Getting an "RPC endpoint not found/not listening" exception when connecting to a remote machine with PowerShell

Lately I've been dealing with a lot of remote management for the purposes of automating our deployment process for the product on which I'm working. I've been able to connect other (pre-configured) machines, but when I wanted to connect to my own machine in unit tests, I've been unable to do so until now. Each time I try to connect, I'd get an exception along the lines of "The remote RPC server is not responding". I double checked that my "Windows Remote Management (WS-Management)" service is up and running, so I was perplexed as to why I still couldn't connect. I had turned off my firewall (temporarily, of course), and as if that wasn't enough, I'd explicitly enabled the rules for Windows Remote Management. As it turns out, (at least when you're running Windows Server 2008 R2) the service runs by default, but is not configured to allow remote management by default. (Totally makes sense, right ? /sarcasm) To remedy this, you need only run the following under and Administrator command line:

winrm quickconfig

This will enable your machine to accept incoming connections. You should also ensure that your firewall has been properly configured to allow the remote management rules (pre-existing, come with Windows). Also make sure that your service is actually running.

Creating a certificate chain of self-signed certificates for development / testing / private environments

As anybody who's ever tried to develop secure services with SSL knows, it's expensive to buy trusted certificates from a certification authority. This is especially true if you're an independent developer who doesn't have a lot of resources. Therefore, we need to be able to generate self-signed certificates in order to develop and test our code before we actually go buy a Trusted Certificate for production. This tutorial will show you how to create a chain of trust and start generating certificates from a self-signing authority. The information here is based off of Microsoft's documentation on MSDN about the matter.

1. Create a signing authority certificate:

• makecert -n "CN=My Signing Authority" -r -sv MySigningCert.pvk MySigningCert.cer

You'll be prompted for passwords for securing the private key. Ensure that you remember them, you'll need them to create the merged file.


2. Merge the private key file and public key file into an encrypted key (this isn't mentioned in the MSDN article linked above, but you can find the documentation here):

• pvk2pfx /pvk MySigningCert.pvk /spc MySigningCert.cer /pfx MySigningCert.pfx /pi mycertpassword /po mycertpassword /f

This step isn't necessary for signing site certificates, but does make things more convenient for storing the certificate and installing it on different machines. Be careful: you should never leave keys laying around file systems on machines, they should always either: a) be stored in an encrypted store like that provided by Windows, or b) be stored on separate storage media that can be physically locked away with access only available to trusted personnel.


3. Start creating site certificates with your signing certificate:

• makecert -iv MySigningCert.pvk -n "CN=www.mywebsite.com" -ic MySigningCert.cer -sv sitekey.pvk sitekey.cer -pe

Now, as above, I recommend that you merge the .pvk and .cer into a .pfx for easy transport and storage.

Retargeting a Windows 8 application to Windows 8.1

Apparently, fuck Windows 8. So says everybody. Including Microsoft. That's why at some point you're going to have to retarget your Windows 8 app (if you were crazy enough to make any) for Windows 8.1. Fortunately, Microsoft provides a guide for doing so in Visual Studio 2013 here. Fortunately, it's as simple as right-clicking on your solution in the Solution Explorer and clicking on Retarget for Windows 8.1

Making TFS builds consistent with desktop builds when invoking MSBuild directly on a .*proj file

As it turns out, MSBuild has more than a few quirks when being invoked through TFS compared to being invoked through a command line or from Visual Studio. Some of them are pretty well documented. Others are not, like the fact that in a .*proj file, the OutDir variable is inherited to sub-MSBuild tasks.  There's also quirks because OutputPath is used to determine OutDir, but not in all cases. If you're going to specify OutputPath in the properties when invoking the MSBuild task, you should also explicitly override the OutDir variable as well to ensure consistency, unless you **TRULY** understand the differences between the two and how MSBuild determines OutDir, and you **REALLY** want it to be that way.

Getting code signing to work with ClickOnce on a TFS Build Agent

Code signing is a giant pain in the butt. You have to :

• Obtain the certificate for signing the code by:
1. buying the certificate from an issuer.
2. generating your own self-signed certificate
• Configure ClickOnce within your project file with the following property elements:
• <signmanifests>true</signmanifes>
• <manifestcertificatethumbprint>A387B95104A9AC19230A123773C7347401CBDC69</manifestcertificatethprint>
• Log into your machine **as the user running the build controller / agents ** and import the key to their user Personal certificate store!
• Run 'certmgr.msc' from the Run command in the start menu (WinKey + R is the hotkey)
• In the Certificate Manager that comes up, go to Personal in the tree, right-click, and select All Tasks -> Import ...
• In the Certificate Import Wizard window that comes up, select Next to move to the 'File To Import' screen.
• Select your certificate file, which has the same thumbprint as specified in your project file, then click Next to move to the 'Certificate Store' screen.
• In the 'Certificate Store' screen, select the 'Place all certificates in the following store' option, then click Browse to select the store. Choose 'Personal' in the selection window. Click Next to move to the "Completing the Certificate Import Wizard" window.
• On the "Completing the Certificate Import Wizard" window that comes up, click Finish to import the certificate.

