Tk Source Code

TIP 474 renumbered the mouse buttons for scripts: "buttons 4 up to 9 in scripts will handle the X11 buttons numbered 8 up to 13". Binding patterns and event generate -button follow this, but the %b substitution still reports the internal (X11) number. So the number in %b cannot be used in a binding, and event generate -button no longer "corresponds to the %b substitution" as event.n says. On Windows this is a regression: Tk 8.6 reports the side buttons as 4 and 5 (since [1fb7af623a]), Tk 9 reports 8 and 9.

pack [frame .f -width 200 -height 100]
bind .f <Button-4> {puts "<Button-4> %b"}
bind .f <Button-8> {puts "<Button-8> %b"}
bind .f <ButtonPress> {puts "<ButtonPress> %b"}
# Then press the mouse's Back button over the frame, and:
event generate .f <ButtonPress> -button 4
event generate .f <ButtonPress> -button 8

(Tk 8.6 refuses to bind <Button-8>: "bad button number". Forward behaves the same with 5/9. On Windows the Tk 9 build was 9.1.)

Expected in Tk 9: <Button-4> 4 in the first three rows, and <Button-8> 8 in the last.

The pattern side adds 4 (GetButtonNumber in generic/tkBind.c, and -button in HandleEventGenerate), and printing a pattern subtracts it again (tkBind.c:5220), but the %b case (tkBind.c:3025) prints xbutton.button unchanged. Applying the same mapping there, (button > 7) ? button - 4 : button, would make %b agree with the patterns, and restore the Tk 8.6 values on Windows.

bind.n still says only "the number of a button (1-9)"; it could say that 4 and 5 are the side buttons and that 4-7 of X11 are the wheel.

Tests and documentation fixes would be a very welcome addition!

Thank you for the fantastically quick fix, Jan! Are the documentation changes and tests needed?

Fixed in [b310da761bbab679|trunk] and core-9-0-branch.

Well done!