I am using 64-bit Vim on windows, this version by Haroogan:
+python27
+python33
+huge
I am trying to use clang_complete, so I took the libclang.dll
library here and I set up my vimrc
correctly. libclang.dll
is found.
However when I open a *.cpp
file I now have a message:
libclang
cannot find the builtin includes this will cause slow
completion code
but I have no completion at all...
I found this post with the guy who build the version of Vim that I use, but there is no clear instruction about what to do. Can anybody help here?
UPDATE
Here is what I did:
- Took Vim YouCompleteMe for Windows
- Took LLVM for Windows
- Took the liclang.dll from it and stuck it in YCM python directory
- Took .ycm_extra_conf.py and stuck it with my cpp file
It almost work (I do not have any error message), but I still get:
clang_complete
Thanks for using Vim for Windows, hope you enjoy.
To solve your problem you have to utilize .clang_complete
file. Read more about it in :h clang_complete
. In brief, you can put this file into the root directory of any of your projects (i.e. for each project this file can be different, what perfectly makes sense, since different projects have different toolchain configurations). Here is the sample for MinGW-w64 toolchain:
-ID:/Toolchains/x64/MinGW-w64/4.8.1/lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++
-ID:/Toolchains/x64/MinGW-w64/4.8.1/lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++/x86_64-w64-mingw32
-ID:/Toolchains/x64/MinGW-w64/4.8.1/lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++/backward
-ID:/Toolchains/x64/MinGW-w64/4.8.1/lib/gcc/x86_64-w64-mingw32/4.8.1/include
-ID:/Toolchains/x64/MinGW-w64/4.8.1/lib/gcc/x86_64-w64-mingw32/4.8.1/include-fixed
-ID:/Toolchains/x64/MinGW-w64/4.8.1/x86_64-w64-mingw32/include
-ID:/Libraries/x64/MinGW-w64/4.8.1/Boost/1.54.0/include
-ID:/Libraries/x64/MinGW-w64/4.8.1/Qt/4.8.5/include
-ID:/Libraries/x64/MinGW-w64/4.8.1/Eigen/3.1.3/include
"-ID:/Libraries/x64/MinGW-w64/4.8.1/Example with Spaces/0.0.1/include"
-std=c++11
-DUNICODE
When you are editing some file in the project with Vim, clang_complete
traverses backward all the parent directories of the edited file until it stumbles across the first .clang_complete
file to read. Then it reads all these flags/switches/definitions and uses them during the invocation of libclang
for completion.
YouCompleteMe
These days, I don't use clang_complete
anymore. There is more powerful semantic autocompletion plugin for Vim our there. It's YouCompleteMe. I highly recommend you try it out. For semantic completion of C-family languages (C/C++/Objective-C/Objective-C++) it uses libclang
as well. It relies on a robust C++ back end, and is therefore incredibly fast. It has great integration with syntastic (another must have plugin for Vim). It has an ability to jump to definitions, and so on.
Since it's written in C++ and glued to Vim via Python, you'd have to compile the C++ back end. To ease the pain you can download prebuilt and ready to use YCM plugin from my Vim YouCompleteMe for Windows. I've built it for both x86 and x64 architectures. The native component is called ycm_core.pyd
. As usual the architecture of the Vim build you chose has to match the YCM build (i.e. ycm_core.pyd
). YCM can work ONLY with Python 2 (not 3), so just make sure that you have Python 2 DLL (e.g. python27.dll
) and Python 2 Interpreter (python.exe
) in the PATH
environment variable.
If you need LLVM/Clang, you can download it from me as well: LLVM for Windows. Again, just make sure that you have libclang.dll
in the PATH
environment variable (recommended) OR right next to ycm_core.pyd
. Once again both x86 and x64 architectures are supported, and once again the architecture should match both Vim's and YCM's ones.
Concerning completion:
Should he magically find where is the STL?
Of course not! This is just an autocompletion system based on Clang front end. How is it supposed to know which toolchain you're currently using to compile your code? You could use anything: Visual C++, Borland C++, GCC, MinGW, MinGW-w64, LLVM/Clang, etc. Each of them has their own standard library and runtime supplied. Thus, in each case you'd have to specify all the paths which your current toolchain uses to find standard includes.
For example, in case of GCC, MinGW, MinGW-w64, you can run the following in POSIX shell:
g++ -E -x c++ - -v < /dev/null
Or in Windows Command Prompt:
g++ -E -x c++ - -v < nul
And look for the following excerpt:
#include <...> search starts here:
d:\toolchains\x64\mingw-w64\4.8.1\posix\seh\bin\../lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++
d:\toolchains\x64\mingw-w64\4.8.1\posix\seh\bin\../lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++/x86_64-w64-mingw32
d:\toolchains\x64\mingw-w64\4.8.1\posix\seh\bin\../lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++/backward
d:\toolchains\x64\mingw-w64\4.8.1\posix\seh\bin\../lib/gcc/x86_64-w64-mingw32/4.8.1/include
d:\toolchains\x64\mingw-w64\4.8.1\posix\seh\bin\../lib/gcc/x86_64-w64-mingw32/4.8.1/include-fixed
d:\toolchains\x64\mingw-w64\4.8.1\posix\seh\bin\../lib/gcc/x86_64-w64-mingw32/4.8.1/../../../../x86_64-w64-mingw32/include
End of search list.
This tells you all the paths to standard includes which GCC, MinGW, MinGW-w64 use implicitly during compilation of your code.
For LLVM/Clang, you can do the same:
clang++ -E -x c++ - -v < /dev/null
Or:
clang++ -E -x c++ - -v < nul
When you know the paths, you can happily add them into .ycm_extra_conf.py
. For instance, in my case:
'-I',
'D:/Toolchains/x64/MinGW-w64/4.8.1/POSIX/SEH/lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++',
'-I',
'D:/Toolchains/x64/MinGW-w64/4.8.1/POSIX/SEH/lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++/x86_64-w64-mingw32',
'-I',
'D:/Toolchains/x64/MinGW-w64/4.8.1/POSIX/SEH/lib/gcc/x86_64-w64-mingw32/4.8.1/include/c++/backward',
'-I',
'D:/Toolchains/x64/MinGW-w64/4.8.1/POSIX/SEH/lib/gcc/x86_64-w64-mingw32/4.8.1/include',
'-I',
'D:/Toolchains/x64/MinGW-w64/4.8.1/POSIX/SEH/lib/gcc/x86_64-w64-mingw32/4.8.1/include-fixed',
'-I',
'D:/Toolchains/x64/MinGW-w64/4.8.1/POSIX/SEH/x86_64-w64-mingw32/include',