You should now be able to build and sign your code on a TFS Build controller / agent.

Converting an existing Windows Store app to using the Prism Framework

I began converting an existing Windows Store App to using the Prism Framework provided by Microsoft. However, I'm running into the following error:

The primary reference "Microsoft.Practices.Prism.StoreApps" could not be resolved because it was built against the ".NETCore,Version=v4.5.1" framework. This is a higher version than the currently targeted framework ".NETCore,Version=v4.5".

This post on stackoverflow.com recommends installing the Microsoft Build Tools 2013 package, which is available here:
http://www.microsoft.com/en-ca/download/details.aspx?id=40760

That didn't work.

I later realized that I had installed Prism with NuGet, so I went and checked the publishing dates on the versions. The latest (and default, which I had installed) was 1.1.0. The date on 1.0.1 was much older, and after reverting to that version, I was able to get my program to compile and run with a few modifications to the steps in this tutorial. The modifications are as follows:

• Change the return type of the App.OnLaunchApplication method to 'void' to match the 1.0.1 version of the Prism.StoreApps library.
• In the App.OnLaunchApplication method, ensure that there's a call to :
• NavigationService.Navigate("Main", null); where "Main" is the initial page name, and there's a MainPage class in your Views folder.
• Move the existing MainPage class into the Views folder in the root of the project.

Creating my first Windows 8 store app

As you may or may not be aware, there are multiple types of applications that can be created for Windows 8:

• Windows store apps, which use the new Metro interface
• Desktop-based apps which are like those created for previous versions of Windows that can still run in the Desktop app.

I'm quite familiar with creating WPF apps for Windows, but Metro apps are new, and those are what I'll be working on. With that in mind, Microsoft provides the Prism framework which helps provide additional classes, interfaces, events etc to help people develop Windows Store apps that keep consistent with Windows 8 design principles and help the apps perform properly. I'll be starting with the MSDN link here.

Beginning to work with Windows Store apps

I really hate Windows 8. I think the majority of the applications that have been written for it are complete pieces of shit, for the following reasons:

• The developers who wrote them didn't pay any attention to Microsoft's best practices and they :
• perform poorly
• don't follow UI conventions and are hard to understand as a result
• crash 
• don't always save data properly
• Many are piss-poorly written and adapted by third party developers for first-party systems because those first-parties don't want to write software in a competing ecosystem, and instead want to force users to use their ecosystem, which has their own set of flaws and deficiencies. Case in point: Google. At the time of this writing, there are no native Windows 8 applications put out by Google. There's no native YouTube app for Windows 8 (which there damn well should be), presumably because those fuckers couldn't find a good way to generate advertising revenue in a Windows 8 app. (can't really blame them for that because if I see ads in an app, I immediately delete it from my device without hesitation. I can't stand that shit.)
• Windows 8 is a shit operating system. It was built on the new Modern interface (aka Metro), and initially had piss-poor integration with the desktop paradigm on which all previous incarnations of Windows were based. Add to this the fact that Microsoft didn't give people an easy choice of which paradigm they wanted to use right off the bat, and the fact that in successive iterations like Windows 8.1 they've tacked on hacky additions to make the Metro interface more like the previous desktop interface, you end up with a shitty operating system that's a pain to use; this pain stems from the fact that it's a horrible amalgamation of multiple user interface paradigms. 

As long as Microsoft continues to force their shitty iterations of Windows on the world, I, as a software developer, will be forced to deal with it because of the immense investment most employers have in Microsoft technology. With that in mind, I'm going to start learning Windows 8 applications so that I can make myself more marketable to employers everywhere. I'm going to document my learning here for my usual reasons:

1. So that I have a reference for myself for the future
2. So that others may learn more easily what I have learned.

Resolving ssh: connect to host xxx.xxx.xxx.xxx port 22: Connection refused

There are a number of reasons why an SSH server may fail to allow a client to connect. Many aren't readily apparent, even from tailing system log files or using ssh -v on the client. Here are some of the ones I've encountered:

1. Incorrect permissions / ownership on the key files in /etc/ssh/
2. Incorrect permissions / ownership on the ~/.ssh/id_rsa private key file of the user as which we're trying to connect. I'll add more to this list as I encounter them.
3. systemd just plain being a piece of shit. Running 'systemctl restart sshd.socket' has fixed the problem in the past.

