Software Developer Blog |

Executing assembly location in a Dotnet (core) single file app

2022-01-19T10:00:00+00:00

How to get Assembly location for modern .NET projects embedded in a single-file app.

Executing Assembly location

To get executing assembly location you can invoke:

var executableDirectory = System.Reflection.Assembly.GetExecutingAssembly().Location;

But when you try to publish app as a single file, for example this way:

dotnet publish -c Release -r win-x64 --self-contained true -p:PublishSingleFile=true -p:IncludeNativeLibrariesForSelfExtract=true

then you will see following error:

error IL3000: 'System.Reflection.Assembly.Location' always returns an empty string for assemblies embedded in a single-file app. If the path to the app directory is needed, consider calling 'System.AppContext.BaseDirectory'

Current directory

You can achieve similar result with BaseDirectory property from AppContext class:

var executableDirectory = System.AppContext.BaseDirectory;

which will give you directory path of your single-file bundle.

Note that this is exactly what error message suggests, but I know that not everyone will read it to the end (maybe that’s why your here?) :)

Example usages

You can use this path for example to set current directory:

var executableDirectory = AppContext.BaseDirectory;
Directory.SetCurrentDirectory(executableDirectory ?? throw new Exception("Could not find out executable directory"));

Or to instruct configuration file provider without changing current directory:

var builder = WebApplication.CreateBuilder(args);
builder.Configuration.SetBasePath(AppContext.BaseDirectory);
builder.Configuration.AddJsonFile("CustomConfig.json");

Pay attention to lines order - setting new base path after adding json file will cause searching for json file in old base path.

Summary

Retrieving application exe directory is still simple thing, but modern .NET is full of small changes which affects our lives, so I hope this short how-to will be helpful.

If in your version of .NET another approach is required, please share it with others in comment.

Private Nuget feed in Docker .Net Core application

2021-04-03T10:00:00+00:00

Using public Nuget from Nuget.org is always easy, you just need internet connection. But when you need private Nuget feed it sometimes doesn’t want to work, especially in isolated Docker container.

Why private nuget feed won’t work in Docker without additional effort

When you restore nuget packages in your Windows, then access to feed is granted because of your credentials stored by OS. When you try to access private nuget packages in Docker then your credentials are not accessible (and you wouldn’t want to use them in CI/CD pipeline) so you need to resolve it some other way.

How to access private nuget packages in Docker

In theory, it should be easy, it should just end up with passing proper credentials to Docker somehow. In my case it was more difficult and I think similar problems are common, so I will share all steps I had to do.

Use nuget config file in Dockerfile to pass credentials to Docker

First you have to pass nuget config file in Dockerfile. To do so, you can use --configfile Nuget.config option in dotnet publish/restore commands. This file should be stored at solution level, not to need copy-paste it for every image from solution. Remember to set proper path (or/and copy it to convenient directory in image), otherwise you will see error claiming that this file doesn’t exist:

/usr/share/dotnet/sdk/2.1.803/NuGet.targets(525,5): error : File '/app/Nuget.config' does not exist.

You can read more about Nuget.config file with examples and explanations in official documentation.

Establishing ssl connection to private Docker feed

Next I faced error while trying to establish encrypted connection to my Azure DevOps server nuget feed.

It was something like this:

System.Net.Http.HttpRequestException: The SSL connection could not be established, see inner exception. 
---> System.Security.Authentication.AuthenticationException: Authentication failed, see inner exception. 
---> System.ComponentModel.Win32Exception: The client and server cannot communicate, because they do not possess a common algorithm
   --- End of inner exception stack trace ---
   at System.Net.Security.SslState.StartSendAuthResetSignal(ProtocolToken message, AsyncProtocolRequest asyncRequest, ExceptionDispatchInfo exception)
   at System.Net.Security.SslState.CheckCompletionBeforeNextReceive(ProtocolToken message, AsyncProtocolRequest asyncRequest)
   at System.Net.Security.SslState.StartSendBlob(Byte[] incoming, Int32 count, AsyncProtocolRequest asyncRequest)
   at System.Net.Security.SslState.ProcessReceivedBlob(Byte[] buffer, Int32 count, AsyncProtocolRequest asyncRequest)
   at System.Net.Security.SslState.StartReadFrame(Byte[] buffer, Int32 readBytes, AsyncProtocolRequest asyncRequest)
   at System.Net.Security.SslState.PartialFrameCallback(AsyncProtocolRequest asyncRequest)

