This shows up in a bunch of reftests. I have no idea why yet.

What does the spec say about float-to-int casting?

(In reply to Benoit Jacob [:bjacob] from comment #2) > What does the spec say about float-to-int casting? Section 4.9, "Floating-integral conversions", paragraph 1: "... The conversion truncates; that is, the fractional part is discarded. The behavior is undefined if the truncated value cannot be represented in the destination type," Is this what you were looking for?

We probably want to use Skia's built in macros for converting. SkScalar{Round,Ceil,Floor}ToInt(), depending on what behaviour we explicitly want.

Thanks Botond, so IIUC this is rounding towards zero.

The right thing to do here is likely to something like cairo's _cairo_fixed_from_double: http://cgit.freedesktop.org/cairo/tree/src/cairo-fixed-private.h#n111 That way we can be more correct and have more speed.

I haven't tested this but it should be faster and more correct than previous patch and upstream skia code.

This patch seems to cause assertion failures on some tests...

(In reply to Jeff Muizelaar [:jrmuizel] from comment #9) > This patch seems to cause assertion failures on some tests... Turns out these seemed to be caused by a conflict with the SkScalarToFDot6 macro in the same file.

I already fixed that issue locally and also on the graphics branch which I'm using to track progress: https://tbpl.mozilla.org/?tree=Graphics

Mike and I agreed on "SkScalarRoundToFDot6()"

Rubber stamp for importing this from upstream?

Oops, a bunch of my other patches got caught up in that.

OK, I think this is correct now: https://tbpl.mozilla.org/?tree=Try&rev=247713774a59

https://hg.mozilla.org/integration/mozilla-inbound/rev/ce39df5dccdb https://hg.mozilla.org/integration/mozilla-inbound/rev/c8d299030e1d

https://hg.mozilla.org/mozilla-central/rev/ce39df5dccdb https://hg.mozilla.org/mozilla-central/rev/c8d299030e1d