Problem:    New regexp engine: look-behind match not checked when followed by
            zero-width match.
Solution:   Do the look-behind match before adding the zero-width state.

Problem:    New regexp engine: addstate() is called very often.
Solution:   Optimize adding the start state.

Problem:    "RO" is not translated everywhere.
Solution:   Put inside _(). (Sergey Alyoshin)

Problem:    When mapping NUL it is displayed as an X.
Solution:   Check for KS_ZERO instead of K_ZERO. (Yasuhiro Matsumoto)

Problem:    Memory leak in ":syntime report".
Solution:   Clear the grow array. (Dominique Pelle)

Problem:    Win32: Check for available memory is not reliable and adds
            overhead.
Solution:   Remove mch_avail_mem(). (Mike Williams)

Problem:    New regexp engine: trying expensive match while the result is not
            going to be used.
Solution:   Check for output state already being in the state list.

Problem:    New regexp engine: negated flag is hardly used.
Solution:   Add separate _NEG states, remove negated flag.

Problem:    New regexp engine: neglist no longer used.
Solution:   Remove the now unused neglist.

Problem:    New regexp engine: collections are slow.
Solution:   Handle all characters in one go.

Problem:    ":func Foo" does not show attributes.
Solution:   Add "abort", "dict" and "range". (Yasuhiro Matsumoto)

Problem:    Compiler warning for unused argument.
Solution:   Add UNUSED.

Problem:    Running test 49 takes a long time.
Solution:   Don't have it grep all files.

Problem:    New regexp engine is a bit slow.
Solution:   Skip ahead to a character that must match.  Don't try matching a
            "^" patter past the start of line.

Problem:    Crash when debugging regexp.
Solution:   Do not try to dump subexpr that were not set.  Skip over count of
            \% items.

Problem:    New regexp engine is a bit slow.
Solution:   Do not clear the state list.  Don't copy syntax submatches when
            not used.

Problem:    Can't build with anything but huge features.
Solution:   Check for FEAT_PROFILE. (Yasuhiro Matsumoto)

Problem:    Can't see what pattern in syntax highlighting is slow.
Solution:   Add the ":syntime" command.

Problem:    Now that the NFA engine handles everything every failure is a
            syntax error.
Solution:   Remove the syntax_error flag.

Problem:    No error for using empty \%[].
Solution:   Give error message.

Problem:    Compiler warning for unitialized variable. (Tony Mechelynck)
Solution:   Assign something to the variable.

Problem:    Error for using \%V in a pattern in tiny Vim.
Solution:   Allow using \%V but never match. (Dominique Pelle)

Problem:    Python: Crash on MS-Windows when os.fchdir() is not available.
Solution:   Check for _chdir to be NULL. (Ken Takata)

Problem:    Can't build tiny Vim on MS-Windows.
Solution:   Adjust #ifdef around using modif_fname(). (Mike Williams)

Problem:    New regexp engine: \%> not supported.
Solution:   Implement \%>.

Problem:    New regexp engine: adding states that are not used.
Solution:   Don't add the states.

Problem:    Crash when regexp logging is enabled.
Solution:   Avoid using NULL pointers. Advance over count argument.

Problem:    Flags in 'cpo' are search for several times.
Solution:   Store the result and re-use the flags.

Problem:    Match failure rate is not very specific.
Solution:   Tune the failure rate for match items.

Problem:    New regexp engine: \%[abc] not supported.
Solution:   Implement \%[abc].  Add tests.

Problem:    Can't build without Visual mode.
Solution:   Add #ifdefs.

Problem:    Many users don't like the cursor line number when 'relativenumber'
            is set.
Solution:   Have four combinations with 'number' and 'relativenumber'.
            (Christian Brabandt)

Problem:    Can't build without the syntax feature.
Solution:   Add #ifdefs. (Erik Falor)

Problem:    New regexp engine: \%'m not supported.
Solution:   Implement \%'m.  Add tests.

Problem:    New regexp engine: \%V not supported.
Solution:   Implement \%V.  Add tests.

Problem:    nfa_recognize_char_class() implementation is inefficient.
Solution:   Use bits in an int instead of chars in a string. (Dominique Pelle)

Problem:    New regexp matching: Using \@= and the like can be slow.
Solution:   Decide whether to first try matching the zero-wdith part or what
            follows, whatever is more likely to fail.

Problem:    Building on MS-Windows doesn't see changes in if_py_both.h.
Solution:   Add a dependency. (Ken Takata)

Problem:    Error message for os.fchdir() (Charles Peacech)
Solution:   Clear the error. (ZyX)