I have a number of native C++ libraries (Win32, without MFC) compiling under Visual Studio 2005, and used in a number of solutions.
I'd like to be able to choose to compile and link them as either static libraries or DLLs, depending on the needs of the particular solution in which I'm using them.
What's the best way to do this? I've considered these approaches:
1. Multiple project files
Example: "foo_static.vcproj" vs "foo_dll.vcproj"
Pro: easy to generate for new libraries, not too much manual vcproj munging.
Con: settings, file lists, etc. in two places get out of sync too easily.
2. Single project file, multiple configurations
Example: "Debug | Win32" vs "Debug DLL | Win32", etc.
Pro: file lists are easier to keep in sync; compilation options are somewhat easier to keep in sync
Con: I build for both Win32 and Smart Device targets, so I already have multiple configurations; I don't want to make my combinatorial explosion worse ("Static library for FooPhone | WinMobile 6", "Dynamic library for FooPhone | WinMobile 6", "Static library for BarPda | WinMobile 6", etc.
Worse Con: VS 2005 has a bad habit of assuming that if you have a configuration defined for platform "Foo", then you really need it for all other platforms in your solution, and haphazardly inserts all permutations of configuration/platform configurations all over the affected vcproj files, whether valid or not. (Bug filed with MS; closed as WONTFIX.)
3. Single project file, selecting static or dynamic via vsprops files
Example: store the appropriate vcproj fragments in property sheet files, then apply the "FooApp Static Library" property sheet to config/platform combinations when you want static libs, and apply the "FooApp DLL" property sheet when you want DLLs.
Pros: This is what I really want to do!
Cons: It doesn't seem possible. It seems that the .vcproj attribute that switches between static and dynamic libraries (the ConfigurationType attribute of the Configuration element) isn't overrideable by the .vsprops file. Microsoft's published schema for these files lists only <Tool> and <UserMacro> elements.
EDIT: In case someone suggests it, I've also tried a more "clever" version of #3, in which I define a .vsprops containing a UserMacro called "ModuleConfigurationType" with a value of either "2" (DLL) or "4" (static library), and changed the configuration in the .vcproj to have ConfigurationType="$(ModuleConfigurationType)". Visual Studio silently and without warning removes the attribute and replaces it with ConfigurationType="1". So helpful!
Am I missing a better solution?
I may have missed something, but why can't you define the DLL project with no files, and just have it link the lib created by the other project?
And, with respect to settings, you can factor them out in vsprop files...
There is an easy way to create both static and dll lib versions in one project.
Create your dll project. Then do the following to it:
Simply create an nmake makefile or .bat file that runs the lib tool.
Basically, this is just this:
lib /NOLOGO /OUT:<your_lib_pathname> #<<
<list_all_of_your_obj_paths_here>
<<
Then, in your project, add a Post Build Event where the command just runs the .bat file (or nmake or perl). Then, you will always get both a dll and a static lib.
I'll refrain from denigrating visual studio for not allowing the tool for this to exist in a project just before Linker (in the tool flow).
I think the typical way this is done is choice 2 above. It is what I use and what I have seen done by a number of libraries and companies.
If you find it does not work for you then by all means use something else.
Good luck.
I prefer 2 configurations way.
Setup all common settings via 'All configurations' item in a project properties windows. After it separated settings. And it's done. Let's go coding.
Also there is very good feature named 'Batch build', which builds specified configurations by turn.
Multiple projects are the best way to go - this is the configuration i have most widely seen in umpteen no of projects that i have come across.
That said, it might be also possible to implement the third option by modifying your vcproj files on the fly from external tools(like a custom vbscript), that you could invoke from a make file. You can use shell variables to control the behavior of the tool.
Note that you should still use use visual studio to make the build, the makefile should only launch your external tool if required to make the mods and then follow that by the actual build command
I use Visual Studio 6.0 (Still) due to issues that are preventing us from Migrating to VS2005 or newer. Rebuilding causes severe issues (everything breaks)... so many of us are considering lobbying a migration to GnuC++ moving forward in a structured way to eventually get us off of licensed Visual Studio products and onto Eclipse and Linux.
