I am writing a software update process on Linux. Application is .NET 5 RC1 (Sept 15 2020 release). When a certain packet is received by my application, it downloads the software update to a sub-folder then spawns off the executable to perform the software update.
Unfortunately, using Process.Start and ProcessStartInfo seems to create a process that is attached to the main process. Since the software update must stop the process in order to update it, it also gets stopped because it is a child of the process, having been spawned via Process.Start.
How do I create a detached process on Linux? On Windows I am using PInvoke and the CreateProcess API with the DETACHED_PROCESS flag, see the following:
var processInformation = new ProcessUtility.PROCESS_INFORMATION();
var startupInfo = new ProcessUtility.STARTUPINFO();
var sa = new ProcessUtility.SECURITY_ATTRIBUTES();
sa.Length = Marshal.SizeOf(sa);
CreateProcess(null, "\"" + fileName + "\" " + arguments, ref sa, ref sa, false, DETACHED_PROCESS, IntPtr.Zero, Path.GetDirectoryName(fileName), ref startupInfo, out processInformation);
Here is my code for Linux. I had read that appending & to a process on Linux creates it detached, but that does not appear to be the case.
ProcessStartInfo info = new ProcessStartInfo
{
// Linux uses " &" to detach the process
Arguments = arguments + " &",
CreateNoWindow = true,
FileName = fileName,
UseShellExecute = false,
WindowStyle = ProcessWindowStyle.Hidden,
WorkingDirectory = Path.GetDirectoryName(fileName)
};
Process.Start(info);
I was unable to get nohup or disown to work from C#. Killing the parent process always resulted in the child process being terminated as well.
I ended up using at, which can be installed via sudo apt install at. The atd service is installed and will stay running even when rebooted.
Here is the C# code that I used:
// the following assumes `sudo apt install at` has been run.
string fileName = "[your process to execute]";
string arguments = "[your command line arguments for fileName]";
string argumentsEscaped = arguments.Replace("\"", "\\\"");
string fullArgs = $"-c \"echo sudo \\\"{fileName}\\\" {argumentsEscaped} | at now\"";
ProcessStartInfo info = new()
{
Arguments = fullArgs,
CreateNoWindow = true,
FileName = "/bin/bash",
UseShellExecute = false,
WindowStyle = ProcessWindowStyle.Hidden,
WorkingDirectory = Path.GetDirectoryName(fileName)
};
using var process = Process.Start(info);
process.WaitForExit();
// make sure to check process.ExitCode == 0
For me, setsid in combination with & makes a spawned child process out-living its parent process when invoked via sh -c.
Example:
var command = $"dotnet \"PathToDll\" param1 param2";
process.StartInfo = new ProcessStartInfo
{
Arguments = $"-c \"setsid {command.Replace("\"", "\\\"")} &\"",
CreateNoWindow = true,
FileName = "/bin/sh",
};
process.Start();
This was tested on Debian and Ubuntu.
Slightly refactored version of #jjxtra solution, so it's easier to understand what's going on in the arguments.
Btw, the echo is not an example, but the way of executing at command.
string command = $"actual command to run";
string atdCommand = $#"echo \""{command}\"" | at now";
string bashCommand = $#"-c ""{atdCommand}"" ";
ProcessStartInfo psi = new ProcessStartInfo
{
FileName = "/bin/bash",
Arguments = bashCommand,
...
};
Related
I am writing a code to execute some commands against wsl, parsing and reading the returned value is important.
Project is a .net core console app 3.1
wsl2 is enabled on the system
for example, listing all the available wsl images on my local machine i am using a snippet found in an answer provided in another "kind of related" SO post.
using (var proc = new Process
{
StartInfo = new ProcessStartInfo
{
FileName = #"cmd.exe",
UseShellExecute = false,
RedirectStandardOutput = true,
RedirectStandardInput = true,
CreateNoWindow = true,
}
})
{
proc.Start();
proc.StandardInput.WriteLine("wsl --list");
System.Threading.Thread.Sleep(500);
proc.StandardInput.Flush();
proc.StandardInput.Close();
proc.WaitForExit(5000);
var c = proc.StandardOutput.ReadToEnd();
Console.WriteLine(c);
Console.ReadLine();
}
now the expected output should be
what i am getting is
if i inspect using breakpoint i get this in "var c"
Ideally i want to be able to have a list that contains the 2 dockers items inside C#, changing the wait time didn't help.
in the ProcessStartInfo you have to set
StandardOutputEncoding = Encoding.Unicode;
StandardErrorEncoding = Encoding.Unicode;
for direct call of wsl use additionally:
FileName = #"wsl.exe";
Arguments = #"-l -v";
I'd like to set up a Visual C++ toolchain to be used inside my C# application. For the toolchain it's recommended to call vcvarsall (or some subvariant). My problem is that the calling process - my application - will not get to keep the environment set up by vcvarsall. Can this somehow be achieved?
