The official errata for the THINK Unix textbook is located at:

http://www.tux.org/~lasser/think-unix/errata.html

The section The Easy Way refers to >&. This is a csh mechanism for merging STDERR with the redirected STDOUT; it is not generally valid in sh-based shells (note: the bash shell does support this csh-style syntax). To test out the examples, start a csh or bash shell and try the problem as is:

$ csh

% find / -name '*' -print >& ~/allfiles.txt

or for all sh type shells use:

$ find / -name '*' -print > ~/allfiles.txt 2>&1

The ^ username should be ^username, with no space between the ^ and the word username.

Same error as on page 67: The ^ username should be ^username.

Again, the >& merge syntax is the csh-style syntax. The 2>&1 syntax is available for all sh type shells.

The statement about signal 9: "It's much harder for a process to avoid being terminated with this signal" might be considered a bit misleading. In fact, it is impossible for a process to avoid being terminated with signal 9. The kernel does not deliver this signal to the process; instead it terminates the process directly and immediately*. Remember signal 9 cannot be caught or blocked by a process.

*In rare cases (typically only with poorly written device drivers and malfunctioning hardware), the kernel will not be able to kill the process if the process is sleeping waiting for an I/O completion event that might never occur.

For problems 6 through 9, the rev and tac programs are open source GNU programs, and may not be included in your UNIX distribution. They are not part of Solaris, for example. Use the cat program for these practice problems instead.

The answer to problem 5 is incorrect. It should be:

telephone [-h|-s] [[1[-]]NNN[-]]NNN[-]NNNN

Note the movement of the brace before the country code 1 in the phone number.

The answer to problem 21 has a mistake. Permissions on one hard link do affect permissions on other hard links.

The answer to problem 4 for chapter 5 has a mistake. The map should not include the w to write out the file. The correct mapping should be:

:map! <F7> ^[:q!

Additionally, the problem did ask that you create an additional mapping, so the answer should not have used the F7 function key again, since it was already mapped in the previous problem. (Thanks Connie N. for this correction, and for the following two corrections).

The answer to problem 8 for chapter 5 also has a mistake. The answer is clearly confusing the meaning of the . and the ? characters in REs (these solutions certainly were not tested!). The ? character does not mean match any character in REs (as it does in shell wildcards), but in EREs means 0 or 1 of the previous atom. The correct solution should be:

grep 'n.t'

The answer to problem 9 for chapter 5 is completely incorrect. First, as in the previous errata entry for problem 8, a . needs to replace the ? as the match single character meta-character. Second, the solution is implying that the syntax [ ^] will match either the space character or the beginning of the line and that [ $] will match the space character or the end of the line, thus designating a word. This is incorrect, and does not work. A ^ or a $ inside of the range brackets has no special meaning (except when ^ is used as the initial character to mean not in the range). To match a word, you must use \<word\>. Finally, the problem required a letter as the second character, not any character as the solution seems to allow. The correct solution would be:

grep \<n[a-zA-Z]t\>.

The answer to problem 12 for chapter 5 uses the i flag for the substitution. This flag is supported on the GNU version of sed which the author uses, but is not supported on the Solaris version of sed. Sun has traditionally not advanced their basic UNIX toolset to keep up with the industry. The correct Solaris implementation would be:

sed 's/[Yy][Ee][Ss]/no/g'

This of course does not preserve the original case of the "yes", since it forces a replacement to the lowercase "no". (Challenge - can you preserve the case?).

The answer to problem 14 for chapter 5 would replace the last occurrence, not the second occurrence as required. In fact, the answer is much simpler than that shown in the book:

sed 's/No/Yes/2'