Also trim or improve a few comments.

That is: elide a second test from the most travelled path: a valid
character.  This adds a second call of mblen() when parse_mbchar()
is called on a terminating zero, but that should never happen.

The fsfromline() function is mostly used by the undo functions, which
are not present in the tiny version.  It is also used by the comment/
uncomment feature, but this feature cannot be enabled when --enable-tiny
is in effect.

Achieve this by eliding two conditions from the inner loop,
which is possible because 'end' will never be NULL.

Instead compute directly whether we're at a softwrapped part or not.

Convert digits() to take a ssize_t instead of an int, since it's
used on ssize_t line numbers.  And properly use the long modifier
when displaying a line number.

Also, conditionalize the digits() prototype.

An added magic linefeed should be removed again /before/ restoring
the x position, as the latter needs to be calculated from the real
last line of the region.

This fixes https://savannah.gnu.org/bugs/?49817

.
Reported-by: Mike Frysinger <vapier@gentoo.org>

It is easier to see the order of steps and what the terminating
condition is.

When wanting to debug something, it is far more useful
to temporarily insert lines like:

    statusline(ALERT, "name = %i", variable);

It provides instant feedback, and it slows things down,
so you can kind of see what happens.

When there are line numbers, the effective available screen width is
smaller than COLS: editwincols.

This fixes https://savannah.gnu.org/bugs/?49821.

Also, be more selective when coloring a git diff line.  And color
an svn index line too.

This addresses https://savannah.gnu.org/bugs/?49803.

Like do_down() does already, do_pageup() and do_right() should use
editwincols instead of COLS.

This addresses https://savannah.gnu.org/bugs/?49796.

The functions read() and fwrite() take size_t, not ssize_t.

And line numbers in the file should be displayed as a long type instead
of an int, since the effective type of ssize_t is not int, but long.

(The variable 'pletion_line' is not conditionalized with this option, as
it would become messy.  The compiler will probably be able to elide it.)

When using --enable-tiny, it is not possible to use --enable-wordcomp,
because the word completion function uses the undo system.

Executing the 'complete_a_word' function will search from the start
of the current buffer for entire words that begin with the fragment
that is before the cursor, and will complete this fragment to the
first word that is found.  Each consecutive call of 'complete_a_word'
will search for the next matching word and will complete the fragment
to that.  By default the function is bound to the ^] keystroke.
Signed-off-by: Sumedh Pendurkar <sumedh.pendurkar@gmail.com>
Signed-off-by: Benno Schulenberg <bensberg@justemail.net>

But do it correctly this time: don't switch it on when replacing.

None of the prompts offer a total-refresh shortcut -- only the
edit window, the help viewer, and the file browser provide this.

It's redundant because the call of total_refresh() wipes everything.

At the very beginning of do_replace(), nothing has been changed
or initialized yet, so there is nothing to refresh or clean up.