char* printed as a pointer, instead of what we need?
Two years ago, I couldn’t tell.
Some guru from Apple answered this in a newsgroup.

Here, take another, simpler C++ quiz about member pointers.

A couple of notes how NOT to use MSBuild.

1. Don’t generate files with AssemblyInfo task. It screws comments, screws InternalsVisibleTo attributes, and whatever else it doesn’t account for.
   Instead: Use
   <FileUpdate Files="@(AssemblyInfoFiles)"
Regex="AssemblyVersion\(".*?"\)\]"
ReplacementText="AssemblyVersion("$(Major).$(Minor).$(Build).$(Revision)")]"
/>
<FileUpdate Files="@(AssemblyInfoFiles)"
Regex="AssemblyFileVersion\(".*?"\)\]"
ReplacementText="AssemblyFileVersion("$(Major).$(Minor).$(Build).$(Revision)")]"
/>
2. Don’t account for TfsXXXX community tasks - it won’t run on VS 2008 with Team Explorer 9.0.
   Instead:  Use <Exec Command=”tf …”>.
   And if you have domain authentication in TFS and don’t run integration under domain user (which happens), you even can’t run tf in cmd-line mode: stupid authentication dialog will pop up.
3. Do not use local files to store version number. You will lose them when moving build to another host, so use only source control-dependent files.
   Might seem evident, but strange how many people miss this.
   So you’ll need to
   <Attrib Files="version.txt" Normal="true" />
4. Per Scott Colestock’s post, “since TfsVersion interrogates the build server’s workspace, and BuildNumberOverrideTarget comes early in the process… you might find that you don’t get the changeset number you expect”.
5. Don’t try to operate on XML node collections with tasks other then Xslt.
   For instance, XmlRead returns a Nodeset only as a string of “;”-joined
   Node.Value-s.
   XmlQuery supports Values property, though I wasn’t able to operate it’s results.
   Instead you can use ReadLinesFromFile (sadly, this will trim the lines).
   My personal choice for not-too-complex XML is… you can guess… yes, FileUpdate.
6. Don’t use FtpUpload. It won’t create folders for you.
   Just <Exec> a ftp…
   By the way, Windows’ ftp.exe client is unable to upload in passive mode. You can download passively, if you issue “LITERAL PASV“, but it won’t help the client.

Another problem with MSBuild bootstrappers that can screw all your installation…

If scheduling reboot after installing .NET or other prerequisites, bootstrapper copies itself to run from TEMP on boot. Why not staying where it is - who knows. And definitely, all the other MSIs are not copied, so it won’t find them. Reproduced for me.

Now give up and go write own bootstrapper. I’m setiously considering going for InnoSetup deployment.

For now, I’m trying a crazy ting, to copy MSI to TEMP for Setup.exe to find it after reincarnation… It’s the way to use given tools.

Now I know it.

It’s Windows Installer error 1618 that I didn’t see somewhy when attempting to fix .NET bootstrapper. Either I was inattentive or it wasn’t there: “Another copy of msiexec is running”. You can’t have two Windows Installers running, at least if one of them installs .NET. Information is from this maillist post.

Why? Implementation limitations, I believe. No sane reasons for such a contract, no article mentioned this. Surely, who’s going to declare own “features” of that kind - just a gentle hints that “you should better use bootstrapper”. Another reason to long for a systems where setup is just a copy.

So go pack a MSI, then make a MSBuild bootstrapper.

Oh wait, do I have to have 2 files in my installation? And my customers can’t download one file?

Yes. Or go for WinZip Self-Extractor, or WinRar, or whatever else.

What about build scripts? For WinRar, here’s a nice sample. For WinZip sfx, there’s also a command-line interface. Do a

wzipse32 %ARC_NAME% -setup -auto -runasuser -le -i ../icon.ico -wait msiexec -c ".\setup.exe"

WinZip will do the thing, though displaying some dialogs we don’t need at all.

Oh, WinRar won’t work, because it erases temporary files upon setup.exe completion. And we need to wait for msiexec, because setup.exe exits earlier. What a pity.

After Microsoft didn’t include .NET 3.5 in any current release of Windows (read “Vista service pack 1″), some of us developers who believed in long-ago C# 3.0 ads, had to roll back to .NET 3.0 and C# 2.0.

