On Mon, Dec 19, 2011 at 12:25:24PM -0800, Jason Gerecke wrote:
> On Sun, Dec 18, 2011 at 10:06 PM, Peter Hutterer
> <[email protected]> wrote:
> > On Tue, Dec 13, 2011 at 01:59:02PM -0800, Jason Gerecke wrote:
> >> Replaces sendWheelStripEvents and getWheelButton with several
> >> small functions to reduce unnecessary code duplication.
> >>
> >> As a side-effect, it is now possible for the driver to handle
> >> simultaneous scrolls on multiple axes. Previously, getWheelButton
> >> would go through all the axes but was limited to returning a
> >> single button/action.
> >>
> >> Signed-off-by: Jason Gerecke <[email protected]>
> >
> > thanks! this patch makes it messy, but at least removes the copy/paste.
> It'd be nice if there were a 'tunable' diff that let you adjust how it
> matches subsequences... Patience diff is usually pretty good, but even
> it coughed up a mess for this particular patch. Its not that bad of a
> patch when done by hand, but the algorithms get caught up matching the
> wrong bits of "common" text. :(

I admit that with diffs like this, I usually apply the patches and look at
the resulting source.

> > aside from the comments below, please reshuffle this one so it applies
> > _before_ the other one in the series. short story: this way we can bisect
> > down to check if this patch broke anything on its own before the new feature
> > was added.
> >
> That's actually the reason I put it as the final patch instead of
> somewhere in the middle. Though, I suppose I can put it at the
> beginning of sequence as well.

it should be relatively easy since you're now just calling the new
functions. To expand on this a little - by default we have to assume that
one day we may have to revert a patch because it broke something. To
identify the patch, we may have to bisect.

so if you have structural changes, they should go in first. A patch, ideally
with the note "no functional changes" if it's just moving stuff around. Then
the extra features (or new users) on top. If something breaks, it's easy to
identify whether it was the structural change or the new feature.

And - if it was the new feature it can be quickly reverted if there is
release pressure. If it was the structurual change, it can be fixed.

If you have the cleanup after the new feature, reverting the feature is
hard. Reverting the cleanup is worse, because it leaves you with copy/pasted
code and that's just generally a big no-no.