Apache fails to handle requests with "libgcc_s.so.1 must be installed for pthread_cancel to work"

I recently had an apache2 server go down while I was using it. I still don't know what caused it, but I do know how I fixed it, thanks to this thread on Launchpad. Adding libgcc_s.so.1 to the ld pre-load got me back up and running.

echo "/lib/i386-linux-gnu/libgcc_s.so.1" >> /etc/ld.so.preload
ldconfig

The value echoed is the path to the libgcc_s.so.1 file on your system. It can be found with:

gcc --print-file-name=libgcc_s.so.1

I hope this helps anybody who has a similar problem.

Waiting for the network to be up on the BeagleBone Black

It finally hit me today as I was reading over this article, and it should have hit me sooner. The article mentions that in order to get past the network startup hurdles in systemd, you need to wait on the NetworkManager service. However, I use connman. It didn't occur to me until today that they provide exactly the same functionality, and I just needed to swap one with the other. Now,  I have all my network dependent programs simply After=connman.service in the systemd service descriptor files, and they're golden.

ImportError: No module named pkg_resources when trying to use pip on the BeagleBone Black

I've recently been trying to use Python on the BeagleBone Black, for a number of reasons:

1. I want to learn a new language which could be useful to me in another job in the future (and this job, even better!)
2. Given all the effort I've put into making our embedded systems work on multiple platforms, I think I've finally got enough infrastructure in place that we can start leveraging other cross platform products to shorten up our development time.
3. As a scripting language capable of using bindings to other languages, Python should help me create very functional code that doesn't need extremely high performance in a very short amount of time and reduce my development time for complicated tasks.

Unfortunately, like many other things on the BeagleBone Black, things aren't going as well or as simply as you'd think they should at first glance. For starters, the pip package manager for Python isn't installed by default on the BeagleBone Black (at least not as of the 2012.12 image). So, I had to install that first:

opkg install python-pip

When I tried to run the package manager, I ran into the following error:

Traceback (most recent call last):

  File "/usr/bin/pip", line 5, in

    from pkg_resources import load_entry_point

ImportError: No module named pkg_resources

After Googling around for a bit, I found these questions on Stack Overflow. Apparently you must also have the setuptools package installed in order to be able to use pip because it's not a simple package manager like apt in Ubuntu. It's more like emerge in Gentoo, where it downloads code packages and is capable of compiling them and performing custom installations. Fortunately, there's an opkg package for that:

opkg install python-setuptools

After that was installed, it became a simple matter of finally importing the actual package that I originally wanted that started all of this:

pip install psutil

... the process utils library for Python

Undefined reference to `log' when compiling on Ubuntu 13.10 with gcc

After I upgraded to Ubuntu 13.10, I inexplicably started getting errors in a build that had been perfect for a very long time. It turned out, there was a significant change in the linker and a bug introduced. The fix is described beautifully on this blog post, but for convenience sake:

Add '-Wl,--no-as-needed' to your LDFLAGS

Slow login times in Ubuntu 13.10 (not just SSH)

I recently setup a new install of Ubuntu 13.10 for a server and found that a lot of my login times were slow when remotely logging in (and not just via SSH). The culprit turned out to the the /etc/nsswitch.conf file:

This line:
hosts: files mdns4_minimal [NOTFOUND=return] dns mdns 4 mdns

Should be changed to this line:
hosts: files dns

... to resolve the issue

Developing a cape for the BeagleBone (Black)

Due to the needs of our company, we're developing an in-house cape for the BeagleBone Black to integrate it with our equipment. Yes, there are numerous capes already in existence, but they each provide a singular function that would require stacking in combination with other capes to meet our needs. Additionally, it increases the number of external vendors on which we must rely to meet our demands. Instead, we've chosen to create a custom cape that has multiple extensions on the same cape and integrates nicely with our existing board stacks. Unfortunately, I'm not that familiar with BeagleBone Black capes, so I'm going to be learning how to create the software necessary for a new cape and configure it using the Device Tree system on the BeagleBone Blacks. I'll be tracking my progress and things that I've learned on this blog for my own future reference and hopefully it'll even help out somebody else.

