+.bP
+The Single Unix Specification, Version 2 states that
+\fBvw_scanw\fR is preferred to \fBvwscanw\fR since the latter requires
+including \fB<varargs.h>\fR, which
+cannot be used in the same file as \fB<stdarg.h>\fR.
+This implementation uses \fB<stdarg.h>\fR for both, because that header
+is included in \fB<curses.h\fR>.
+.bP
+X/Open Curses, Issue 5 (December 2007) marked \fBvwscanw\fP (along with
+\fBvwprintw\fP and the termcap interface) as withdrawn.
+.LP
+Both XSI and The Single Unix Specification, Version 2 state that these
+functions return \fBERR\fP or \fBOK\fP.
+.bP
+Since the underlying \fBscanf\fR(3) can return the number of items scanned,
+and the SVr4 code was documented to use this feature,
+this is probably an editing error which was introduced in XSI,
+rather than being done intentionally.
+.bP
+This implementation returns the number of items scanned,
+for compatibility with SVr4 curses.
+As of 2018, NetBSD curses also returns the number of items scanned.
+Both ncurses and NetBSD curses call \fBvsscanf\fP to scan the string,
+which returns \fBEOF\fP on error.
+.bP
+Portable applications should only test if the return value is \fBERR\fP,
+since the \fBOK\fP value (zero) is likely to be misleading.
+.IP
+One possible way to get useful results would be to use a "%n" conversion
+at the end of the format string to ensure that something was processed.