I'm trying figure out how to compile C++ code from an executable during runtime using Visual Studio compiler under Windows.
I'll be using Visual Studio IDE to build main project into an executable and use CreateProcess to compile other C++ files and create a DLL to later load/use/unload this DLL.
I understand that one way of doing this requires setting environment variables(mainly PATH, INCLUDE and LIB) and there's a .bat file called "vcvarsall.bat" which does this.
The part I'm stuck with is the argument(s) passed to this batch file. I see that first argument is the platform with some of the options being x86, amd64, arm, etc. But how do I programmatically figure out which one of these arguments I should be using considering main executable could've been built with any one of these?
You can prepare a regular solutionfor this purpose, containing one project with a single file, and use it to compile your file easily.
Now, all you need is to reame your file to the file name in the project and compile a solution with command line. Alternatively, you can also edit the project and replace the existing filename with your file name.
To do so you need to resolve the environment variable %DevEnvDir% and run the folowing command with the platform name (x64, win32 etc.) and configuration name(Release or Debug)
like this:
%DevEnvDir%\devenv.com \path\to\yoursolution.sln /ReBuild "Release|x64"
Related
I have a Visual Studio 2019 CMake project and i need to run a postbuild script which copys a file (The file is not generated by the build process). What i did so far is add a custom command in CMake with
add_custom_command(TARGET testExec POST_BUILD COMMAND "../postbuild.bat")
The postbuild.bat would copy the file. This works great most of the time but when my build fails due to some compile error, the postbuild script won't be executed.
How can i run the postbuild script even if the build fails ? I know there is a similar question here but sadly no proper solution. If there is a way to configure a postbuild event directly inside a Visual Studio CMake project this would also be suitable, but it seems like this is not possible (because in a cmake project i don't have a project file).
Since The copied file is not generated by the build you can use PRE_BUILD. On Visual Studio Generators, it runs before any other rules are executed within the target.(https://cmake.org/cmake/help/latest/command/add_custom_command.html).
The other solution could be to use add_custom_command(OUTPUT as the file seems independent of the build.
I'm trying to take a couple of projects normally compiled on Windows with Microsoft C++ and compile them with clang instead.
On the upside, there exists clang-cl.exe which is designed to be a drop-in replacement for cl.exe. However, even when I copy clang-cl.exe into the current directory as cl.exe, msbuild still in some cases calls Microsoft's cl.exe.
Is there a way to tell msbuild 'here, when executing Task CL, use this cl.exe instead of the usual one'? msbuild's command line options don't contain anything obvious in that direction.
Also, is there a way to tell it to supply or override command line parameters for cl without changing the project file?
This is easy to do from either command line or project file. The properties you need to configure are $(CLToolExe) and $(CLToolPath).
From the command line:
msbuild MyProj.vcxproj /p:CLToolExe=clang-cl.exe /p:CLToolPath=c:\whatever\path\to\the\tool
Alternatively, inside your .vcxproj file:
<PropertyGroup>
<CLToolExe>clang-cl.exe</CLToolExe>
<CLToolPath>c:\whatever\path\to\the\tool</CLToolPath>
</PropertyGroup>
If you are calling task CL directly inside your .vcxproj file, as opposed to just relying on common targets, just set corresponding parameters ToolExe and ToolPath of the CL task.
Since Visual Studio 2019 16.2, Microsoft provide an integration of MSbuild and ClangCl. So this can be achieved by:
Installing the “C++ Clang Tools for Windows” component
Choosing the "LLVM (clang-cl)” toolset in the IDE
Microsoft's blog post has more information on this: https://devblogs.microsoft.com/cppblog/clang-llvm-support-for-msbuild-projects/
I were using #seva solution for sometime, though in Visual studio version Version 16.10.1 it works for me only if the 'CL' prefix is omitted from the command line parameters. i.e.:
msbuild MyProj.vcxproj /p:ToolExe=clang-cl.exe /p:ToolPath=c:\whatever\path\to\the\tool
I am learning how to use OpenCV and as practice I ran a program (in Release mode, x64). I had 0 compiler errors but got a pop-up screen that said:
"the program can't start because opencv_core243.dll is missing"
However, I made sure I declared the correct environment variables and specified the necessary libraries/directories. My problem was fixed when I copied the following .dll files into x64/Release:
opencv_core243.dll
opencv_highgui243.dll
opencv_imgproc243.dll
My program compiles fine now and works. However, I would like to know why. It feels cheap to copy and paste these .dll files. Did I miss a step where these .dll files would be generated automatically?