// First set up the toolchain with vcvarsall
var vcvarsallProc = new Process
{
StartInfo = new ProcessStartInfo
{
FileName = vcvarsallPath,
Arguments = "x86",
UseShellExecute = false,
}
};
vcvarsallProc.Start();
vcvarsallProc.WaitForExit();
// Invoke the linker for example
var linkerProc = new Process
{
StartInfo = new ProcessStartInfo
{
FileName = "LINK.exe",
Arguments = " foo.obj /OUT:a.exe",
UseShellExecute = false,
}
};
linkerProc.Start();
linkerProc.WaitForExit();
// ERROR: 'LINK' is not recognized as an internal or external command
Turns out, this is pretty much impossible without a hassle.
The best one can do is chain multiple commands, which will run in the same environment. It can be done like so:
var linkerProc = new Process
{
StartInfo = new ProcessStartInfo
{
FileName = "cmd.exe",
Arguments = $"/C ({vcvarsallPath} {vcvarsallArgs}) && (LINK.exe {linkArgs})",
UseShellExecute = false,
}
};
Hiding this behind a function like InvokeWithEnvironment, and it's pretty painless to use.
Mostly just as a curiosity, I wrote a little app to start up Terminator shell on Windows, using Ubuntu/WSL and Xming window server.
Doing things manually from the shell, I can run Firefox, gedit, Terminator, etc on Windows, it's pretty cool.
So I checked the location of bash.exe using where bash and it returned...
C:\Windows\System32\bash.exe
However when I tried to run this code...
using (var xminProc = new Process())
{
xminProc.StartInfo.FileName = #"C:\Program Files (x86)\Xming\Xming.exe";
xminProc.StartInfo.Arguments = ":0 -clipboard -multiwindow";
xminProc.StartInfo.CreateNoWindow = true;
xminProc.Start();
}
using (var bashProc = new Process())
{
bashProc.StartInfo.FileName = #"C:\Windows\System32\bash.exe";
bashProc.StartInfo.Arguments = "-c \"export DISPLAY=:0; terminator; \"";
bashProc.StartInfo.CreateNoWindow = true;
bashProc.Start();
}
I get the error...
System.ComponentModel.Win32Exception: 'The system cannot find the file specified'
And checking my entire system for bash.exe reveals it really be in another place altogether...
I'm not sure if this location is one that I can rely on, I'm worried it's ephemeral and can change during a Windows Store update, although I may be wrong about that.
Why does the command prompt show bash.exe to be in System32 but it's really in another location altogether?
Can I get C# to also use the System32 location?
As #Biswapriyo stated first set the platafrom to x64 on your solution:
Then you may run on your ubuntu machine from c# as:
Console.WriteLine("Enter command to execute on your Ubuntu GNU/Linux");
var commandToExecute = Console.ReadLine();
// if command is null use 'ifconfig' for demo purposes
if (string.IsNullOrWhiteSpace(commandToExecute))
{
commandToExecute = "ifconfig";
}
// Execute wsl command:
using (var proc = new Process
{
StartInfo = new ProcessStartInfo
{
FileName = #"cmd.exe",
UseShellExecute = false,
RedirectStandardOutput = true,
RedirectStandardInput = true,
CreateNoWindow = true,
}
})
{
proc.Start();
proc.StandardInput.WriteLine("wsl " + commandToExecute);
System.Threading.Thread.Sleep(500); // give some time for command to execute
proc.StandardInput.Flush();
proc.StandardInput.Close();
proc.WaitForExit(5000); // wait up to 5 seconds for command to execute
Console.WriteLine(proc.StandardOutput.ReadToEnd());
Console.ReadLine();
}
I want to run a gpu accelerated python script on windows using conda environment (dlwin36).
I’m trying to activate dlwin36 and execute a script:
1) activate dlwin36
2) set KERAS_BACKEND=tensorflow
3) python myscript.py
If I manually open cmd on my machine and write:"activate dlwin36"
it works.
But when I try opening a cmd from c# I get:
“activate is not recognized as an internal or external command, operable program or batch file.”
