Finding current executable's path without /proc/self/exe

Viewed 128096

It seems to me that Linux has it easy with /proc/self/exe. But I'd like to know if there is a convenient way to find the current application's directory in C/C++ with cross-platform interfaces. I've seen some projects mucking around with argv[0], but it doesn't seem entirely reliable.

If you ever had to support, say, Mac OS X, which doesn't have /proc/, what would you have done? Use #ifdefs to isolate the platform-specific code (NSBundle, for example)? Or try to deduce the executable's path from argv[0], $PATH and whatnot, risking finding bugs in edge cases?

14 Answers

Some OS-specific interfaces:

There are also third party libraries that can be used to get this information, such as whereami as mentioned in prideout's answer, or if you are using Qt, QCoreApplication::applicationFilePath() as mentioned in the comments.

The portable (but less reliable) method is to use argv[0]. Although it could be set to anything by the calling program, by convention it is set to either a path name of the executable or a name that was found using $PATH.

Some shells, including bash and ksh, set the environment variable "_" to the full path of the executable before it is executed. In that case you can use getenv("_") to get it. However this is unreliable because not all shells do this, and it could be set to anything or be left over from a parent process which did not change it before executing your program.

If you ever had to support, say, Mac OS X, which doesn't have /proc/, what would you have done? Use #ifdefs to isolate the platform-specific code (NSBundle, for example)?

Yes, isolating platform-specific code with #ifdefs is the conventional way this is done.

Another approach would be to have a have clean #ifdef-less header which contains function declarations and put the implementations in platform specific source files.

For example, check out how POCO (Portable Components) C++ library does something similar for their Environment class.

In addition to mark4o's answer, FreeBSD also has

const char* getprogname(void)

It should also be available in macOS. It's available in GNU/Linux through libbsd.

AFAIK, no such way. And there is also an ambiguity: what would you like to get as the answer if the same executable has multiple hard-links "pointing" to it? (Hard-links don't actually "point", they are the same file, just at another place in the file system hierarchy.)

Once execve() successfully executes a new binary, all information about the arguments to the original program is lost.

Well, of course, this doesn't apply to all projects. Still, QCoreApplication::applicationFilePath() never failed me in 6 years of C++/Qt development.

Of course, one should read documentation thoroughly before attempting to use it:

Warning: On Linux, this function will try to get the path from the /proc file system. If that fails, it assumes that argv[0] contains the absolute file name of the executable. The function also assumes that the current directory has not been changed by the application.

To be honest, I think that #ifdef and other solutions like that shouldn't be used in modern code at all.

I'm sure that smaller cross-platform libraries also exist. Let them encapsulate all that platform-specific stuff inside.

If you're writing GPLed code and using GNU autotools, then a portable way that takes care of the details on many OSes (including Windows and macOS) is gnulib's relocatable-prog module.

Just my two cents. You can find the current application's directory in C/C++ with cross-platform interfaces by using this code.

void getExecutablePath(char ** path, unsigned int * pathLength)
{
    // Early exit when invalid out-parameters are passed
    if (!checkStringOutParameter(path, pathLength))
    {
        return;
    }

#if defined SYSTEM_LINUX

    // Preallocate PATH_MAX (e.g., 4096) characters and hope the executable path isn't longer (including null byte)
    char exePath[PATH_MAX];

    // Return written bytes, indicating if memory was sufficient
    int len = readlink("/proc/self/exe", exePath, PATH_MAX);

    if (len <= 0 || len == PATH_MAX) // memory not sufficient or general error occured
    {
        invalidateStringOutParameter(path, pathLength);
        return;
    }

    // Copy contents to caller, create caller ownership
    copyToStringOutParameter(exePath, len, path, pathLength);

#elif defined SYSTEM_WINDOWS

    // Preallocate MAX_PATH (e.g., 4095) characters and hope the executable path isn't longer (including null byte)
    char exePath[MAX_PATH];