> > coincidentally, writing tests for these so they keep exhibiting the same
> > behaviour would be great...
> >
> Good call -- I always seem to forget that bit when adding new
> functions. I'll whip some up during the rebase.
> 
> >> ---
> >>  src/wcmCommon.c |  206 
> >> ++++++++++++++++++++++++++++++++-----------------------
> >>  1 files changed, 121 insertions(+), 85 deletions(-)
> >>
> >> diff --git a/src/wcmCommon.c b/src/wcmCommon.c
> >> index 0e719cf..170d1d0 100644
> >> --- a/src/wcmCommon.c
> >> +++ b/src/wcmCommon.c
> >> @@ -317,98 +317,106 @@ static void sendAButton(InputInfoPtr pInfo, int 
> >> button, int mask,
> >>                  first_val, num_val, valuators);
> >>  }
> >>
> >> -/*****************************************************************************
> >> - * getWheelButton --
> >> - *   Get the wheel button to be sent for the current device state.
> >> - 
> >> ****************************************************************************/
> >> -
> >> -static int getWheelButton(InputInfoPtr pInfo, const WacomDeviceState* ds,
> >> -                       unsigned int **fakeKey)
> >> +/**
> >> + * Get the distance an axis was scrolled. This function is aware
> >> + * of the different ways different scrolling axes work and strives
> >> + * to produce a common representation of relative change.
> >> + *
> >> + * @param current  Current value of the axis
> >> + * @param old      Previous value of the axis
> >> + * @param wrap     Value after which wraparound occurs (0 if axis does 
> >> not wrap)
> >> + * @param invert   Scroll axis is inverted
> >> + * @param bitwise  Scroll axis values change in a bitwise manner
> >> + * @return         Relative change in axis value
> >> + */
> >> +static int getScrollDelta(int current, int old, int wrap, Bool invert, 
> >> Bool bitwise)
> >
> > I'd really like either a usage of flags or enums here so that the code is
> > easier to read.
> >        delta = getScrollDelta(ds->stripx, priv->oldStripX, 0, TRUE, TRUE);
> > is not at all obvious, getScrollDelta(..., INVERT|BITWISE) would be.
> >
> Flags work for me.
> 
> >>  {
> >> -     WacomDevicePtr priv = (WacomDevicePtr) pInfo->private;
> >> -     int fakeButton = 0, value = 0;
> >> +     int delta;
> >>
> >> -     /* emulate events for relative wheel */
> >> -     if ( ds->relwheel )
> >> +     if (bitwise)
> >>       {
> >> -             value = ds->relwheel;
> >> -             fakeButton = (value > 0) ? priv->relup : priv->reldn;
> >> -             *fakeKey = (value > 0) ? priv->wheel_keys[0+1] : 
> >> priv->wheel_keys[1+1];
> >> +             current = (int)log2(current);
> >> +             old = (int)log2(old);
> >>       }
> >>
> >> -     /* emulate events for absolute wheel when it is a touch ring (on 
> >> pad) */
> >> -     if ( (ds->abswheel != priv->oldWheel) && IsPad(priv) &&
> >> -         (priv->oldProximity == ds->proximity))
> >> -     {
> >> -             int wrap_delta;
> >> -             value = priv->oldWheel - ds->abswheel;
> >> -
> >> -             /* Wraparound detection. If the distance oldvalue..value is
> >> -              * larger than the oldvalue..value considering the
> >> -              * wraparound, assume wraparound and readjust */
> >> -             if (value < 0)
> >> -                     wrap_delta = ((MAX_PAD_RING + 1) + priv->oldWheel) - 
> >> ds->abswheel;
> >> -             else
> >> -                     wrap_delta = priv->oldWheel - ((MAX_PAD_RING + 1) + 
> >> ds->abswheel);
> >> +     delta = current - old;
> >>
> >> -             DBG(12, priv, "wrap detection for %d (old %d): %d (wrap 
> >> %d)\n",
> >> -                 ds->abswheel, priv->oldWheel, value, wrap_delta);
> >> +     if (invert)
> >> +             delta = -delta;
> >>
> >> -             if (abs(wrap_delta) < abs(value))
> >> -                     value = wrap_delta;
> >> -
> >> -             fakeButton = (value > 0) ? priv->wheelup : priv->wheeldn;
> >> -             *fakeKey = (value > 0) ? priv->wheel_keys[2+1] : 
> >> priv->wheel_keys[3+1];
> >> -     }
> >> -
> >> -     /* emulate events for 2nd absolute wheel when it is a touch ring (on 
> >> pad) */
> >> -     if ( (ds->abswheel2 != priv->oldWheel2) && IsPad(priv) &&
> >> -         (priv->oldProximity == ds->proximity))
> >> +     if (wrap != 0)
> >>       {
> >> +             /* Wraparound detection. If the distance old..current
> >> +              * is larger than the old..current considering the
> >> +              * wraparound, assume wraparound and readjust */
> >>               int wrap_delta;
> >> -             value = priv->oldWheel2 - ds->abswheel2;
> >>
> >> -             /* Wraparound detection. If the distance oldvalue..value is
> >> -              * larger than the oldvalue..value considering the
> >> -              * wraparound, assume wraparound and readjust */
> >> -             if (value < 0)
> >> -                     wrap_delta = ((MAX_PAD_RING + 1) + priv->oldWheel2) 
> >> - ds->abswheel2;
> >> +             if (delta < 0)
> >> +                     wrap_delta = (wrap + 1) + delta;
> >>               else
> >> -                     wrap_delta = priv->oldWheel2 - ((MAX_PAD_RING + 1) + 
> >> ds->abswheel2);
> >> +                     wrap_delta = -((wrap + 1) - delta);
> >>
> >> -             DBG(12, priv, "wrap detection for %d (old %d): %d (wrap 
> >> %d)\n",
> >> -                 ds->abswheel2, priv->oldWheel2, value, wrap_delta);
> >> +             if (abs(wrap_delta) < abs(delta))
> >> +                     delta = wrap_delta;
> >> +     }
> >>
> >> -             if (abs(wrap_delta) < abs(value))
> >> -                     value = wrap_delta;
> >> +     return delta;
> >> +}
> >>
> >> -             fakeButton = (value > 0) ? priv->wheel2up : priv->wheel2dn;
> >> -             *fakeKey = (value > 0) ? priv->wheel_keys[4+1] : 
> >> priv->wheel_keys[5+1];
> >> -     }
> >> +/**
> >> + * Get the scroll button/action to send given the delta of
> >> + * the scrolling axis and the possible events that can be
> >> + * sent.
> >> + *
> >> + * @param delta        Amount of change in the scrolling axis
> >> + * @param button_up    Button event to send on scroll up
> >> + * @param button_dn    Button event to send on scroll down
> >> + * @param action_up    Action to send on scroll up
> >> + * @param action_dn    Action to send on scroll down
