Log Message:
Update to 4.0.24. Following bugs were fixed since 4.0.10:

When using the ** globbing operator, bash will incorrectly add an extra
directory name when the preceding directory name ends with `*' or an empty
string when there is no preceding directory name.

If the prompt length exactly matches the screen width, and the prompt ends
with invisible characters, readline positions the cursor incorrectly.

When parsing case statements in command substitutions, the shell did not
note that a newline is a shell metacharacter and can legally be followed
by a reserved word (e.g., `esac').

When not in a locale supporting multibyte characters, readline will
occasionally not erase characters between the cursor position and the end
of the line when killing text backwards.

If a SIGWINCH arrives while bash is performing redisplay, multi-line prompts
are displayed incorrectly due to the display code being called recursively.

Using an external command as part of the DEBUG trap when job control is
enabled causes pipelines to misbehave.  The problem has to do with process
groups assigned to the pipeline and terminal.

A missing include file results in an empty function definition and a no-op
when checking whether or not the window size has changed.

Adding a null line to a here-document (e.g., by hitting EOF) causes the
shell to dump core attempting to dereference the NULL pointer.

There are several problems with the handling of $LINENO in an ERR trap.

Deferring handling of signals which should cause the shell to terminate until
it is "safe" to run the handler functions does not work for some \ 
terminating
signals.

When the fc builtin is run in a command substitution from a shell with history
enabled, it does not correctly calculate the command on which to operate.

Though references to $@ when there are no positional parameters will now
cause the shell to exit if the `errexit' option has been enabled, constructs
such as ${@:-foo} should not cause an exit.

A case statement using the ;& pattern terminator followed immediately by
"esac" caused a core dump due to a null pointer dereference.

When using the new |& operator following a simple command with a redirection,
the redirection of stderr through the pipe was not performed under certain
circumstances.