At work, i'm working on a large Visual Studio c++ project with several thousands source files, which take 2 hours + to build. To pull third party libraries in, we use scripts and conan, which tend to copy over header files, even when they didn't change. Any time this happens Visual Studio starts to rebuild everything from scratch, which is quite annoying.
In linux, it's usually sufficient to touch all object (.o) files and let make/cmake/whatever directly go to the linker stage, which may take some few minutes.
How could i do this in Visual Studio? When i touch all .obj files it will still recompile everything. It seems like the build state is stored elsewhere, and the time stamps on the .obj files are disregarded. How can i override Visual Studio's up-to-date check, and let it (for instance) just execute the link.exe?
In the build directory, i start up a cygwin shell and execute find . -name '*.obj' | xargs touch. The object files now have a newer timestamp than all the source and header files. I was expecting Visual Studio would just try to link the object files to produce the build targets, instead it still starts to compile everything.
Related
I want to version my visual studio 2010 debug builds. I'm going to use git. Right now all the project files are in 'project' and the output binaries are put by visual studio in the directory 'project\debug'. I've read a few questions on here that have ignore lists for Visual Studio but most of them have an ignore list that ignores the binary files, like the obj exe idb pdb ilk etc.
If I want to be able to debug my exe in the future what files do I need to keep? Is there any reason to keep the obj files? Does VS use them at all during debugging or tracing a crash? I thought I should keep at least the the pdb files and the exe. So that would be project.exe, project.pdb and vc100.pdb. But what about vc100.idb, project.ilk, project.res... etc. Is there an ignore list that will ignore everything that isn't needed to keep binary builds but keep everything that is?
I want to make sure if a user sends me a crash dump or reports a problem at some point in the future that I can go back to the version that was being used and be able to debug it. I had also hoped to make the binaries and debug files available in a remote repo so that anyone else could as well. Thanks!
Don't keep binaries in version control. Instead, build from source.
Since the binaries are built from source, and you have the source versioned already, just re-build the particular version as-needed. It will both keep your version control system lean and fast, as well as not cause merge headaches (which binaries are notorious for).
So, the answer is don't keep any binaries in version control.
If you feel you really must keep them, however, then just zip them up and store them like normal files using the tag of the build they came from.
I'm going to keep project.exe and project.pdb in the branch. The other files like *.obj, *.ilk, and *.sdf are not needed. Neither are the vc100.* files. I made my decision after speaking to a number of people in person and reviewing this:
File Types Created for Visual C++ Projects
and this:
Which Visual C++ file types should be committed to version control?
Although that SO question/answer does not answer my question it gave me a lot of good information. I also looked through my project.pdb for all the file references to confirm which files in my project that I needed to keep around. I used this command from the debugging tools for windows:
srcsrv\srctool -r project.pdb
My dev environment at work consists of a Visual Studio 2005 Solution with many sub-projects (by "project" I always mean VS project). Some of the projects build libraries which are used by other projects. By convention, a fair amount of test-related code ends up in header files which end up getting modified frequently. I've noticed that when I hit F7 to Build the solution, Visual Studio does not detect changes to header files that are in library projects. It will report that everything is up to date when it's not. To force it to rebuild the libary, I have to change (touch) one of the .c files in that particular project, or do Rebuild All which is quite slow.
Is there something I can change in the Solution or project settings to change this behavior so Build works as expected? I've actually gone so far as to hack together a script that "touches" one of the library .c files in a library when it detects an .h file has been updated, but there has got to be a VS solution to this.
Are the header files actually members of the library project - not just in an include file search path?
I've written a VS addin which intercepts Visual Studio's build command and uses another build system to do a build. I've got my build showing errors in the right format so that you can click on them in VS but the one step remaining for completely seamless integration is to prevent VS's "run" or "debug" commands from whining the it doesn't think the project is built (when of course it is) does anyone know how I can trick VS into thinking the project is built?
Shortly, MSBuild / Visual Studio checks what is output of defined targets in configuration of particular project. Next it compares timestamp of file(s) specified as target output with timestamps of all files specified as input for that particular target. If input is more up-to-date than output, then it requests to regenerate output (re-compile source code or re-link object files, etc.)
How I can trick VS into thinking the project is built?
Update timestamp of input files, namely source files (.cpp, .cs) or binary files like .obj used as input for linker and other files that are of your interest.
Sometimes I do it manually issuing the following command using touch utility from GnuWIn32
touch myfile.obj
Visual Studio uses the date/time stamp of the source files. If the source file is more up-to-date then the compiled binary located in bin\debug or bin\release, Visual Studio will rebuild the solution.
Sometimes, the rebuild process can fail, the quick solution is to delete the intermediate debug files *.pdb within the solution's bin\debug or bin\release folder. Then issue a build on the solution.
Hope this helps,
Best regards,
Tom.
Google reveals many tools for taking Visual Studio format sln/vcproj files, and producing a make file from them. But I can't find one that solves the opposite problem - I have a make file that references hundreds of .c and .h files and (for convenience, for debugging, for writing code in the VS IDE) would like to open it as a Visual Studio project.
Where can I find a tool to take an arbitrary make file as input, and produce Visual Studio project/solution files as output?
Makefile Project Wizard
You might be able to find a converter for a well-constrained set of makefiles, but a converter for any arbitrary makefile would be tricky. They are mini-programs, after all, that would have to be evaluated. And not all makefile concepts map directly to Visual Studio projects.
If you only have one project to worry about, I'd just manually put together a project in Visual Studio (tip: you can select more than one file in the "Add existing file..." dialog). If you do this regularly, perhaps look into a tool like Premake or CMake to help automate the generation of the projects (and if necessary, a new Makefile) for you.
I have a (C++) project that I originally developed under Linux using make to build it. I would like to also have it run in Windows and am using Visual Studio 2005 to build it. The problem I'm running into is that Visual Studio places all objects into the same output directory. This doesn't work for me because I have source files with the same name in different sub-directories. This means that the last object (with the common name) overwrites all previous ones and thus I get link errors for the "missing" code.
Setting the output directory (even using the Visual Studio variables like $(InputDir)) at the project level doesn't work because Visual Studio passes all of the source files to cl.exe at once (i.e. $(InputDir) is evaluated once rather than for each input file). It appears that this problem can be solved by manually setting the output directory for each file (at a minimum, for the affected files), but this is less than optimal and error-prone.
Is there a better way to solve this problem? I would even appreciate it even if someone is able to suggest a way to get Visual Studio to compile files one-at-a-time so that setting the output directory at the project level would have the desired effect.
You might consider using different project for each directory or so. Otherwise, using the exactly same filename within a certain project might seem a bit strange. (Sort of a hierarchy within the project structure; something I've never seen before anyway.)