> >> + * @param[out] action  Action that should be performed
> >> + * @return             Button that should be pressed
> >> + */
> >> +static int getWheelButton(int delta, int button_up, int button_dn,
> >> +                          unsigned int *action_up, unsigned int 
> >> *action_dn,
> >> +                          unsigned int **action)
> >> +{
> >> +     int button = 0;
> >> +     *action = NULL;
> >>
> >> -     /* emulate events for left strip */
> >> -     if ( ds->stripx != priv->oldStripX )
> >> +     if (delta)
> >>       {
> >> -             value = ds->stripx - priv->oldStripX;
> >> -
> >> -             fakeButton = (value < 0) ? priv->striplup : priv->stripldn;
> >> -             *fakeKey = (value < 0) ? priv->strip_keys[0+1] : 
> >> priv->strip_keys[1+1];
> >> +             button  = delta > 0 ? button_up : button_dn;
> >> +             *action = delta > 0 ? action_up : action_dn;
> >>       }
> >>
> >> -     /* emulate events for right strip */
> >> -     if ( ds->stripy != priv->oldStripY )
> >> -     {
> >> -             value = ds->stripy - priv->oldStripY;
> >> +     return button;
> >> +}
> >>
> >> -             fakeButton = (value < 0) ? priv->striprup : priv->striprdn;
> >> -             *fakeKey = (value < 0) ? priv->strip_keys[2+1] : 
> >> priv->strip_keys[3+1];
> >> -     }
> >> +/**
> >> + * Send button or actions for a scrolling axis.
> >> + *
> >> + * @param button     X button number to send if no action is defined
> >> + * @param action     Action to send
> >> + * @param pInfo
> >> + * @param first_val
> >> + * @param num_vals
> >> + * @param valuators
> >> + */
> >> +static void sendWheelStripEvent(int button, unsigned int *action, 
> >> InputInfoPtr pInfo,
> >> +                                 int first_val, int num_vals, int 
> >> *valuators)
> >> +{
> >> +     WacomDevicePtr priv = (WacomDevicePtr) pInfo->private;
> >>
> >> -     DBG(10, priv, "send fakeButton %x with value = %d \n",
> >> -             fakeButton, value);
> >> +     unsigned int button_action[1] = {button | AC_BUTTON | 
> >> AC_KEYBTNPRESS};
> >> +     if (!action || !(*action)) {
> >> +             DBG(10, priv, "No wheel/strip action set; sending button %d 
> >> (action %d).\n", button, button_action[0]);
> >> +             action = &button_action[0];
> >> +     }
> >>
> >> -     return fakeButton;
> >> +     sendAction(pInfo, 1, action, ARRAY_SIZE(action), first_val, 
> >> num_vals, valuators);
> >> +     sendAction(pInfo, 0, action, ARRAY_SIZE(action), first_val, 
> >> num_vals, valuators);
> >>  }
> >> +
> >>  /*****************************************************************************
> >>   * sendWheelStripEvents --
> >>   *   Send events defined for relative/absolute wheels or strips
> >> @@ -418,31 +426,59 @@ static void sendWheelStripEvents(InputInfoPtr pInfo, 
> >> const WacomDeviceState* ds,
> >>                                int first_val, int num_vals, int *valuators)
> >>  {
> >>       WacomDevicePtr priv = (WacomDevicePtr) pInfo->private;
> >> -     int fakeButton = 0;
> >> +     int fakeButton = 0, delta = 0;
> >>       unsigned int *fakeKey = NULL;
> >>
> >>       DBG(10, priv, "\n");
> >>
> >> -     fakeButton = getWheelButton(pInfo, ds, &fakeKey);
> >> +     /* emulate events for left strip */
> >> +     delta = getScrollDelta(ds->stripx, priv->oldStripX, 0, TRUE, TRUE);
> >> +     if (delta && IsPad(priv) && priv->oldProximity == ds->proximity)
> >> +     {
> >> +             DBG(10, priv, "Left touch strip scroll delta = %d\n", delta);
> >> +             fakeButton = getWheelButton(delta, priv->striplup, 
> >> priv->stripldn,
> >> +                                         priv->strip_keys[0+1], 
> >> priv->strip_keys[1+1], &fakeKey);
> >
> > urgh, named defines for the indices please. which would prevent...
> >
> >> +             sendWheelStripEvent(fakeButton, fakeKey, pInfo, first_val, 
> >> num_vals, valuators);
> >> +     }
> >>
> >> -     if (!fakeButton && (!fakeKey || !(*fakeKey)))
> >> -             return;
> >> +     /* emulate events for right strip */
> >> +     delta = getScrollDelta(ds->stripy, priv->oldStripY, 0, TRUE, TRUE);
> >> +     if (delta && IsPad(priv) && priv->oldProximity == ds->proximity)
> >> +     {
> >> +             DBG(10, priv, "Right touch strip scroll delta = %d\n", 
> >> delta);
> >> +             fakeButton = getWheelButton(delta, priv->striprup, 
> >> priv->striprdn,
> >> +                                         priv->strip_keys[2+1], 
> >> priv->strip_keys[3+1], &fakeKey);
> >> +             sendWheelStripEvent(fakeButton, fakeKey, pInfo, first_val, 
> >> num_vals, valuators);
> >> +     }
> >>
> >> -     if (!fakeKey || !(*fakeKey))
> >> +     /* emulate events for relative wheel */
> >> +     delta = getScrollDelta(ds->relwheel, 0, 0, FALSE, FALSE);
> >> +     if (delta && IsCursor(priv) && priv->oldProximity == ds->proximity)
> >>       {
> >> -             /* send both button on/off in the same event for pad */
> >> -             xf86PostButtonEventP(pInfo->dev, is_absolute(pInfo), 
> >> fakeButton & AC_CODE,
> >> -                                  1, first_val, num_vals, 
> >> VCOPY(valuators, num_vals));
> >> +             DBG(10, priv, "Relative wheel scroll delta = %d\n", delta);
> >> +             fakeButton = getWheelButton(delta, priv->relup, priv->reldn,
> >> +                                         priv->wheel_keys[0+1], 
> >> priv->wheel_keys[1+1], &fakeKey);
> >> +             sendWheelStripEvent(fakeButton, fakeKey, pInfo, first_val, 
> >> num_vals, valuators);
> >> +     }
> >
> > confusion here, the relwheel has the same index as the lstrip? and the lring
> > the same as the rstrip?
> >
> > the diff is virtually so this may be correct, but having named defines with
> > comments would help greatly here.
> >
> If you're confused by two different arrays (strip_keys and wheel_keys)
> having different data at the same indicies, I think that's a sign of a
> more fundamental problem. Personally, I never understood the need for
> a distinction between strip_keys and wheel_keys. A single
> "scroll_keys" (or, as I'd rather name it, "scroll_actions") array that
> houses all the scroll events would make things a *lot* less confusing.
> Throw named indicies on top of that, and things get better still.

