RFR: 8284184: Crash in GraphicsContextJava::drawLinesForText on https://us.yahoo.com/

[Kevin Rushforth]

On Sat, 2 Apr 2022 08:29:43 GMT, Jay Bhaskar <jbhaskar at openjdk.org> wrote:

> Issue: Floating point overflow , when making end point for line drawing as FloatPoint endPoint = startPoint + FloatPoint(widths.last(), 0);
> Solution: traverse widths to calculate end point and increment start point.

I instrumented the code, and discovered that in the failing case, the list of `widths` is empty. I see that other implementations of GraphicsContext have a check for this at the beginning of the method, and do an early return in that case. That seems a good idea regardless. So I recommend adding this around line 345:


    if (widths.size() == 0)
        return;


That alone is sufficient to fix this bug.

As for the change to loop over the widths array, I'm not sure about that part. I see code in some of the ports that treat the widths array as pairs of start / end values, but that doesn't match what I see in a simple strike-through test case where there is only a single width value.

For the case of a simple underline (with or without dashing), the widths array is of length 2 with the first value being 0.0, which tends to support the definition used by the other ports. This will need more analysis.

So I think the simplest (and safest) fix is to just add the check for size == 0, but leave the drawing part of the code unchanged, and file a follow-up bug to investigate.

-------------

PR: https://git.openjdk.java.net/jfx/pull/765