I tried using the following methods:
Command chaining:
var start = new ProcessStartInfo();
start.FileName = "cmd.exe";
start.Arguments = "/c activate dlwin36&&set KERAS_BACKEND=tensorflow&&python myscript.py";
Process.Start(start).WaitForExit();
(I’ve tested several variations of UseShellExecute, LoadUserProfile and WorkingDirectory)
Redirect standard input:
var commandsList = new List<string>();
commandsList.Add("activate dlwin36");
commandsList.Add("set KERAS_BACKEND=tensorflow");
commandsList.Add("python myscript.py");
var start = new ProcessStartInfo();
start.FileName = "cmd.exe";
start.UseShellExecute = false;
start.RedirectStandardInput = true;
var proc = Process.Start(start);
commandsList.ForEach(command => proc.StandardInput.WriteLine(command));
(I’ve tested several variations of LoadUserProfile and WorkingDirectory)
In both cases, I got the same error.
It seems that there is a difference between manually opening cmd and opening it from c#.
The key is to run activate.bat in your cmd.exe before doing anything else.
// Set working directory and create process
var workingDirectory = Path.GetFullPath("Scripts");
var process = new Process
{
StartInfo = new ProcessStartInfo
{
FileName = "cmd.exe",
RedirectStandardInput = true,
UseShellExecute = false,
RedirectStandardOutput = true,
WorkingDirectory = workingDirectory
}
};
process.Start();
// Pass multiple commands to cmd.exe
using (var sw = process.StandardInput)
{
if (sw.BaseStream.CanWrite)
{
// Vital to activate Anaconda
sw.WriteLine("C:\\PathToAnaconda\\anaconda3\\Scripts\\activate.bat");
// Activate your environment
sw.WriteLine("activate your-environment");
// Any other commands you want to run
sw.WriteLine("set KERAS_BACKEND=tensorflow");
// run your script. You can also pass in arguments
sw.WriteLine("python YourScript.py");
}
}
// read multiple output lines
while (!process.StandardOutput.EndOfStream)
{
var line = process.StandardOutput.ReadLine();
Console.WriteLine(line);
}
You need to use the python.exe from your environment. For example:
Process proc = new Process();
proc.StartInfo.FileName = #"C:\path-to-Anaconda3\envs\tensorflow-gpu\python.exe";
or in your case:
start.Arguments = "/c activate dlwin36&&set KERAS_BACKEND=tensorflow&&\"path-to-Anaconda3\envs\tensorflow-gpu\python.exe\" myscript.py";
I spent a bit of time working on this and here's the only thing that works for me: run a batch file that will activate the conda environment and then issue the commands in python, like so. Let's call this run_script.bat:
call C:\Path-to-Anaconda\Scripts\activate.bat myenv
set KERAS_BACKEND=tensorflow
python YourScript.py
exit
(Note the use of the call keyword before we invoke the activate batch file.)
After that you can run it from C# more or less as shown above.
ProcessStartInfo start = new ProcessStartInfo();
start.FileName = "cmd.exe";
start.Arguments = "/K c:\\path_to_batch\\run_script.bat";
start.UseShellExecute = false;
start.RedirectStandardOutput = true;
start.RedirectStandardError = true;
start.WorkingDirectory = "c:\\path_to_batch";
string stdout, stderr;
using (Process process = Process.Start(start))
{
using (StreamReader reader = process.StandardOutput)
{
stdout = reader.ReadToEnd();
}
using (StreamReader reader = process.StandardError)
{
stderr = reader.ReadToEnd();
}
process.WaitForExit();
}
I am generating the batch file on the fly in C# to set the necessary parameters.
If this is gonna help anyone in the future. I found that you must run the activation from C:\ drive.
I'm trying to automate some operations in Mainframe. For that I made a C# program that connects to Mainframe using terminal emulator wc3270 and send keys to it.
This part works fine.
My problem is that I need to read the content displayed in terminal emulator screen so I can take better decisions about which keys I'll send to it.
How could I do it? I didn't find out any API that would let me do it.
Thanks.
I suggest using IBM personal communication tools:
http://www-03.ibm.com/software/products/en/pcomm
Together with the C# EHLLAPI wrapper, URL to example:
https://www.codeproject.com/Articles/9615/Using-EHLLAPI-in-C
To find out the session key for the emulator, you can use handle.exe from Windows Sysinternals
Code that im using for handle:
public String getSessionKey(String pid)
{
Process p = new Process();
p.StartInfo = new ProcessStartInfo()
{
UseShellExecute = false,
CreateNoWindow = true,
WindowStyle = System.Diagnostics.ProcessWindowStyle.Hidden,
FileName = "cmd.exe",
Arguments = "/C C:\\handle.exe -a -p " + pid,
RedirectStandardError = true,
RedirectStandardOutput = true
};
p.Start();
String str = (p.StandardOutput.ReadToEnd());
String[] arr = str.Split('\n');
foreach (String s in arr)
{
if (s.Contains("Owned"))
{
return s.Substring(s.Length - 3, 1);
}
}
return "";
}