Parsing source file with libclang - problems with linking include files

Question

I've written a script to parse C/C++ files using the libclang library.

The file I was trying to parse: osf_sys.c . The program prints each function/variable etc. in the file. Sample output:

...
CURSORKIND: 6; __user at line: 469
CURSORKIND: 43; struct proplistname_args at line: 469
CURSORKIND: 6; nbytes at line: 470
...

So I run my script ./script osf_sys.c and parsing mostly works. However, I notice that certain parts of osf_sys.c were not being printed (missing variables and whatnot) and I suspect it had to do with the header files (which were located in a separate directory). I added a code snippet to help with debugging / printing of the error messages shown below.

Attempt 1

After building the script with CMake , I proceed to run it. This is the error that is printed:

/Linux_Kernel/linux/arch/alpha/kernel/osf_sys.c:13:10:
fatal error: 'linux/errno.h' file not found

It seems like it is having some issues locating the header files of this osf_sys.c . So I added this to CMakeLists.txt (which is probably wrong since the problem doesn't seem to be with the compilation of my program but with how libclang locates stuff):

-I/Linux_Kernel/linux/include

Note that the full path of errno.h is /Linux_Kernel/linux/include/linux/errno.h

Attempt 2

I decided to drag the entire include/ folder into the same directory that osf_sys.c resides in.

Now, running it gives me a new error printout:

/Linux_Kernel/linux/arch/alpha/kernel/osf_sys.c:13:10: error:
'linux/errno.h' file not found with <angled> include; use "quotes" instead

followed by:

/Linux_Kernel/linux/arch/alpha/kernel/linux/errno.h:4:10:
fatal error: 'asm/errno.h' file not found

Okay, now it seems to be reading something. Sure enough, the line in question is #include <linux/errno.h> from the osf_sys.c file. The second error message occurs presumably because I didn't drag the asm/ folder into the same working directory as osf_sys.c .

Attempt 3

So, as requested, I manually edit osf_sys.c to change #include <linux/errno.h> to #include "linux.errno.h"

This time, I'm left with the following error message:

/Linux_Kernel/linux/arch/alpha/kernel/linux/errno.h:4:10: 
fatal error: 'asm/errno.h' file not found

Attempt 4

Passing -I/Linux_Kernel/linux/include as an argument into the script. The script uses the following parameters for the CXTranslationUnit :

CXTranslationUnit translationUnit = clang_parseTranslationUnit(index, nullptr, argv, argc, 0, 0, CXTranslationUnit_KeepGoing);

This yields other kinds of error messages:

warning: 'linker' input unused [-Wunused-command-line-argument]
/Linux_Kernel/linux/arch/alpha/kernel/linux/errno.h:4:10: error: 'asm/errno.h' file not found
/Linux_Kernel/linux/include/linux/types.h:4:10: error: 'asm/types.h' file not found
...

What's left to do?

I have little intention of editing all the #include statements (in every single the linux kernel file) to use double quotes AND dragging the include/ folder into every single directory with a .c file.

Would appreciate some advice on how to make libclang search the include/ file directory I specify.

Any suggestions / alternatives as to how I can go about solving this?

Answer 1

To parse the source file correctly you have to pass to clang_parseTranslationUnit (via argv) all the compilation flags that were used for compilation of the kernel. The easiest way to get all of them is to construct compilation database . You could try using bear or look for other tools. Also, look at similar question: AST of a project by Clang