There are a lot of answers on this topic but in order to aggregate the steps I usually follow for future reference I'm noting them in this blog post.

There are few things more annoying than Intellisense going wobbly and flagging successfully compiling code with errors. Obviously the first step is to restart Visual Studio but if the problem persists you need to try something more. These steps are for Visual Studio 2017 Community with Resharper.

You can check whether the underlines have disappeared after each step or run them all:

1. Unload then reload the problematic project from Solution Explorer. To do this, right click the project, select Unload Project and then Reload Project. This sometimes helps clear incorrect highlighting due to Resharper especially after merges.
2. Clear the Resharper cache. This is accessed by going to Resharper > Options > General > Clear caches in the menu. You will need to restart Visual Studio to see if this step worked.
3. Disable Resharper from Tools > Options > Resharper > Suspend Now. Then start it again from the same location.
4. Close Visual Studio and then delete the .vs folder from the source folder. This is a hidden folder at the same level as the .sln file.
5. With Visual Studio closed delete the obj folders from the problematic project folders.

Today was spent tracking down a very weird bug. In production we were seeing an important part of our document reading fail. We kept getting NullReferenceExceptions when calling AppendLine on a non-null StringBuilder. It didn't prevent us reading the document however the result would be significantly different to the same document on a local instance of our program.

It only started occurring after the production server had been running for a few days which meant we couldn't debug it locally. We were running .NET 4.5.2.

Luckily we had lots of logging to track down the issue. The problem was in a class like this:

public class AlgorithmLogicLogger
{
    private readonly StringBuilder stringBuilder;

    public AlgorithmLogicLogger()
    {
        stringBuilder = new StringBuilder();
    }

    public void Append(string s)
    {
        var message = BuildStringDetails(s);

        stringBuilder.AppendLine(message);
    }

    private static string BuildStringDetails(string s)
    {
        return $"{DateTime.UtcNow}: {s}";
    }
}

This was a class which was originally intended to provide detailed logging for a complicated algorithm.

The call to StringBuilder.AppendLine() inside Append was throwing a NullReferenceException.

After ensuring no weird reflection was taking place and using Ildasm to inspect the compiled code we were sure it wasn't possible for stringBuilder to be null. It was always instantiated in the constructor and never changed elsewhere.

The next working theory was that a multi-threading issue was somehow calling Append prior to the field being set. This was also discounted both because it wouldn't have been possible and also because the code in question was not called from multiple threads.

After ensuring that it wasn't the case that the garbage collector wasn't somehow incorrectly collecting the string builder (because it was only written to, never read, the reading hadn't been implemented yet) we were beginning to run out of ideas.

Following on from my much older posts about using ASP.NET Identity 2 to manage user accounts in MVC 4 sites, today I needed to use Identity on an ASP.NET Core MVC site.

As with previous versions, the current Identity library for .NET Core 1.1 uses Entity Framework out-the-box. Luckily it's much easier to change this behaviour for the simplest register/log-in flow.

Install

Firstly you need to get the right NuGet package. For the Identity library without Entity Framework this is:

Microsoft.AspNetCore.Identity

Once this is installed in your project you need to provide your own implementation of the IUserStore<TUser> and IRoleStore<TRole>.

Unlike previous versions the TUser no longer needs to implement a specific interface.

For my project I was using PostgreSQL as the database and Dapper.Contrib as the ORM. My User class was simply:

[Table("\"user\"")]
public class User
{
    [ExplicitKey]
    public Guid Id { get; set; }

    public string UserName { get; set; }

    public string NormalizedUserName { get; set; }

    public string Email { get; set; }

    public string PasswordHash { get; set; }
}

(I had to do some faffing around to make sure the table name was detected correctly in my schema but could have called the table anything, it doesn't have to be called "user").

I've spent a few days getting very in-depth with White for WPF UI automation so in order to store the information while it's still fresh in my mind I am writing a blog post.

White is a framework for automating interactions with desktop Windows applications. It is built on top of the UIAutomation library which is part of the .NET framework. The project is currently not very active on GitHub but despite this it's still very good at dealing with desktop interactions on Windows.

Automate all the things

For this tutorial I've decided to automate interactions with the Visual Studio 2015 application; since it's software that most people interested in following this tutorial will probably have.

I'm creating my application as a .NET 4.6.1 console application. Generally White is used for automated UI tests, so you might be using a test project. Whichever project type you choose the information in this tutorial will still be relevant.

First we need to add the TestStack.White NuGet package. Right click your project and select Manage NuGet Packages:

I've done quite a bit of work with Selenium on and off, mainly for browser automation tests. There are many high quality official libraries for Selenium covering a range of languages. It's an amazing technology and a real joy to work with; however sometimes stuff can go wrong and it's useful to know what's going on under the hood in order to debug where the problem lies.

For this reason I present my very-probably-wrong guide to Selenium.

Gecko, drivers!? What's going on?

There are 3 components in code that interacts with Selenium (ignoring RemoteWebDriver use cases):

1. Your code
2. A driver
3. The browser

For each browser that supports Selenium, the browser vendor provides a "driver". This is usually a small executable (.exe) program tailored for a specific browser:

• Chrome / Chromium
• Firefox
• Edge
• Other drivers available

This driver is the part your code actually interacts with. The driver then communicates with the browser using whatever magic the vendors design. In order to change browsers while using the same code, the drivers expose interfaces/endpoints that implement the WebDriver specification.

Put simply, this means the drivers are a simple RESTful API.

Demo

To demonstrate this I'll use the ChromeDriver version 2.29 which is the latest version at the time of writing. Drivers can only talk to the versions of the browser they are compatible with. In this demo I'm running Chrome on version 58.