In Unix/Linux it is easy to build for all configurations.. so I can't believe what a time and productivity sink it is to try and accomplish the same task in Visual Studio. For VS6.0 I have so far found that only having two separate projects seems to be workable. I haven't yet tried the multiple configuration technique, but will see if it works in the older VS6.0.
Why not go for version 1 and generate the second set of project files from the first using a script or something. That way you know that the differences are JUST the pieces required to build a dll or static lib.
Related
I am developing a project on C++ which relies on many of third-party libraries (*.lib files and *.h files). I store these libraries in a folder which is not dependant to project, like C:/thirdpartylib. Relative paths is not an option, since it becomes way too long. I have defined connections to libraries in linker setting and in general C++ settings.
But when I pass the project to supervisor he has to reset all paths to libraries to match his environment. We use git, and the project file is being tracked. He stores thirdparty libraries in another way than me.
Is there any way to make a project more portable? Maybe it is possible to store paths in some sort of config files?
As #gaurav says, the way to deal with this in Visual Studio is with property sheets. It's unfortunate that this term is used for two different things in VS, but I guess they just ran out of names (spoiler alert).
These are very powerful, once you learn how they work, and they're just what you need here because they let you define macros, and these macros can in turn be used in the rest of your project to refer to the (volatile) location of your various libraries. This is a trick that everyone who uses VS should know, but it seems that a lot of people don't.
I don't think it's worth me trying to walk you through the mechanics of setting one up here because Microsoft already document it in the Visual Studio help file. Suffice to say, you do it in the Property Manager, that should help you track down the relevant information.
There is also an excellent blog post here which I recommend you read before you do anything else:
http://www.dorodnic.com/blog/2014/03/20/visual-studio-macros/
It's also on Wayback Machine here:
https://web.archive.org/web/20171203113027/http://www.dorodnic.com/blog/2014/03/20/visual-studio-macros/
OK, so now we know how to define a macro, what can we do with it?
Well, that's actually the easy part. If we have a macro called, say, FOO, then wherever we want to expand that macro in some project setting or other we can just use $(FOO). There's also a bunch of macros built into the IDE as listed here:
https://msdn.microsoft.com/en-us/library/c02as0cs.aspx
So, you, I imagine, will want to define macros for the include and lib directories for each of your external libraries and you can then use these to replace the hard-coded paths you are currently using in your project.
And that, I reckon, should sort you out, because the definitions of the macros themselves are stored in a separate file, external to your project file, and different users / build machines can use different files. IIRC, these have extension .props.
Also, you can define a macro in terms of another macro or macros, and that makes the job easier still.
So, who still thinks that Microsoft don't know how to create a build system? Visual Studio is a fantastic piece of software once you get used to it, there's just a bit of a learning curve.
The way to go for large project is to use a package manager. There are some good options out there. Perhaps in windows and visual studio you can use vcpkg or NuGet unmanaged.
If you cannot use a package manager for some reason, the next thing to do is to commit all the dependencies to the GIT repo. If you only target windows platforms like windows 8 or 10 and want to support only VS2017 then committing the compiled dependencies is not a problem. The downside is that the repo will become huge.
For a tiny school project the latter option is viable.
Background
Our build uses ant and a custom task to build Visual Studio projects/solutions as well as some Java projects. There structure is basically a large tree and artifacts from the projects are typically copied upwards to a common build directory.
This was previously a complete mess and I've greatly simplified the ant scripts and now I'm most of the way through the Visual Studio projects/solutions. These projects are extremely old and have been upgraded through every version of Visual Studio up to 2013. Part the changes I have done was to use as many of the default project properties and macros as possible. Most of these used to be hard coded.