I solved this one by adding DOTNET_SYSTEM_NET_HTTP_USESOCKETSHTTPHANDLER=0 before using dotnet publish, like this:

RUN DOTNET_SYSTEM_NET_HTTP_USESOCKETSHTTPHANDLER=0 dotnet publish project -c Release --configfile Nuget.config

Still not able to authenticate - Response status code does not indicate success: 401 (Unauthorized)

After doing above steps I still faced issue while trying to use this nuget feed. I knew that credentials I used were corrected, the account which I used was created especially for nuget purposes and I knew that it had all proper permissions set and still was not able to connect. Error said:

/usr/share/dotnet/sdk/2.1.803/NuGet.targets(123,5): error : Unable to load the service index for source http://myazuredevopsserver:8080/tfs/DefaultCollection/_packaging/PackageFeed/nuget/v3/index.json. [/app/App.csproj]

/usr/share/dotnet/sdk/2.1.803/NuGet.targets(123,5): error : Response status code does not indicate success: 401 (Unauthorized). [/app/App.csproj]

The command '/bin/sh -c DOTNET_SYSTEM_NET_HTTP_USESOCKETSHTTPHANDLER=0 dotnet restore App --configfile Nuget.config' returned a non-zero code: 1

Azure DevOps server admin didn’t know why and couldn’t read it from server logs so suggested trying another approach, which by the way may be good for security - using authentication with PAT (personal access token). He generated such a token, and we passed this token to Nuget.config file and feed finally started to work. Now it looks like this:

<?xml version="1.0" encoding="utf-8"?>
<configuration>
<config>
    <add key="http_proxy" value="http://myazuredevopsserver:8080@company.corp" />
</config>
  <packageSources>
    	<clear />
    <add key="nuget.org" value="https://api.nuget.org/v3/index.json" />
    <add key="Internal" value="http://myazuredevopsserver:8080/tfs/DefaultCollection/_packaging/CompanyPackageFeed/nuget/v3/index.json" />
  </packageSources>
	<activePackageSource>
		<add key="All" value="(Aggregate source)" />
	</activePackageSource>
  <packageSourceCredentials>
          <Internal>
            <add key="Username" value="xxx" />
            <add key="ClearTextPassword" value="xxx" />
            <add key="ValidAuthenticationTypes" value="basic" />
          </Internal>
  </packageSourceCredentials>
</configuration>

Where xxx is the token (exactly the same value passed both to Username and ClearTextPassword field). Note that (as far as I know) this token can be generated max for 1 year, so better set some reminder in your calendar, or red builds in build server will remind it for you after a year :)

Local Nuget feed as a last resort

For a while till I didn’t get this working I created local Nuget feed, to have some workaround until I figured out proper solution. With local feed you can store Nugets in local directory and commit it to repo. But of course you need to maintain it by yourself, manually or with some automation which will commit new Nuget versions for you, so I don’t recommend it until it’s really the only way to get job done.
There are articles on the internet how to set it up, I will just show you screenshot of how it looked in my case (it’s actually screenshot from commit which removed it from my project):

Summary

I must admit that this problem ate much of my time, moreover I had to ask Azure DevOps server admin for help with reading server logs, and trying to figure out how to solve it. I didn’t have time to write this post earlier, but questions about private feed were common in my inbox, so I know that many of you face similar issues.