    // Return written bytes, indicating if memory was sufficient
    unsigned int len = GetModuleFileNameA(GetModuleHandleA(0x0), exePath, MAX_PATH);
    if (len == 0) // memory not sufficient or general error occured
    {
        invalidateStringOutParameter(path, pathLength);
        return;
    }

    // Copy contents to caller, create caller ownership
    copyToStringOutParameter(exePath, len, path, pathLength);

#elif defined SYSTEM_SOLARIS

    // Preallocate PATH_MAX (e.g., 4096) characters and hope the executable path isn't longer (including null byte)
    char exePath[PATH_MAX];

    // Convert executable path to canonical path, return null pointer on error
    if (realpath(getexecname(), exePath) == 0x0)
    {
        invalidateStringOutParameter(path, pathLength);
        return;
    }

    // Copy contents to caller, create caller ownership
    unsigned int len = strlen(exePath);
    copyToStringOutParameter(exePath, len, path, pathLength);

#elif defined SYSTEM_DARWIN

    // Preallocate PATH_MAX (e.g., 4096) characters and hope the executable path isn't longer (including null byte)
    char exePath[PATH_MAX];

    unsigned int len = (unsigned int)PATH_MAX;

    // Obtain executable path to canonical path, return zero on success
    if (_NSGetExecutablePath(exePath, &len) == 0)
    {
        // Convert executable path to canonical path, return null pointer on error
        char * realPath = realpath(exePath, 0x0);

        if (realPath == 0x0)
        {
            invalidateStringOutParameter(path, pathLength);
            return;
        }

        // Copy contents to caller, create caller ownership
        unsigned int len = strlen(realPath);
        copyToStringOutParameter(realPath, len, path, pathLength);

        free(realPath);
    }
    else // len is initialized with the required number of bytes (including zero byte)
    {
        char * intermediatePath = (char *)malloc(sizeof(char) * len);

        // Convert executable path to canonical path, return null pointer on error
        if (_NSGetExecutablePath(intermediatePath, &len) != 0)
        {
            free(intermediatePath);
            invalidateStringOutParameter(path, pathLength);
            return;
        }

        char * realPath = realpath(intermediatePath, 0x0);

        free(intermediatePath);

        // Check if conversion to canonical path succeeded
        if (realPath == 0x0)
        {
            invalidateStringOutParameter(path, pathLength);
            return;
        }

        // Copy contents to caller, create caller ownership
        unsigned int len = strlen(realPath);
        copyToStringOutParameter(realPath, len, path, pathLength);

        free(realPath);
    }

#elif defined SYSTEM_FREEBSD

    // Preallocate characters and hope the executable path isn't longer (including null byte)
    char exePath[2048];

    unsigned int len = 2048;

    int mib[] = { CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1 };

    // Obtain executable path by syscall
    if (sysctl(mib, 4, exePath, &len, 0x0, 0) != 0)
    {
        invalidateStringOutParameter(path, pathLength);
        return;
    }

    // Copy contents to caller, create caller ownership
    copyToStringOutParameter(exePath, len, path, pathLength);

#else

    // If no OS could be detected ... degrade gracefully
    invalidateStringOutParameter(path, pathLength);

#endif
}

You can take a look in detail here.

But I'd like to know if there is a convenient way to find the current application's directory in C/C++ with cross-platform interfaces.

You cannot do that (at least on Linux)

Since an executable could, during execution of a process running it, rename(2) its file path to a different directory (of the same file system). See also syscalls(2) and inode(7).

On Linux, an executable could even (in principle) remove(3) itself by calling unlink(2). The Linux kernel should then keep the file allocated till no process references it anymore. With proc(5) you could do weird things (e.g. rename(2) that /proc/self/exe file, etc...)

In other words, on Linux, the notion of a "current application's directory" does not make any sense.

Read also Advanced Linux Programming and Operating Systems: Three Easy Pieces for more.

Look also on OSDEV for several open source operating systems (including FreeBSD or GNU Hurd). Several of them provide an interface (API) close to POSIX ones.

Consider using (with permission) cross-platform C++ frameworks like Qt or POCO, perhaps contributing to them by porting them to your favorite OS.

Related