Including me.

Not mentioning pain of rolling back all LINQ tasties in code and debugging our replacements (do you know why you can’t implement Where() just with yield-style function?), it means installing .NET 3.0.

It seemed to be not a problem, having a MSBuild code snippet that worked for 3.5 - just replace 3.5 with 3.0:

<ItemGroup>
<BootstrapperFile Include="Microsoft.Net.Framework.3.0">
<ProductName>.NET Framework 3.0</ProductName>
</BootstrapperFile>
<BootstrapperFile Include="Microsoft.Windows.Installer.3.1">
<ProductName>Windows Installer 3.1</ProductName>
</BootstrapperFile>

<Target Name="Bootstrapper">
<GenerateBootstrapper
ApplicationName="blah blah blah blah"
BootstrapperItems="@(BootstrapperFile)"
BootstrapperKeyFile=""
OutputPath="obj"
ComponentsLocation="HomeSite" CopyComponents="false"
SupportUrl="http://www.microsoft.com/downloads/details.aspx?familyid=10CC340B-F857-4A14-83F5-25634C3BF043"
Culture="en" Path="$(BootstrapperPath)"/>
</Target>

Won’t work.

Hell why? Who knows. Logs in temp don’t give proper diagnostics, nor system events do. WPF fails to install at some point onto every XP installations we have, including clean special test XP SP2 VMWare images downloaded from Microsoft.

Update: Here’s the answer.

Now I rolled back to dotnetfx3setup.exe redistributable. It’s 2.8M compared to 300K bootstrapper generated by MSBuild, and it’s a pain to see installer grow…And if .NET 3.0 is already installed it starts to uninstall it… and there is no key to make it just “check or install”, need to trick it via WiX.

But we have to pay it. Oh my.

All over the Web I see no evidence that any man walked this path before. For.NET 2.0 - a lot of mentions and code samples. And completely empty for .NET 3.0.

Did you ever try parsing C++? What about parsing for every combination of possible define-s at once?

Microsoft has a static code checker tools for C++, like FxCop for .NET.

That’s awfully complex task, you now see. In .NET you can analyze bytecode, and luckily don’t really need to account for conditional compilation. In C++ where everthing is about #ifdef - it’s hell.

I never heared of anything like that. Though Insure++ or AQTime do way more things when profiling runtime, they of course find only things that were on necessary execution paths.

Some pages that somewhy didn’t appear at top of Google searches for WiX help.

• WiXWiki.com - pretty useful and contains some snippets you won’t find in WiX manual or tutorial. Many of rare code snippets
• WiX development releases; some broken (last two broken in VS 2008), some contain the very things you’ll need;
• Justin Rockwood’s blog entry with Votive (Visual Studio addon) preprocessor variables. The blog itself is also thing to read;
• Alex Shevcuk’s blog entry on checking pre- and postconditions in WiX. His whole cycle on WiX - I also highly recommend it;
• Separately, .NET: WixWiki pages about checking for .NET runtime and deploying it with a bootstrapper generated with MSBuild;
• Another WiX wiki on mindcapers.net, I particularily needed ICE38 error explanation;
• John Robbins’ blog entry on patching patching with WiX. You don’t want number of auto-update patches to grow quadratically?
• Last but not the least, MSI properties reference on MSDN.

All the software is equally bad…

Well, I know some pieces of software that are good, but I won’t mention them today.

One of our msbuild vs nant choice criteria is  a product versioning support.

Sad or funny things are:

• both don’t include sane tool out of box;
• both need installing NAntContrib or MSBuild Community Tasks;
• even then both can’t sanely write AssemblyInfo.cs in order to leave other properties intact; if you had InternalsVisibleTo satellite unit test assembly - it won’t build then;

NAntContrib Version task is way more functional.

Additionally, MSDN page design where MSBuild reference manual lives, is awful.

I wonder if it’s the reason to choose a build tool, while they are approximately equal.

OTOH, MSBuild has a great GenerateBootstrapper task to install .NET framework from MSI (WiX sample included in the referenced wiki).

Windows Installer on itself is a pain. WiX saves you from it, mostly wrapping the complexity into XML constructs understandable to non-gurus.

