In Section 4.3.2, we noted that it is not possible to open a file for writing if it is currently being executed (open() returns -1, with errno set to ETXTBSY). Nevertheless, it is possible to do the following from the shell:

$ cc -o longrunner longrunner.c 
$ ./longrunner &                          Leave running in background
$ vi longrunner.c                         Make some changes to the source code
$ cc -o longrunner longrunner.c 

					  

The last command overwrites the existing executable of the same name. How is this possible? (For a clue, use ls -li to look at the i-node number of the executable file after each compilation.)

Why does the call to chmod() in the following code fail?

mkdir("test", S_IRUSR | S_IWUSR | S_IXUSR);
chdir("test");
fd = open("myfile", O_RDWR | O_CREAT, S_IRUSR | S_IWUSR);
symlink("myfile", "../mylink");
chmod("../mylink", S_IRUSR);

Implement realpath().

Modify the program in Example 18-2 (list_files.c) to use readdir_r() instead of readdir().

Implement a function that performs the equivalent of getcwd(). A useful tip for solving this problem is that you can find the name of the current working directory by using opendir() and readdir() to walk through each of the entries in the parent directory (..) to find an entry with the same i-node and device number as the current working directory (i.e., respectively, the st_ino and st_dev fields in the stat structure returned by stat() and lstat()). Thus, it is possible to construct the directory path by walking up the directory tree (chdir(“..”)) one step at a time and performing such scans. The walk can be finished when the parent directory is the same as the current working directory (recall that /.. is the same as /). The caller should be left in the same directory in which it started, regardless of whether your getcwd() function succeeds or fails (open() plus fchdir() are handy for this purpose).

Modify the program in Example 18-3 (nftw_dir_tree.c) to use the FTW_DEPTH flag. Note the difference in the order in which the directory tree is traversed.

Write a program that uses nftw() to traverse a directory tree and finishes by printing out counts and percentages of the various types (regular, directory, symbolic link, and so on) of files in the tree.

Implement nftw(). (This will require the use of the opendir(), readdir(), closedir(), and stat() system calls, among others.)

In Section 18.10, we showed two different techniques (using fchdir() and chdir(), respectively) to return to the previous current working directory after changing the current working directory to another location. Suppose we are performing such an operation repeatedly. Which method do you expect to be more efficient? Why? Write a program to confirm your answer.