Category Archives: software

fixing MSB3644 build errors and the point of .NET targeting packs

Build errors on build servers suck. If it builds locally, why the hell doesn’t it build on the build server? Well there’s plenty of reasons for that, but as a .NET developer is usually means that something you have on your machine that came with your Visual Studio install isn’t installed on the build server.
But you probably don’t want to install the full-blown VS on the build server, so the question now is: what bit do I need to install?

Recently I ran into the following build error on 1 specific build server (yep, not on another one, fun, fun, fun).

C:\Program Files (x86)\Microsoft Visual Studio\2017\BuildTools\MSBuild\15.0\bin\Microsoft.Common.CurrentVersion.targets(1111, 5): error MSB3644: The reference assemblies for framework ".NETFramework,Version=v4.5.2" were not found. To resolve this, install the SDK or Targeting Pack for this framework version or retarget your application to a version of the framework for which you have the SDK or Targeting Pack installed. Note that assemblies will be resolved from the Global Assembly Cache (GAC) and will be used in place of reference assemblies. Therefore your assembly may not be correctly targeted for the framework you intend.

It clearly has something to do with a project which targets .NET framework v4.5.2.
The answer in fact is right there. You need to install a “targeting pack”.
But WTF is that thing and why do I need it?

Apparently having the .NET framework itself installed on your machine isn’t enough to be able to build a project if that project targets a specific version of the .NET framework. It makes sense as the .NET framework installation is actually the runtime, used to run .NET applications, not build them.
If you want to build apps against a specific version of the .NET framework, you need that targeting pack as well on your build machine. This is either your development box, which has VS on it, and thus the required targeting packs which come with the VS installation. Or this can be your build server, where you have to install the targeting packs as well.

You can check what targeting packs you already have on a machine by checking the sub-folders in “C:\Program Files (x86)\Reference Assemblies\Microsoft\Framework\.NETFramework
For each supported framework version, there’s a v<version number> folder there. In my case there was no v4.5.2 folder on that one machine.

So where do you find those targeting packs for the various .NET framework versions? Microsoft luckily compiled a nice list where you can find all the download links and instructions.
For the versions listed as “included in VS 2017”, you can see them all listed in the VS 2017 installer if you go to the “Individual components” tab.

Look at all those .NET targeting packs.

So a shortcut to install all the packs at once, is to grab the VS2017 installer and use that one. But you might want to disable all the IDE specific stuff you won’t need on the build server though. clearscriptv8 load error after publishing a website

Crash Here’s another ReactJS.NET quirk I ran into lately. While working on an ASP.NET site using the ReactJS.NET Nuget package to render content using React.JS server side we got this error message on the web server after deployment:

Cannot load V8 interface assembly. Load failure information for ClearScriptV8-64.dll:
C:\Windows\Microsoft.NET\Framework64\v4.0.30319\Temporary ASP.NET Files\root\4e3fedda\f5d1e0ef\assembly\dl3\68f03368\00cf5237_117bd201\ClearScriptV8-64.dll: Could not load file or assembly 'file:///C:\Windows\Microsoft.NET\Framework64\v4.0.30319\Temporary ASP.NET Files\root\4e3fedda\f5d1e0ef\assembly\dl3\68f03368\00cf5237_117bd201\ClearScriptV8-64.dll' or one of its dependencies. The system cannot find the file specified.

The annoying bit here is that everything works fine locally, but not when it was deployed to the server using an Octopus package.
After debugging, searching online and going through the build log files I figured out that the problems could be caused by :

  • Missing DLL’s because of a missing Nuget package (quite obvious but not a problem in my case).
  • The VS 2013 C++ Redistributables are not installed on the server, which is a common cause for this error related to the ClearScript assemblies.

In my case the problem was a variation of the 1st problem. Once the web app was published, the DLL files where not in the expected bin\x64 & bin\x32 sub-folders, but in the root of the application’s bin folder. So the DLL’s where there, but not in the correct spot.

The cause of this problem is the _CopyWebApplicatonLegacy MS Build task which lives in the Microsoft.WebApplication.targets file.
This creates a _PublishedWebsites folder containing all the files you need to deploy to make your website run, yet it doesn’t trigger the build task used in the JavaScriptEngineSwitcher NuGet package which places those DLL’s in those x64 & x32 sub-folders. This special build task is included in the ReactJS.NET Nuget package, so you don’t have to do anything extra for this.
Because this task doesn’t get triggered in the step build to create the deployment package, the ClearScript-V8-64.dll, v8-x64.dll & ClearScript-V8-32.dll, v8-x32.dll don’t end up in their x64 & x32 sub-folders.

I fixed this by moving the files to their rightful location while creating the package for Octopus deploy with a PowerShell script. There’s probably a way to fix this with an extra build tasks too, but man, I spent so much time on finding this issue in the first place and I really didn’t feel like getting myself into that mess as well.

Photo by Ted Van Pelt, cc-licensed. clearscriptv8 compiler error

Here’s another one for the error message Googlers. Recently I ran into a nasty build error on TeamCity after adding ReactJS.NET Nuget packages to an ASP.NET MVC solution.
Locally everything built just find, but on TeamCity the build failed with the following error message when trying to compile the MVC views:

[AspNetCompiler] ASPNETCOMPILER error ASPCONFIG: Could not load file or assembly 'ClearScriptV8-32.DLL' or one of its dependencies. The specified module could not be found.

Normally you get this type of errors when an assembly can’t be found or isn’t copied to the bin folder for some reason.
In this case it turns out to be more or less the opposite case. The DLL is in the bin folder, but .NET should be ignoring it. Apparently ASP.NET tries to load all DLL files in the bin folder, which it should not do for the v8 ones, making it crash and burn.
The clue came from this StackOverflow post and this blog post. The fix in the blog post is a bit hacky but pointed me in the right direction. The SO answer to change the web.config is spot on.

