Hi all,

I am working on a C program in Unix that basically is supposed to have the same functionality as a shell. I have most commands up and running. I cannot find solutions to two of the commands I have to implement though. They are:

'where filename' - does a search for the file called filename in each directory listed in the PATH and says the path it's located in if found

'disk' - prints amount of available disk space in system

Any answers will be appreciated.

Just as an aside, AFAIK, the which command is often an external command, that is not built into the shell.

But if you wish to write a C language program that does what the which command does, you might wish to take a look at information on the string related functions, such as:

stpcpy, strcasecmp, strcat, strchr, strcmp, strcoll, strcpy, strcspn, strdup, strfry, strlen, strncat, strncmp, strncpy, strncasecmp, strpbrk, strrchr, strsep, strspn, strstr, strtok, strxfrm, index, rindex

You need to first obtain the path list, e.g. using getenv("PATH"). Allocate a temporary string long enough to contain both the longest directory (delimited by colons) in the path list, a slash, the file name, and a trailing EOS (zero byte) to mark the end of the string.

Write a loop that tries each directory specified in the path list in the order they're listed. Remember to skip over empty ones, sometimes PATH contains multiple successive colons (like ::). Append a slash and the file name to the directory name, followed by an end of string mark '\0', then check using e.g access(pathname,X_OK) to see if that one is executable. If it is, this is the result; don't check any further. If none of them match, then most which implementations return nothing.

Most of this is string manipulation. Personally, I would only need strlen(), strcspn() and strspn() (the latter two to find where the next colon is, and to skip over the colon -- these can be written using a trivial loop too, of course).

Usually most people tend to write interfaces that save the pathname to a buffer provided by the caller. I disagree, strongly. Compare the GNU/POSIX.1-2008 getline() interface to old fgets(). As the example at the end of the getline man page shows, a trivial program can accept any input without problems, as long as there is enough resources available. A corresponding fgets() program is either wasteful (using a lot of memory for a buffer that is only partially used), or will get confused with long lines.

I shudder to think what kind of security problems could be caused by a part of a shell silently truncating the PATH string! Especially if it truncated to one length at one place, and to another length at another place. No, please! Don't do it! Noo.......

Therefore, I'd also use malloc() to allocate the pathname buffer (also used to return the string to the caller). Hint: malloc(strlen(pathlist)+strlen(filename)+2) is guaranteed to be long enough, and is not wasteful as the path list is always quite short. One extra char is for the slash, the other for the end-of-string '\0'. Just remember that when a problem occurs, the function should remember to free() the temporary/result buffer, and that all callers should remember to free() the pointer they receive after they've done with it.

1 members found this post helpful.