If your case is yet different, feel free to ask question in comment. And of course feedback is welcome, if this article has helped you or is missing something, it is very valuable for me to know it :)

Disable ctrl+shift shortcut in Windows 10 to stop language switch keyboard layout hot key

2020-05-01T12:00:00+00:00

Every Windows user knows that typing CTRL + SHIFT changes keyboard layout and that typing it by mistake is easy, which can be really annoying. Luckily this can be changed.

Removing CTRL + SHIFT shortcut hot key in old Windows version

We used to change this behaviour in following location:
Control Panel » All Control Panel items » Language » Advanced settings » Change language bar hot keys » Advanced Key Settings » Between input languages » Change Key Sequence…

Windows 10 solution which didn’t work in my case

With new Windows 10 installation (which happened to me) I tried to disable this language hot keys, but the way which I found on the internet didn’t work in my case (configuration in given location didn’t exist).
There is a chance that it will work for you, so just in case I’ll write it here: Control Panel\Clock, Language and Region\Language\Advanced settings\Change Language Bar hot keys.

Change language hot keys to switch keyboard layout in Windows 10

After spending a while doing manual searching in all Windows configuration places, I’ve finally found it. I’ve disabled CTRL + SHIFT shortcut here:
Advanced keyboard settings (you can find it with Start search or with Windows Settings search), choose Input language hot keys, next Advanced Key Settings, then Between input languages and set Not Assigned, or replace with other keys combination.
Like in following screenshot:

Summary

If you don’t need to switch keyboard layout often, you can easily disable this hot key. Those who do need to change language keyboard layout on a daily basis (working with multiple languages at the same time) may leave this untouched or change shortcut to version which is less likely to type by mistake.

If your version of Windows have this setting in yet other location, please share it with others in comment :)

Error in MSSMS: Attempt to retrieve data for object failed. Invalid Urn filter…

2019-12-30T06:00:00+00:00

When you work with MS SQL Server in Docker container, you may come across problem with Microsoft SQL Server Management Studio. See what’s going on.

How Invalid Urn Filter error looks like

What you see when error occurs is Microsoft SQL Server Management Studio message box saying:

Attempt to retrieve data for object failed for Server ‘PC-NAME’. (Microsoft.SqlServer.Smo)

Additional information: Invalid Urn filter on server level: filter must be empty, or server attribute must be equal with the true server name. (Microsoft.SqlServer.Smo)

When you press Copy message text you will have the same message extended with For help, click: http://go.microsoft.com/fwlink?ProdName=… but in my case this url doesn’t work.

When you press Show technical details you will see Advanced information:

===================================

Attempt to retrieve data for object failed for Server 'PC-NAME'.  (Microsoft.SqlServer.Smo)

------------------------------
For help, click: http://go.microsoft.com/fwlink?ProdName=Microsoft+SQL+Server&ProdVer=14.0.17230.0+((SSMS_Rel_17_4).180313-0650)&EvtSrc=Microsoft.SqlServer.Management.Smo.ExceptionTemplates.FailedOperationExceptionText&EvtID=Attempt+to+retrieve+data+for+object+Server&LinkId=20476

------------------------------
Program Location:

   at Microsoft.SqlServer.Management.Smo.Server.GetSmoObject(Urn urn)
   at Microsoft.SqlServer.Management.UI.VSIntegration.ObjectExplorer.OpenTableHelperClass.SelectFromTableOrView(Server server, Urn urn, Int32 topNValue, Boolean scriptForSelectingRows, Boolean isDw)
   at Microsoft.SqlServer.Management.UI.VSIntegration.ObjectExplorer.OpenTableHelperClass.GetScriptForTopNRows(NodeContext parentContext, Int32 topNValue, Boolean scriptForSelectingRows)
   at Microsoft.SqlServer.Management.UI.VSIntegration.ObjectExplorer.OpenTableHelperClass.SelectTopNRows(NodeContext parentContext, Int32 topNValue)
   at Microsoft.SqlServer.Management.UI.VSIntegration.ObjectExplorer.SelectTopNRows.Invoke()
   at Microsoft.SqlServer.Management.UI.VSIntegration.ObjectExplorer.ToolsMenuItemBase.MenuInvokedHandler(Object sender, EventArgs args)

