http://bugs.winehq.org/show_bug.cgi?id=23282
--- Comment #15 from Dylan Smith dylan.ah.smith@gmail.com 2011-11-03 08:35:50 CDT --- Richedit version 1.0 (used by this app) responds to the enter key in WM_CHAR, but version 3.0 responds to the enter key in WM_KEYDOWN. This seems to be the source of the regression.
I simply tested by commenting out TranslateMessage(&msg); in wordpad.c, since that seems to produce the WM_CHAR events from the WM_KEYDOWN events. Afterword typing in wordpad will quickly reveal what events are handled in WM_KEYDOWN, and those that don't work are handled in WM_CHAR.