Well, not completely - you still have to know:

• why the heck shortcut didn’t appear in start menu;
• how to create a web site shortcut;
• why to create RegistryValue for a program directory you create, and that’s the simplest things. Still better then MFC…

But a very good tutorial exists, maillist is very active and it’s easy to get support.
Maybe it’s another reason to love Unix-es, where “install IS just a copy”.
Though, one has to admit that installing Windows application involves somewhat more desktop integration.

I was looking for an auto-update, so that:

• application could say “New version available, download? Restart now?“;
• would use MSI;
• download should preferably use BITS.

Simple automatic updaters like this on Codeproject won’t do, because you can’t just download necessary files: Windows Installer will revert them back. Or you would have to “disable WI’s Resiliency“, which is also a bad idea.

Solution of my choice

ClickThrough subproject of WiX promises that, but it’s still beta and crashes. Still you can easily compare two MSI files, generate a MSP patch, and a RSS feed that client can check, download and even read by eye, so simple it is - a pretty tasty feature.

There is no .NET client code. For now, there is only a separate update.exe that will find, download and install new version. Still, you cannot ask user whether to download now or not, cannot tell what’s new, when to restart, etc.
But you can painlessly add a Start menu shortcut that will update your application.

Come later, I’m going to create a C# code that will check for updates, and probably publish it.

Other solutions

You can use Updater ApplicationBlock, or more exactly, latest version now lives at CodePlex. But Enterprise Library wasn’t updated for .NET 3.5, and I believe it’s abandoned.

You can use NSIS. Though, it’s not Windows Installer MSI packager, it creates own files and scripts.

For ClickOnce application, you can use Application.Deployment.

We’re choosing a build tool for VS 2008 project.
I’m not expert in any, only had put together a couple of NAnt scripts. So I give no objective technical information in this post, just impressions.

• First discussion of this choice, back in 2005, is almost strictly in favor of NAnt. Though, read it to know about MSBuild differences and advantages.
• Recent blog post by Oren Eini and int comments are also generally for NAnt. Additionally, you can use <msbuild> task in NAnt.
• The latter mentions Yahoo! ALT.NET group discussion. Newsgroup fan can read a lot there, I also had some. There are reasonable votes there for MSBuild because it’s just deployed with .NET, ant it’s nearly functionally equivalent to NAnt.
  Though, I can’t forget first discussion that mentioned lack of tasks like editing files or grabbing text output, or incrementing build number - (you know, devil is in details!) - it seems to me that NAnt wins.
  Here’s the most useful message there, as for me.

You see, I didn’t go too deep down simple Google search. That was enough for me, you’re welcome to research deeper and to correct me… especially until I started creating build procedure <grin>

And I trust mature and popular opensource products.

Till now, I did understand but didn’t remember the very situation, as I wasn’t the one creating thread pools and alike: all the core things were already there.
It’s a blessing to work for start-up.
You can think about

Monitor.Wait(toRetrieve); // will throw SynchronizationLockException

Now, after a look into docs, I see that one needs to hold the monitored object lock to avoid race conditions: a thread wakes up already holding a lock for an object it needs. The thread that did Pulse() acquired that lock and safely passed it to Wait()ing one.
Just watch out for deadlocks: it’s easy to wake up with one object lock, try to acquire another and die then…

lock(toRetrieve)
{
  Monitor.Wait(toRetrieve); // will throw SynchronizationLockException
}

As Doug Lea recommends in his excellent eternal “Concurrent Programming in Java“, just make it impossible. Lock only on a single object. If you really really need two locks, make them ordered: always acquire A then B.

If you do:

lock(a){
  // :
  lock (b)
  {
    //:
  }
  // or:
  b.SomethingThatLocks();
}

//then will you never ever do

lock(b)
{
  //:
  a.SomethingThatMightLock();
}

as calling other objects’ methods while holding a lock badly affects your karma.

For more in-detail look, read “EventWaitHandle and Monitors: how to pick one?” by Joe Duffy. He goes very deep down to system details.
He also did a nice brief overview of missed pulse problem - why do you always check wait condition in a loop:

while (toRetrieve.Count == 0)
{
  lock(toRetrieve)
  {
    Monitor.Wait(toRetrieve);
  }
}