Problem:    New regexp state dump is hard to read.
Solution:   Make the state dump more pretty. (Taro Muraoka)

Problem:    New regexp logging is a bit messy.
Solution:   Consistently use #defines, add explanatory comment. (Taro Muraoka)

Problem:    \Z does not work properly with the new regexp engine.
Solution:   Make \Z work.  Add tests.

Problem:    New regexp engine is inefficient with multi-byte characters.
Solution:   Handle a character at a time instead of a byte at a time.  Also
            make \Z partly work.

Problem:    New regexp: adding \Z makes every character match.
Solution:   Only apply ireg_icombine for composing characters.
            Alsl add missing change from patch 1008. (Ken Takata)

Problem:    Compiler warning for ambiguous else.
Solution:   Add curly braces.

Problem:    Test 95 fails on MS-Windows.
Solution:   Set 'nomore'. Change \i to \f.  Change multi-byte character to
            something that is not matching \i. (Ken Takata)

Problem:    Can't build on Minix 3.2.1.
Solution:   Add a condition to an #ifdef. (Gautam Tirumala)

Problem:    NFA engine not used for "\_[0-9]".
Solution:   Enable this, fixed in patch 1005.

Problem:    Get stuck on regexp "\n*" and on "%s/^\n\+/\r".
Solution:   Fix handling of matching a line break. (idea by Hirohito Higashi)

Problem:    No error when option could not be set.
Solution:   Report an error. (ZyX)

Problem:    Python interface does not compile with Python 2.2
Solution:   Fix thread issues and True/False. (ZyX)

Problem:    Valgrind errors for Python interface.
Solution:   Fix memory leaks when running tests. (ZyX)

Problem:    Duplicate condition in if.
Solution:   Remove one condition.

Problem:    Typo in char value causes out of bounds access.
Solution:   Fix character value.  (Klemens Baum)

Problem:    New regexp engine sets curbuf temporarily.
Solution:   Use reg_buf instead, like the old engine.

Problem:    Python: garbage collection issues.
Solution:   Fix the GC issues: Use proper DESTRUCTOR_FINISH: avoids negative
            refcounts, use PyObject_GC_* for objects with tp_traverse and
            tp_clear, add RangeTraverse and RangeClear, use Py_XDECREF in some
            places. (ZyX)

Problem:    Vim and Python exceptions are different.
Solution:   Make Vim exceptions be Python exceptions. (ZyX)

Problem:    Python: Can't check types of what is returned by bindeval().
Solution:   Add vim.List, vim.Dictionary and vim.Function types. (ZyX)

Problem:    Python: Module initialization is duplicated.
Solution:   Move to shared file. (ZyX)

Problem:    Python: using magic constants.
Solution:   Use descriptive values for ml_flags. (ZyX)

Problem:    Python: Later patch does things slightly differently.
Solution:   Adjusted argument type changes. (ZyX)

Problem:    Python: Too many type casts.
Solution:   Change argument types. (ZyX)

Problem:    More can be shared by Python 2 and 3.
Solution:   Move more stuff to if_py_both. (ZyX)

Problem:    Memory leak in new regexp engine.
Solution:   Jump to end of function to free memory. (Dominique Pelle)

Problem:    New regexp engine compares negative numbers to character.
Solution:   Add missing case statements.

Problem:    New regexp engine is slow.
Solution:   Break out of the loop when the state list is empty.

Problem:    No easy to run an individual test.  Tests 64 fails when
            'encoding' is not utf-8.
Solution:   Add individual test targets to the Makefile.  Move some lines from
            test 64 to 95.

Problem:    Test 95 doesn't pass when 'encoding' isn't utf-8. (Yasuhiro
            Matsumoto)
Solution:   Force 'encoding' to be utf-8.

Problem:    GTK vim not started as gvim doesn't set WM_CLASS property to a
            useful value.
Solution:   Call g_set_prgname() on startup. (James McCoy)

Problem:    A Visual mapping that uses CTRL-G works differently when started
            from Insert mode. (Ein Brown)
Solution:   Reset old_mapped_len when handling typed text in Select mode.

Problem:    Uneccessary temp variable.
Solution:   Remove the variable.

Problem:    In the new regexp engine \p does not work on multi-byte
            characters.
Solution:   Don't point to an integer but the characters.

Problem:    In the old regexp engine \i, \I, \f and \F don't work on
            multi-byte characters.
Solution:   Dereference pointer properly.

Problem:    Regexp logs may contain garbage. Character classes don't work
            correctly for multi-byte characters.
Solution:   Check for end of post list.  Only use "is" functions for
            characters up to 255. (Ken Takata)

Problem:    Complex NFA regexp doesn't work.
Solution:   Set actual state stack end instead of using an arbitrary number.
            (Yasuhiro Matsumoto)

Problem:    Regexp debug logs don't have a good name.
Solution:   Use clear names and make it possible to write logs for the old and
            new engines separately. (Taro Muraoka)

Problem:    Compiler warnings on 64 bit Windows.
Solution:   Add type casts. (Mike Williams)  Also fix some white space and
            uncomment what was commented-out for testing.

Problem:    Can't build on HP-UX.
Solution:   Remove modern initialization. (John Marriott)