The actual solution for this problem is to append the path of opencv bin directory to the System PATH environment variable.
Have a look at this answer describing the complete procedure to install OpenCV in Visual Studio 2010.
There is a disadvantage of this approach. The prebuilt OpenCV x86 and x64 binaries have same names. So by appending the path of OpenCV to the PATH variable, you can execute either the 32 bit version or the 64 bit version at a time. To run the other version, you have to change the PATH variable.
An alternative to this, (my personal favorite) also involves copying the dlls to output directory, but this is done automatically at the end of compilation. What I do, is to create new environment variables for x86 and x64 dll paths. I create the User Variables CV_BIN32 and CV_BIN64 containing the paths of x86 and x64 dlls respectively.
After creating the User Variables, open your OpenCV project, go to Project Properties -> Build Events -> Post-Build Event -> Command Line.
Add the copy commands for the dlls you require at the runtime.
This one is for Win32 Release Configuration:
copy "$(CV_BIN32)\opencv_core243.dll" "$(OutDir)"
copy "$(CV_BIN32)\opencv_highgui243.dll" "$(OutDir)"
You can change them for all the 4 configurations, (Debug/Release),(Win32/x64)
Now when the project build procedure completes, the specified dlls will be copied to the output directory automatically, and the error will not be shown.
In VS 2008 and VS 2010, one can easily create a solution and modify the "Solution Configuration". We can choose what configuration each project gets built in when we trigger a build at the solution level.
Is such a facility available in the Visual Studio 6.0?
In my experience:
when a configuration is chosen (form the list available) in VS6 for a VC++ project, the dependencies (which themselves have multiple configurations defined) get built in some random order. There is no way to control the configurations of dependencies at build time.
"Batch Build" does come close to this but is not as flexible for my purpose.
I have tried various options in the VS6.
Hope I am clear.
Here is a link on the MSDEV command line.
https://msdn.microsoft.com/en-us/library/aa699274(v=vs.60).aspx
There is a way to control the building of dependencies. Specify /NORECURSE and dependencies will not be built.
I use /REBUILD with /NORECURSE to keep the dependencies from getting built.
And I build each project one at a time inside the workspace in a bat file by doing a chdir to the subdirectory and calling MSDEV just for that subproject:
msdev myproject.dsp /MAKE "myproject - Win32 Debug" /REBUILD /NORECURSE > Build.log
Then I cd to the next project directory one at a time.
On a side note, I had difficulties for several years where NMAKE would not work for my specific tasks. Turns out that the PATH environment variable inside MSDEV (Visual Studio 6.0) is different from the PATH environment variable of a command shell you would run NMAKE on.
The Path used by the MSDEV shell is the %PATH% at the time Visual Studio 6 was installed. We use this and poke the registry as needed for MSDEV to get the correct path setup when switching revisions of our software; however this doesn't help update the %PATH%. The MSDEV path can be queried with a query script. I don't have my example handy.
That is why builds in MSDEV sometimes work when builds using the command line don't, as the path to DLLs differ, and any custom build steps that run .exe will not work outside of the MSDEV environment unless the path is updated.
I have a script somewhere that reads the queries the registry to extract the MSDEV path and update PATH of a shell so that batch scripts doing nmake will work as they would inside the MSDEV shell environment. The problem with the REGISTRY QUERY is that the return arguments differ with different flavors of Windows (XP/SERVER2003/...).
One thing I just discovered is that Incredibuild works with the old VS6.0 MSDEV IDE. This is a game changer. It distributes builds. I'm evaluating it now, but it might be useful to anyone waiting for long VS6.0 builds.
What is the need of Make File? IN which situation we have to alter that file for Brew Application?
The makefile is usually used to build for the device (ARM) build of the project (with visual studio being used to build for the emulator). Visual Studio will run the makefile when you do the device build.
Small addition - makefiles used in Visual Studio only for MOD1 application, whereas they aren't used for compiling MOD apps.
As joseph said makefiles are used to build the source files in linux projects, advantage of makefile is you can write a general script irrespective of dependency files. and even if you change a single file in your file then you dont have to compile the whole source code again. only the file which is changed is compiled again and the files which are dependent on them are compiled. you can check out tutorial on the makefile at http://sagarsakre.blogspot.in/2012/09/understanding-makefile-for-beginners.html .....