Although I have modified the projects to use the $(Configuration) macro to separate the artifacts from different configurations these still get copied to a common location for other dependent projects in other solutions. So to avoid confusion and actually make sure our debug builds link to all the debug libraries (which wasn't happening previously) I have been adding suffixes to the Target Name. E.g. The Target Name of a Debug Unicode build will be $(ProjectName)DU.
Problem
This has been great so far but now I'm not sure how to complete these changes for one of our COM libraries. This library has an IDL file and the MIDL compiler generates a TLB file. Maybe this isn't a good way to do it but for now I wanted the TLB file to also have a different suffix depending on the build. The problem is, when I change the Type Library property for the MIDL configuration then this breaks the Compile-time directives in the RC file. I figured it might be possible to use #ifdefs in the TEXTINCLUDE block depending on whether _UNICODE or _DEBUG are set (provided I do it via the Resource Includes dialog so I don't break the RC file). It also means that there are other importlib attributes that will also need #ifdef checks.
At the moment it kinda works without renaming the TLB files but that's only because they are currently only being used within this solution.
Has anyone ever done anything like this or know a better solution?
Update
I guess what I really need to know here is what is the best way to use types from one COM DLL in another one? Should I even be using importlib? The MSDN documentation says that in most cases you should be using the import instead. I tried this but broke a whole bunch of stuff.
I am trying to import an existing c++ application's source into visual studio to take advantage of some specific MS tools. However, after searching online and playing with visual studio, I cannot seem to find an easy way to import existing c++ source code into visual studio and keep it structurally intact.
The import capacity I did find flattens out the directories and puts them all into one project. Am I missing something?
(This is all unmanaged C++, and contains specific builds for win/unix)
With no project/solution loaded, in Visual Studio 2005 I see this menu item:
File > New Project From Existing Code...
After following the wizard, my problem is solved!
Switching the "Show All Files" button shows the complete hierarchy with all directories and files within.
If the New Project From Existing Code... option isn't available, you'll need to add it in Tools > Customize...
I am not aware of any general solution under the constraints given - specifically having to create many projects from a source tree.
The best option I see is actually creating the project files by some script.
Creating a single project manually (create empty project, then add the files),
Configure it as close as possible as desired (i.e. with precompiled headers, build configurations, etc.)
Use the .vcproj created as skeleton for the project files to be created
A very simple method would file list, project name etc. with "strange tokens", and fill them in with your generator. If you want to be the good guy, you can of course use some XML handling library.
Our experience: We actually don't store the .vcproj and .sln in the repository (git) anymore, but a python script that re-genrates them from the source tree, together with VS 2008 "property sheet templates" (or whatever they are called). This helps a lot making general adjustments.
The project generation script contains information about all the projects specialties (e.g. do they use MFC/ATL, will it create DLL or an EXE, files to exclude).
In addition, this script also contains dependencies, which feeds the actual build script.
This works quite well, the problems are minor: python requried in build systems, not forgetting to re-gen the project files, me having to learn some python to make adjustments to some projects.
#Michael Burr "How complex are the python scripts and whatever supporting 'templates' you might need?"
I honestly can't tell, since I gave the task to another dev (who picked python). The original task was to provide a build script, as the VS2008 solution build was not good enough for our needs, and the old batch file didn't support parallelization. .vcproj generation was added later. As I understand his script generates the .vcproj and .sln files from scratch, but pulls in all the settings from separate property sheets.
Pros:
Adding new configurations on the fly. Some of the projects already had six configurations, and planning for unicode support meant considering doubling them for a while. Some awkward tools still build as MBCS, so some libs do have 8 configs now. Configuring that from hand is a pain, now it just doesn't bother me anymore.
Global changes, e.g. moving around relative project paths, the folder for temp files and for final binaries until we found a solution we were happy with
Build Stability. Merging VC6 project files was a notable source of errors for various reasons, and VC9 project files didn't look better. Now things seem isolated better: compile/link settings in the property sheets, file handling in the script. Also, the script mostly lists variations from our default, ending up easier to read than a project file.