sorry, I really didn't notice the different array names.
/me goes to find a brown paper bag.

> If you don't mind, I'd like to move this work to a separate patch
> which does unification and named indicies.

of course, sorry for the noise.

Cheers,
  Peter

> 
> > Cheers,
> >  Peter
> >
> >>
> >> -             xf86PostButtonEventP(pInfo->dev, is_absolute(pInfo), 
> >> fakeButton & AC_CODE,
> >> -                                  0, first_val, num_vals, 
> >> VCOPY(valuators, num_vals));
> >> +     /* emulate events for left touch ring */
> >> +     delta = getScrollDelta(ds->abswheel, priv->oldWheel, MAX_PAD_RING, 
> >> TRUE, FALSE);
> >> +     if (delta && IsPad(priv) && priv->oldProximity == ds->proximity)
> >> +     {
> >> +             DBG(10, priv, "Left touch wheel scroll delta = %d\n", delta);
> >> +             fakeButton = getWheelButton(delta, priv->wheelup, 
> >> priv->wheeldn,
> >> +                                         priv->wheel_keys[2+1], 
> >> priv->wheel_keys[3+1], &fakeKey);
> >
> >> +             sendWheelStripEvent(fakeButton, fakeKey, pInfo, first_val, 
> >> num_vals, valuators);
> >>       }
> >> -     else
> >> +
> >> +     /* emulate events for right touch ring */
> >> +     delta = getScrollDelta(ds->abswheel2, priv->oldWheel2, MAX_PAD_RING, 
> >> TRUE, FALSE);
> >> +     if (delta && IsPad(priv) && priv->oldProximity == ds->proximity)
> >>       {
> >> -             sendAction(pInfo, 1, fakeKey, 
> >> ARRAY_SIZE(priv->wheel_keys[0]),
> >> -                        first_val, num_vals, valuators);
> >> -             sendAction(pInfo, 0, fakeKey, 
> >> ARRAY_SIZE(priv->wheel_keys[0]),
> >> -                        first_val, num_vals, valuators);
> >> +             DBG(10, priv, "Right touch wheel scroll delta = %d\n", 
> >> delta);
> >> +             fakeButton = getWheelButton(delta, priv->wheel2up, 
> >> priv->wheel2dn,
> >> +                                         priv->wheel_keys[4+1], 
> >> priv->wheel_keys[5+1], &fakeKey);
> >> +             sendWheelStripEvent(fakeButton, fakeKey, pInfo, first_val, 
> >> num_vals, valuators);
> >>       }
> >>  }
> >>
> >> --
> >> 1.7.7.3
> 
> Jason
> 
> ---
> Day xee-nee-svsh duu-'ushtlh-ts'it;
> nuu-wee-ya' duu-xan' 'vm-nvshtlh-ts'it.
> Huu-chan xuu naa~-gha.
> 

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create 
new or port existing apps to sell to consumers worldwide. Explore the 
Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join
http://p.sf.net/sfu/intel-appdev
_______________________________________________
Linuxwacom-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linuxwacom-devel

Reply via email to