C.1.3 Here-Documents

Don’t rely on ‘\’ being preserved just because it has no special meaning together with the next symbol. In the native sh on OpenBSD 2.7 ‘\"’ expands to ‘"’ in here-documents with unquoted delimiter. As a general rule, if ‘\\’ expands to ‘\’ use ‘\\’ to get ‘\’.

With OpenBSD 2.7’s sh

$ cat <<EOF
> \" \\
> EOF
" \

and with Bash:

bash-2.04$ cat <<EOF
> \" \\
> EOF
\" \

Using command substitutions in a here-document that is fed to a shell function is not portable. For example, with Solaris 10 /bin/sh:

$ kitty () { cat; }
$ kitty <<EOF
> `echo ok`
> EOF
/tmp/sh199886: cannot open
$ echo $?
1

Some shells mishandle large here-documents: for example, Solaris 10 dtksh and the UnixWare 7.1.1 Posix shell, which are derived from Korn shell version M-12/28/93d, mishandle braced variable expansion that crosses a 1024- or 4096-byte buffer boundary within a here-document. Only the part of the variable name after the boundary is used. For example, ${variable} could be replaced by the expansion of ${ble}. If the end of the variable name is aligned with the block boundary, the shell reports an error, as if you used ${}. Instead of ${variable-default}, the shell may expand ${riable-default}, or even ${fault}. This bug can often be worked around by omitting the braces: $variable. The bug was fixed in ‘ksh93g’ (1998-04-30) but as of 2006 many operating systems were still shipping older versions with the bug.

Empty here-documents are not portable either; with the following code, zsh up to at least version 4.3.10 creates a file with a single newline, whereas other shells create an empty file:

cat >file <<EOF
EOF

Many shells (including the Bourne shell) implement here-documents inefficiently. In particular, some shells can be extremely inefficient when a single statement contains many here-documents. For instance if your configure.ac includes something like:

if <cross_compiling>; then
  assume this and that
else
  check this
  check that
  check something else
  …
  on and on forever
  …
fi

A shell parses the whole if/fi construct, creating temporary files for each here-document in it. Some shells create links for such here-documents on every fork, so that the clean-up code they had installed correctly removes them. It is creating the links that can take the shell forever.

Moving the tests out of the if/fi, or creating multiple if/fi constructs, would improve the performance significantly. Anyway, this kind of construct is not exactly the typical use of Autoconf. In fact, it’s even not recommended, because M4 macros can’t look into shell conditionals, so we may fail to expand a macro when it was expanded before in a conditional path, and the condition turned out to be false at runtime, and we end up not executing the macro at all.

Be careful with the use of ‘<<-’ to unindent here-documents. The behavior is only portable for stripping leading TABs, and things can silently break if an overzealous editor converts to using leading spaces (not all shells are nice enough to warn about unterminated here-documents).

$ printf 'cat <<-x\n\t1\n\t 2\n\tx\n' | bash && echo done
1
 2
done
$ printf 'cat <<-x\n 1\n  2\n x\n' | bash-3.2 && echo done
 1
  2
 x
done