Generally: I don't see a big benefit when your projects are already set up, they are rather stable, and you don't have real issues. However, when moving into the unknown (for us: mostly VC6 -> VC9 and Unicode builds), the flexibility reduced the risk of experiments greatly.
Create a new empty solution and add your source code to it.
For example,
File>New>Project...
Visual C++>Win32>Win32 Console Application
Application Settings>
- Uncheck "Precompiled Header"
- Check "Empty Project"
Project is then created. To add existing code:
Project>Add Existing Item...>
- Select file(s) to add
Recompile, done!
In the "Solution Explorer" you can click on the "Show All Files" button to have Visual Studio display the files as they exist on the file system (directories and all).
In my opinion this is an imperfect workaround, but I believe it's the best available. I'm unaware of a plug-in, macro or other tool that'll import a directory into an actual project with folders that mirror the file system's.
I know this question is already marked correct, but I was able to import existing code into a project with Visual Studio 2008 by doing "File" -> "New Project from existing code". The directory structure of my code was retained.
You can always switch view from project menu
For eg. Project->Show All Files
The above will display the files in unformated raw file system order
Not sure of older versions but it works on VS 2010
I understand you, I have the same problem: many .cpp and .h files organized in many folders and subfolders with include paths written for this folder structure. The only way you can do to import this folder structure together with the source files is to use "Show All Files" and then right-click on folders and select "Import in Project". This works for me when I am using C-Sharp projects. But it does not work for my C++ Projects. I am still searching for a solution...
So I have main.cpp and main2.cpp with int main in each. I want to get 2 exes out of it. Is it possible and what would be instruction to create such project?
Nope, Visual Studio's project model is rigidly built around the assumption that "one project generates one output".
If you need two executables, you have to create two projects. You can keep them in the same solution to make things easier for yourself, but they have to be separate projects.
Edit
Ok, as other answers have pointed out, it can of course be done, if you're desperate. You can add a custom build step, which does anything you like, including building another executable. (However, the build system won't understand that this file should be considered project output, and so some scenarios may fail: for example the file won't be automatically copied to the output folder, and when checking dependencies before a rebuild, it might not be able to understand which files to check, and what (or how) to rebuild.)
Visual Studio (at least up to 2008, not sure about 2010) also allows the use of nmake files, but then I think you're stretching the definition of "a Visual Studio project".
But under "normal" circumstances, one project implies one output. And in order to get two executables, you'd normally create two projects.
You need a solution which includes two projects. Have a read of the Visual Studio documentation on solutions and projects.
Here's my solution, since nobody in a Google search seems to suggest this method. It's quite simple and I've used/seen it used in other IDEs (like Code::Blocks).
Within your project, create a configuration for each output that you want. Then, only include one main source file in each configuration.
In VS, this means for each source file with main: right-click -> Properties -> Excluded From Build = Yes. So, once you're done, only one main source is built for each configuration. You can then specify a different output for each configuration in the Project Properties. I did this on VS 2010, but it should probably work with other versions.
I'm using this approach so that I can have several tests for one project, without cluttering the solution with more test projects than actual code projects.
I don't know if it can be done ,but the only change you have ,to do this ,is with custom build step.
EDIT: Since someone downvoted this ,i did a test making a dummy configuration.
In the custom build step I two Link-cmds (copied form original link-cmdline and modified it a bit) taking as input main1.obj resp. main2.obj and outputting App1.exe resp. App2.exe.
It's executed after Compiling and before linking.
It worked !
The downside is I cannot prevent (yet) the linking ot the orinal exe (which errors on duplicate main function).
Solution to this could be to have a lib project excluding the sources with main()from build and build them in the custum-step too.
So the answer to the question should : Yes ,it can be done!
You can't have more than one main() function in a single visual studio project. So you can't compile 2 executables, the only way is to make two different project in the same solution
You can create a solution with two project one for each output you want. Then head to Build menu and select Batch Build.. and select both projects.
If you want both exe files to be in one place you can specify a custom Post-build action:
For both project:
view the project properties page and in Build events select Post-Build events, then in the Command line field enter the command that will copy the output to the location you want, something like:
copy $(TargetPath) c:\yourlocation /Y
Then after you build the solution the output files will be copied to that location.
