endless loop printing a line

> My print routine does these basic steps:
> - open a window on my #printSpec
> - get a new print job with a MockMedium on a ScalingWrapper on the print
> window
> - ask the medium to display
> - send the completed print job to the printer
>
> When the scale of the print job is extreme enough (like when printing a
> very large visual on too small a paper size), display primitives (such
> as HostPrinterGraphicsContext>>primDisplayLineFrom:to:) can fail because
> the coordinates exceed the number ranges allowed in the primitives. The
> prim failure code leads to #displayClippedPolygon:, which begins the
> endless loop.
>
> My partial work around is to override #displayClippedPolygon: to raise a
> Notification if pointCollection cannot be clipped (at the end of the
> method, is clippedPointCollection asArray = pointCollection asArray?). I
> have to handle this notification in the #displayOn: method for my main
> visual object to abort the endless loop (I abort when unable to clip the
> same pointCollection more than once consecutively, because I'm unsure
> whether any paths through this code would resolve after a single
> clipping failure).
> However, this is only a partial work around because there remain the
> coordinate limitations of the graphics primitives. When will those
> limits be raised? Are they higher in the 64 bit VM?
>
> Dave Stevenson
> [hidden email]
>
>
> ------------------------------------------------------------------------
> *From:* Dave Stevenson <[hidden email]>
> *To:* [hidden email]
> *Sent:* Mon, May 2, 2011 3:16:31 PM
> *Subject:* [vwnc] endless loop printing a line
>
> I've got an endless loop in VW 7.6 trying to print the border of a
> widget to a scaled MockMedium. In particular, the problem happens when
> printing the bottom of line of the border. Here is the repeating part of
> the stack:
> ScreenGraphicsContext(GraphicsContext)>>displayClippedPolygon:
> ScreenGraphicsContext>>roundedDisplayPolygon:at:
> ScreenGraphicsContext>>primDisplayPolygon:at:
> ScreenGraphicsContext(GraphicsContext)>>displayPolygon:at:
> ScreenGraphicsContext(GraphicsContext)>>displayPolygon:
> ScreenGraphicsContext(GraphicsContext)>>displayClippedPolygon:
> ...
>
> And the relevant data:
> pointCollection = #(3@25 3@24 797@24 797@25)
> clippingBounds = 0@0 corner: 800@25
> Note that the pointCollection is aligned with the bottom of the
> clippingBounds.
>
> At issue is an apparent mismatch between Rectangle and GraphicsContext
> as to whether the bottom points (3@25 or 797@25) are inside the
> clippingBounds. Rectangle thinks it is out:
> Rectangle>>containsPoint:
> optimized [] in GraphicsContext>>isPolygon:inside:
> Array(ArrayedCollection)>>allSatisfy:
> ScreenGraphicsContext(GraphicsContext)>>isPolygon:inside:
> ScreenGraphicsContext(GraphicsContext)>>displayClippedPolygon:
>
> because it checks for aPoint < corner, but aPoint y is not less than
> corner y, so it answers false:
>
> Rectangle>>containsPoint: aPoint
> "Answer whether the argument aPoint is within the receiver."
> ^origin <= aPoint and: [aPoint < corner]
>
> Point>>< aPoint
> "Answer whether the receiver is 'above and to the left' of the
> argument, aPoint."
> | p |
> p := aPoint asPoint.
> ^x < p x and: [y < p y]
>
> But GraphicsContext thinks it is in:
> ScreenGraphicsContext(GraphicsContext)>>clipPolygon:toBottom:
> ScreenGraphicsContext(GraphicsContext)>>displayClippedPolygon:
>
> because it checks for point y <= aNumber, and they are indeed equal:
>
> GraphicsContext>>clipPolygon: pointCollection toBottom: aNumber
> ^self
> clipPolygon: pointCollection
> insideBlock: [:point | point y <= aNumber]
> intersectionBlock: [:point1 :point2 | aNumber - point1 y * (point2 x -
> point1 x) // (point2 y - point1 y) + point1 x @ aNumber]
>
> So the difference of #< in #containsPoint: and #<= in
> #clipPolygon:toBottom: means the pointCollection never gets clipped, and
> we happily spin the cpu. The relevant code appears unchanged in the 7.8
> beta.
>
> Any recommendations as to how to fix or work around this?
>
> Thanks,
>
> Dave Stevenson
> [hidden email]
>
>
>
> _______________________________________________
> vwnc mailing list
> [hidden email]
> http://lists.cs.uiuc.edu/mailman/listinfo/vwnc