===================================

Invalid Urn filter on server level: filter must be empty, or server attribute must be equal with the true server name. (Microsoft.SqlServer.Smo)

------------------------------
For help, click: http://go.microsoft.com/fwlink?ProdName=Microsoft+SQL+Server&ProdVer=14.0.17230.0+((SSMS_Rel_17_4).180313-0650)&LinkId=20476

------------------------------
Program Location:

   at Microsoft.SqlServer.Management.Smo.Server.CheckValidUrnServerLevel(XPathExpressionBlock xb)
   at Microsoft.SqlServer.Management.Smo.Server.GetSmoObjectRec(Urn urn)
   at Microsoft.SqlServer.Management.Smo.Server.GetSmoObjectRec(Urn urn)
   at Microsoft.SqlServer.Management.Smo.Server.GetSmoObjectRec(Urn urn)
   at Microsoft.SqlServer.Management.Smo.Server.GetSmoObject(Urn urn)

Why MSSMS with MS SQL Server in Docker container causes problems

When you recreate Docker container with Microsoft SQL Server from scratch, it’s like you create new virtual machine, with whole new SQL Server installation. As far as I know, MSSMS was not designed for such situation (even in virtualization era it was not usual case, like it is in containerization era), and it has some internal state cached. It’s probably SQL Server internal instance name, which isn’t the same for new installation, that’s why you see message: server attribute must be equal with the true server name.

How to resolve MSSMS error: Attempt to retrieve data for object failed for Server, Invalid Urn filter on server level

Unfortunately at the time of writing I didn’t find permanent solution. All you can do is to restart Microsoft SQL Server Management Studio every time you recreate Docker container. But what you should do first is to install the newest version, maybe some upgrade will fix this. Moreover, I don’t remember exactly what was the case, but older MSSMS versions got some other related issue, which got fixed.

Summary

To summarize: install the newest MSSMS version, but you will probably still need to restart Management Studio every time you start new instance of SQL Server in Docker container.

If you know something more, please let me know in comment, and of course if you have some questions :)
And share it with others who deals with SQL Server in Docker container too.

Docker run problem no such file or directory (or other strange message)

2019-12-23T08:00:00+00:00

When you cannot start Docker container correctly and have problems like No such file or directory or some other unintuitive message, there is one thing you should check first…

Correct line endings (EOL) in files copied to Docker

Check if you pass to Docker image scripts (or other text files) with Unix line endings (LF)!

With wrong EOL you will be lucky if message somehow indicates it, for example:
line 3: $’\r’: command not found

but it often looks in a way which gives no clue, what is wrong.

I once spent more than 2 hours because of it, so I hope that with this blog post you will not have to :)

If you have problem which looks similar but is not caused by wrong EOL, please share it in comment, I will try to help you figure it out, then update this post for others.

Docker logs from last run container

2019-12-23T06:00:00+00:00

Running Docker containers is what we do often, so it’s good to know how to do it to see Docker logs immediately.

Running Docker container in foreground mode

When you run container without detach flag (-d) then you usually need to stop container when you want to close terminal or type another command. You can manipulate this with -a flag, but it’s easier just to run in detached mode.

Running Docker container in detached mode

When you run container in detached mode (with –detach, -d or -d=true flag) you can safely close terminal or type another command without affecting running containers, but in this mode by default you don’t see logs.

If you know your container name you can quickly run docker logs -f CONTAINER_NAME but when you don’t know it (which is often the case), then many people usually end up with this solution: quickly run docker ps, check container name, and quickly pass it to docker logs command. With terminal with poor autocompletion this takes long.