Recovering the root password on a BeagleBone Black

I recently went to use a BeagleBone Black board on which I'd never booted from the eMMC before (that I could recall) and had been instead booting off of a microSD card. To the best of my knowledge, the root password on this board should never, ever have been changed from the default (blank password), but apparently it was. None of the passwords that I had ever used for any of my BeagleBone Blacks was working, which created a problem: I needed to recover the root password to this BeagleBone Black so that I could use it for my projects again. I was in luck. The BeagleBone Black has the tools that I needed to do it. I'm writing this blog post (like many other posts in the blog) so that I can have a reference to come back to if I ever need it in the future.

To recover the root password of a BeagleBone Black, you'll need the following items:

• A 5V, minimum 1 A dedicated power source (you shouldn't really be powering the board off of the micro USB port)
• An SD card (and reader, of course)
• An FTDI USB-to-serial cable for accessing the debug serial port of the BeagleBone Black.

Once you've acquired all of the required items above, perform the following steps:

1. Flash the SD card with an image that you can obtain from BeagleBone.org/GettingStarted.
2. Insert the SD card into the *unpowered* BeagleBone Black.
3. Apply power.
4. The BeagleBone Black should boot from the SD card (assuming that you've flashed the correct image)
5. Connect the FTDI serial cable to the board and your compter, and open your serial client of choice to connect to the board.
6. Hit enter once or twice, and a command prompt should come up, assuming you've used the correct settings:
1. 115200
2. 8N1
7. Log in using root and a blank password (should be the default on the SD card image that you downloaded from the BeagleBoard.org website.)

You're now logged into the image running on the SD card. However, we can't change the root password now because that will only change it for the root user on the SD card image. We need to change it for the permanent image stored on the BeagleBone Black's embedded eMMC flash. Being logged in as you are, perform the following steps to mount the eMMC image and change the password for that image's root user:

1. Mount the eMMC flash:  mount /dev/mmcblk1p2 /media/card
2. Change the root of the file system to be the partition on the eMMC flash which you just mounted: chroot /media/card
3. Change the password of the root user to something you know: passwd root
4. Exit out of the changed root: exit
5. Shutdown the BeagleBone Black : shutdown -h now
6. Disconnect the power from the board
7. Eject the microSD card.
8. Reconnect the power to the board
9. Watch the board boot up, and log in as root. You should be able to log in with the password that you just set.

Implicit rules with Makefile

In my ongoing quest to make my builds less complex and faster, I've been going through my Makefiles and trying to learn as much as possible to simplify it and leverage Make as much as I can. To that end, I discovered something incredibly useful today that I suppose I would have known had I taken the time to read the man page for Make:

make -p

This command will list all of the implicit rules for make, which you can then use to optimize the living crap out of your Makefile.

Installing NTP on an Ubuntu server

NTP is used to synchronized time between machines. You can read the Ubuntu HOWTO here.

Installing a TFTP server using xinetd on Ubuntu

I recently had need to redo an old server that had been running an ancient Gentoo installation for which there was no longer a software upgrade path, so I chose to install Ubuntu on it. Part of the requirements of the server were that it runs a TFTP server for hosting files for configuring embedded devices. I had previously found an article on how to setup TFTP through inetd, but I could find it, so I'm cobbling this tutorial together from various sources.

1. # apt-get install xinetd tftpd
2. Ensure that the following lines are in the /etc/services file:
1. tftp     69/tcp
2. tftp     69/udp
3. Open the /etc/xinetd.d/tftp file (create it if it doesn't exist) and ensure that it contains the following:

   # default: off
   # description: The tftp server serves files using the Trivial File Transfer \
   #    Protocol.  The tftp protocol is often used to boot diskless \
   #    workstations, download configuration files to network-aware printers, \
   #    and to start the installation process for some operating systems.
   service tftp
   {
       socket_type     = dgram
       protocol        = udp
       wait            = yes
       user            = root
       server          = /usr/sbin/in.tftpd
       server_args     = -s /tftpboot
       disable         = no
   }
   

4. # /etc/init.d/xinetd restart

5. Test the server:# tftp localhost

   tftp> get hello.txt
   Received 23 bytes in 0.1 seconds
   tftp> quit 

You should now be good to go. The (abridged) instructions for this were retrieved from this article.

Digging into Raspberry Pi: Part 2

So, further into my quest to build a control system with the Raspberry Pi, I've been looking around at ways to have simple input from the user for a simple system. The first thing that came to my mind was using an Xbox controller to provide input from the user:

• It's simple for the user to use, and given that many people grew up with game consoles, it'll be intuitive and not require retraining the user.
• The Xbox controller plugs into USB and should therefore provide an easy way to integrate a controller for my system.
• The command to download joystick support for the Raspberry Pi was dead simple: 
• sudo apt-get install joystick

The post located here also shows some options for the joystick on the Raspberry Pi. I'll post more information as I learn how to use the library to access the device.