[PATCH] Implemented support for mouse scrolling in weston-terminal
Magnus Hoff
maghoff+wayland at gmail.com
Tue Aug 5 10:09:52 PDT 2014
On Tue, Aug 5, 2014 at 4:44 PM, Marek Chalupa <mchqwerty at gmail.com> wrote:
> Hi,
> In the first place I'd lake to say: great feature!
Thanks :)
> On 08/05/14 15:05, maghoff+wayland at gmail.com wrote:
>>
>> From: Magnus Hoff <maghoff at gmail.com>
>>
>> ---
>> clients/terminal.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++++
>> 1 file changed, 48 insertions(+)
>>
>> diff --git a/clients/terminal.c b/clients/terminal.c
>> index eb133cd..d238f04 100644
>> --- a/clients/terminal.c
>> +++ b/clients/terminal.c
>> @@ -447,6 +447,7 @@ struct terminal {
>> int width, height, row, column, max_width;
>> uint32_t buffer_height;
>> uint32_t start, end, saved_start, log_size;
>> + wl_fixed_t smooth_scroll;
>> int saved_row, saved_column;
>> int scrolling;
>> int send_cursor_position;
>> @@ -2785,6 +2786,52 @@ motion_handler(struct widget *widget,
>> return CURSOR_IBEAM;
>> }
>> +// This magnitude is chosen rather arbitrarily. Really, the scrolling
>> +// should happen on a (fractional) pixel basis, not a line basis.
>> +#define AXIS_UNITS_PER_LINE 256
>> +
>
>
> Personally, I liked 512 more - but that is just my subjective feeling. It'd
> be nice to have it configurable.
I'm hoping to implement pixel granularity scrolling in short order,
hopefully obsoleting this constant altogether. But I figured I would
try sending a smaller patch first :)
>> +static void
>> +axis_handler(struct widget *widget,
>> + struct input *input, uint32_t time,
>> + uint32_t axis,
>> + wl_fixed_t value,
>> + void *data)
>> +{
>> + struct terminal *terminal = data;
>> +
>> + if (axis != WL_POINTER_AXIS_VERTICAL_SCROLL) return;
>> +
>> + terminal->smooth_scroll += value;
>> + int lines = terminal->smooth_scroll / AXIS_UNITS_PER_LINE;
>> + terminal->smooth_scroll -= lines * AXIS_UNITS_PER_LINE;
>> +
>> + if (lines > 0) {
>> + if (terminal->scrolling) {
>> + if ((uint32_t)lines > terminal->saved_start -
>> terminal->start)
>> + lines = terminal->saved_start -
>> terminal->start;
>> + } else {
>> + lines = 0;
>> + }
>> + } else if (lines < 0) {
>> + uint32_t neg_lines = -lines;
>> + if (neg_lines > terminal->log_size + terminal->start -
>> terminal->end)
>> + lines = terminal->end - terminal->log_size -
>> terminal->start;
>> + }
>> +
>> + if (lines) {
>> + if (!terminal->scrolling)
>> + terminal->saved_start = terminal->start;
>> + terminal->scrolling = 1;
>> +
>> + terminal->start += lines;
>> + terminal->row -= lines;
>> + terminal->selection_start_row -= lines;
>> + terminal->selection_end_row -= lines;
>> +
>> + widget_schedule_redraw(widget);
>> + }
>> +}
>> +
>> static void
>> output_handler(struct window *window, struct output *output, int enter,
>> void *data)
>> @@ -2880,6 +2927,7 @@ terminal_create(struct display *display)
>> widget_set_button_handler(terminal->widget, button_handler);
>> widget_set_enter_handler(terminal->widget, enter_handler);
>> widget_set_motion_handler(terminal->widget, motion_handler);
>> + widget_set_axis_handler(terminal->widget, axis_handler);
>> widget_set_touch_up_handler(terminal->widget, touch_up_handler);
>> widget_set_touch_down_handler(terminal->widget,
>> touch_down_handler);
>> widget_set_touch_motion_handler(terminal->widget,
>> touch_motion_handler);
>
>
> So far I tried works good for most usecases. There's one usecase that is
> behaving incorrectly I think. When you put something into buffer, then
> scroll it up, back down, then resize the terminal so that it has k less
> lines than before and scroll again up and down, than the scrolling stops k
> lines above the end of buffer. The same way you can resize to the other side
> (and end up k lines after the buffer). Pressing enter clears it. But that's
> not critical, I think.
It is also possible to scroll up and down with ctrl+shift+up/down
arrow, and it looks like I can replicate the behavior you are
describing with this scrolling functionality. So this is not a new bug
:)
It might be time to refactor the scrolling functionality soon and
maybe fix this bug in one place.
> Regards,
> Marek
>
More information about the wayland-devel
mailing list