So the trick is to exclude the offending binaries by listing them in the web.config compilation section:

        <trace autoflush="true" />
                <remove assembly="ClearScriptV8-64" />
                <remove assembly="ClearScriptV8-32" />
                <remove assembly="v8-ia32.dll" />
                <remove assembly="v8-x64.dll" />

This way the DLL’s are no longer automatically scanned, and your build can nicely go on compiling those MVC views without any trouble.

Photo by strange little woman on stream, cc-licensed.

disable flash and silverlight for safer surfing

Flash, Silverlight and (*gasp*) QuickTime plugins in your browser with the modern web are about as necessary as a horse whip is on a Tesla. Well I might be exaggerating a bit. There are still some useful sites out there that actually use these things. Intranet sites that run on IE only for example, or flaky game sites. But any self-respecting web developer has long ditched them in favor of fancy new HTML5 features.

So why would you still run these things in your favorite browser (Firefox right?) where they only take up extra memory and have a bunch of security problems that might end up causing you trouble. There have been enough exploits for the Flash plugin out there to be sure to actually update those plugins every time they ask for it. Which is about every week if I recall correctly.

Anyway, it’s better to turns those damn things off completely and only turn them on when you hit one of those web sites maintained by a dinosaur. That way you’re stopping that evil hacker from taking over your machine with his Flash exploit and you’re gaining some free performance along the way.

In Firefox you can turn those plugins off in your Tools menu, under Add-ons. Just select “Never activate” and you’ll be fine.
Switch it back to “Ask to Activate” if you’d need them again. That way they’ll never activate by accident either, if you forget to turn it back off.
On Chrome it’s a bit more elaborate, but the option “Let me choose when to run plugin content” sounds like a safe bet instead of having plugin code be ran willy-nilly.
IE? Ha! Who cares right?! For anything else, a properly aimed search query should find you the answer in no time.

Oh, and don’t forget to tweak your Flash security settings if you decide to keep it on after all.

The Firefox add-in screen with all plugins disabled. Just like it should be.

indent selected lines with tab in vim

.vimrcOne thing that annoyed me about using Vim was how much keystrokes it took to indent or un-indent a few selected lines of code. My (probably less than ideal) way of doing that was to go into visual mode, select the lines with the movement keys J or K, then use the keys to change the indenting which are < or >.
To indent another level, pressing dot after this would work.

In Visual Studio or a typical Windows text editor I’m used to simply selecting the lines by holding shift & moving the cursor keys up or down, then pressing TAB to indent and shift-TAB to un-indent.

I’m so used to using the cursor keys for text manipulation that it’s hard to unlearn this, so I was looking for key mappings to do the same thing in Vim.
Luckily this turned out to be rather easy. If you add the following to your vimrc file, you can shift-tab away to indent your code:

" TAB-mappings to allow indenting of selected text instead of using < & >
vnoremap <Tab> >
vnoremap <S-Tab> <

warmup your site or wordpress blog with a single command line statement


GNU Wget is a powerful tool when it comes to downloading files from the web or mirroring sites. It’s command line features can be daunting and not very obvious. With some experimentation, reading the (f..) manual and some Googling you can get it to do some pretty neat tricks for you.
All of that is from the command line too, which is great if you want to schedule this kind of magic or use it in a script.

For example, you might want to warm-up your site or WordPress blog so your homepage and all posts linked from it are present in your cache when a visitor arrives. I’m assuming you are using a caching on your site otherwise this is pretty pointless. For WordPress you can use a caching plugin like W3 Total Cache for example.

With Wget, it goes like this:

wget.exe --spider --no-directories --level=1 --recursive [1..9].*

The command line parameters (in order) mean something like:

  1. Crawl
  2. Crawl it like a spider (follow the links).
  3. Don’t create directories for downloads.
  4. Crawl 1 level deep (so anything linked on the homepage is OK, but don’t go deeper).
  5. Do this recursively (so it actually goes 1 level deep).
  6. Follow only links that start with "/201..." to "/209..." (it’s a regular expression).
    This one is a trick to have it only follow links to blog-posts because my URL scheme begins with the year of the post (2015, 2016, …). It’s good until 2099, which should do the trick I guess. :)
    This way I’m also avoiding it loading all tag, category or page links.

If your site has a different URL scheme you’ll have to change the accept regex pattern to fit your scheme.

You can download Wget from the GNU site. It’s Open Source and is available for Windows, Mac and various Unix systems.
For Chocolatey users, there is a wget package available to install it on your system.

verifying an md5 file hash with just powershell

You see this possibility for a lot of software downloads but if you’re like me you hardly ever end up doing it: verifying an install file’s MD5 or SHA-something hash.
For Open Source software this is however recommended if you aren’t downloaded from the official mirror (and even then) and certainly if it’s anything related to security (like Keepass for example).

But to verify that MD5 hash you probably need to install yet another piece of software you’re hardly going to need, so you end up not bothering at all.

Hold on a second.

If you have a recent Windows system with PowerShell installed, you probably have all you need to verify that MD5 hash.

Try this in the PowerShell command prompt:

Get-FileHash .\KeePass-2.30-Setup.exe -Algorithm md5

It should print out something like this:

Algorithm       Hash
---------       ----
MD5             CD430EB0F108BB192D2155C68EB7BB48

Which happens to be exactly the MD5 hash code listed on the site for that version of the Keepass installer. Yay!
Without that -Algorithm parameter it prints out the SHA-256 hash by default, but that’s longer and harder to compare visually even though it’s more precise.

That was easy and required no additional software.
Pretty damn sweet.

Photo by Julien Dumont, cc-licensed.