How to see docker logs from last ran container?

We usually don’t focus on Docker run output, which in detached mode is… container ID. So we don’t need to use Docker ps to check container name which just got started, we already have it’s ID, which can be used in Docker logs command as well. So we can copy and paste it after Docker run. Example:

$ docker run -d app
d6a2d30a9a85abf110192e944963bb0ce157dde143139c24abb4bbf038403451
$ docker logs -f d6a2d30a9a85abf110192e944963bb0ce157dde143139c24abb4bbf038403451

It’s already more convenient, but… better solution is to use it without need to do copy and paste. How? Simply use xargs to pass Docker run output (container id), to Docker logs. Example:

docker run -d app | xargs docker logs -f

And we have it :) Very simple, very convenient.

Why pass -f flag to Docker logs command?

In all previous examples I have passed -f flag to Docker logs command, which stand for follow. It’s not necessary, but without it Docker will only show already created logs. With follow option, it will show already created logs and will attach our terminal to see new logs, which is what we usually want, to monitor what is going on.

Summary

As Docker comes from Linux world it’s meant to be used from command line and is optimized for this. So having container ID returned from Docker run command (in detached mode) is probably meant to be passed to next command (or script), so we can safely use it to follow logs immediately.

Let me know in comment if you like this trick or share it with others with following buttons or direct url :)

Error 459 - permission denied while trying to run git bisect with script

2019-11-16T06:00:00+00:00

How to fix error occurring during git bisect run - 459: /usr/lib/git-core/git-bisect: script.sh: Permission denied.

Full logs of git-bisect with 459 error

As you see, problem with running script doesn’t stop git-bisect so it finishes job. I hope in the future Git team will improve this, to stop when something goes wrong - such result is not reliable, there is no point to continue process.

~/Projects/app ((no branch, bisect started on fix_bug)) $ git bisect run /usr/lib/git-core/scripts/check-app.sh
running /usr/lib/git-core/scripts/check-app.sh
/usr/lib/git-core/git-bisect: 459: /usr/lib/git-core/git-bisect: /usr/lib/git-core/scripts/check-app.sh: Permission denied
Bisecting: 3 revisions left to test after this (roughly 2 steps)
[0e804f177d84772564ef74cb3905c4c1147d2f78] commit title
running /usr/lib/git-core/scripts/check-app.sh
/usr/lib/git-core/git-bisect: 459: /usr/lib/git-core/git-bisect: /usr/lib/git-core/scripts/check-app.sh: Permission denied
Bisecting: 1 revision left to test after this (roughly 1 step)
[e02f6d38ed24ee37caddd494e04742dcfa2eef84] another commit title
running /usr/lib/git-core/scripts/check-app.sh
/usr/lib/git-core/git-bisect: 459: /usr/lib/git-core/git-bisect: /usr/lib/git-core/scripts/check-app.sh: Permission denied
Bisecting: 0 revisions left to test after this (roughly 0 steps)
[4abe5155e0b4781b3247b289ebea23c081889150] yet another commit title
running /usr/lib/git-core/scripts/check-app.sh
/usr/lib/git-core/git-bisect: 459: /usr/lib/git-core/git-bisect: /usr/lib/git-core/scripts/check-app.sh: Permission denied
4abe5155e0b4781b3247b289ebea23c081889150 is the first bad commit
commit 4abe5155e0b4781b3247b289ebea23c081889150
(COMMIT DETAILS)
:040000 040000 92dfff033a3f883db1311fe9a8a3b50a5c901cbd a60d2bf636421197930d7194fbd56b954880e50e
bisect run success

Quick solution

Quick solution is to allow everyone to access directory with script.
To do so you just need to use chmod command like this:

# Make directory available for all 
chmod -R 777 PATH_TO_DIRECTORY

In my case it was fine, I have even moved whole script to /usr/lib/git-core/scripts directory, so it’s globally available now, so I didn’t investigate what exactly has gone wrong.