Another option you have is to use conditional compilation with sth like
main()
{
#ifdef VERSION1
call_main_logic();
#else
call_main2_logic();
#endif
}
and then define different configurations within the project. For each configuration you will then need to define preprocessor symbols appropriately (in: project settings -> configuration properties -> C/C++ -> preprocessor).
Please note, that it will be only one executable created at a time, but with switching configurations you'll get the one that does what you want at the moment.
... which may suit your needs or not, depending on more specific situation that you are in.
Edit: In addition, since different configurations create their separate output folders, you will have your both execs as outputs.
I have a few things that I cannot find a good way to perform in Visual Studio:
Pre-build step invokes a code generator that generates some source files which are later compiled. This can be solved to a limited extent by adding blank files to the project (which are later replaced with real generated files), but it does not work if I don't know names and/or the number of auto-generated source files. I can easily solve it in GNU make using $(wildcard generated/*.c). How can I do something similar with Visual Studio?
Can I prevent pre-build/post-build event running if the files do not need to be modified ("make" behaviour)? The current workaround is to write a wrapper script that will check timestamps for me, which works, but is a bit clunky.
What is a good way to locate external libraries and headers installed outside of VS? In *nix case, they would normally be installed in the system paths, or located with autoconf. I suppose I can specify paths with user-defined macros in project settings, but where is a good place to put these macros so they can be easily found and adjusted?
Just to be clear, I am aware that better Windows build systems exist (CMake, SCons), but they usually generate VS project files themselves, and I need to integrate this project into existing VS build system, so it is desirable that I have just plain VS project files, not generated ones.
If you need make behavior and are used to it, you can create visual studio makefile projects and include them in your project.
If you want less clunky, you can write visual studio macros and custom build events and tie them to specific build callbacks / hooks.
You can try something like workspacewhiz which will let you setup environment variables for your project, in a file format that can be checked in. Then users can alter them locally.
I've gone through this exact problem and I did get it working using Custom Build Rules.
But it was always a pain and worked poorly. I abandoned visual studio and went with a Makefile system using cygwin. Much better now.
cl.exe is the name of the VS compiler.
Update: I recently switched to using cmake, which comes with its own problems, and cmake can generate a visual studio solution. This seems to work well.
Specifically for #3, I use property pages to designate 3rd party library location settings (include paths, link paths, etc.). You can use User Macros from a parent or higher level property sheet to designate the starting point for the libraries themselves (if they are in a common root location), and then define individual sheets for each library using the base path macro. It's not automatic, but it is easy to maintain, and every developer can have a different root directory if necessary (it is in our environment).
One downside of this approach is that the include paths constructed this way are not included in the search paths for Visual Studio (unless you duplicate the definitions in the Projects and Directories settings for VS). I spoke to some MS people at PDC08 about getting this fixed for VS2010, and improving the interface in general, but no solid promises from them.
(1). I don't know a simple answer to this, but there are workarounds:
1a. If content of generated files does not clash (i.e. there is no common static identifiers etc.), you can add to the project a single file, such as AllGeneratedFiles.c, and modify your generator to append a #include "generated/file.c" to this file when it produces generated/file.c.
1b. Or you can create a separate makefile-based project for generated files and build them using nmake.
(2). Use a custom build rule instead of post-build event. You can add a custom build rule by right-clicking on the project name in the Solution Explorer and selecting Custom Build Rules.
(3). There is no standard way of doing this; it has to be defined on a per-project basis. One approach is to use environment variables to locate external dependencies. You can then use those environment variables in project properties. Add a readme.txt describing required tools and libraries and corresponding environment variables which the user has to set, and it should be easy enough for anyone to set up.
Depending on exactly what you are trying to do, you can sometimes have some luck with using a custom build step and setting your dependencies properly. It may be helpful to put all the generated code into its own project and then have your main project depend on it.