Summary

To summarize: 459 access problem can be fixed without investigation, but only if giving permission for every user is acceptable in your case. If this quick solution is not enough for you or doesn’t work in your case, please share your own with others in comments :)

TFS Linux Agent with error: The user’s home directory could not be determined

2019-06-08T06:00:00+00:00

How to fix problem with dotnet restore on TFS 2017 Linux agent - The user’s home directory could not be determined.

I have recently configured Linux agent for Team Foundation Server 2017 to be able to run dotnet core project there. But during dotnet restore step error was thrown: Error: /usr/bin/dotnet failed with return code: 1 with more explanation: The user’s home directory could not be determined. Set the ‘DOTNET_CLI_HOME’ environment variable to specify the directory to use. It may occur also on newer versions of TFS and on its online version - Visual Studio Team Services (VSTS), currently called Azure DevOps.

After quick search I knew that dotnet needs to know where home is located for some operations.

Agent logs

Full agent logs for dotnet restore step:

[section]Starting: Restore
==============================================================================
Task         : .NET Core
Description  : Build, test and publish using dotnet core command-line.
Version      : 1.0.1
Author       : Microsoft Corporation
Help         : [More Information](https://go.microsoft.com/fwlink/?linkid=832194)
==============================================================================
[command] /usr/bin/dotnet restore /home/tfs-agent/PS-TFS-AGENT-LNX01_agent01/_work/2/s/ProjectName/ProjectName.csproj
The user's home directory could not be determined. Set the 'DOTNET_CLI_HOME' environment variable to specify the directory to use.
[error] Error: /usr/bin/dotnet failed with return code: 1
[error] Dotnet command failed with non-zero exit code on the following projects : /home/tfs-agent/PS-TFS-AGENT-LNX01_agent01/_work/2/s/ProjectName/ProjectName.csproj
[section] Finishing: Restore

What didn’t help

First I checked if $HOME variable can be seen on this Linux Machine, logged in as user which is given for TFS. It was ok. Then I set DOTNET_CLI_HOME as Linux environment variable, pointing to $HOME. It didn’t help. There were much more solutions to check, looking through the internet, but next one worked for me.

Solution

This solutions is not elegant, but it works and in my view is acceptable, and it has one big advantage - it doesn’t require access to Linux Machine ssh. I set DOTNET_CLI_HOME as build definition variable, pointing to /tmp directory. That’s it, now dotnet restore and build works :)

Summary

I don’t know if this solution will work for everyone, but it takes only 2 clicks to try, so definitely it’s good to start with :)

Error using jquery-csv: $.csv is undefined

2019-05-15T06:00:00+00:00

Javascript has nice tool to read and transform csv data, with handy api but with unexpected inconvenience as well.

Recently in one web project I had to convert csv data to javascript objects. Quick DuckDuckGo research has shown me library which suits my needs: jquery-csv. I did everything the same way as in official demo, but exception $.csv is undefined kept blocking me. My IDE was able to go to declaration without any problems, still browser was unable to got it.

Solutions

Wait till document is ready

The only solution which I found on the internet was to wait till document is ready.

$( document ).ready(function() {
    // Use jquery-csv here
});

Unfortunately this was not my case, I already did it.

Check libraries load order

Finally, comparing my code with demo line by line, I found the solution. I had jquery-csv import after usual jQuery. So all I had to do was moving <script src="jquery-csv.js"></script> line after <script src="jquery-3.4.0.min.js"></script>. That’s because this library tries to follow jQuery api, which seems to be convenient idea, but practice shows that it has big disadvantage - jQuery deletes or mutates library namespace.

Summary

That’s it, but coming to this point took me some time and internet was not helpful back then, so I decided to share this solution. It might be helpful for people stuck in the same situation :)