Re: Missing test assertion, reconstructing (searchRange, entrySelector and rangeShift).

On Nov 29, 2011, at 1:21 PM, Chris Lilley wrote:

> On Tuesday, November 29, 2011, 3:56:13 PM, Tal wrote:
> 
> TL> On Nov 29, 2011, at 8:49 AM, Chris Lilley wrote:
> 
>>> On a third hand, the assertion could be broadened to include authoring tools which perform round-trip or bidirectional conversion,and would be testable (perhaps the bitwise-identical authoring tool tests already cover this).
> 
> TL> The Authoring Tool cases invalidsfnt-searchrange-001,
> TL> invalidsfnt-entryselector-001 and invalidsfnt-rangeshift-001 have
> TL> invalid values for these fields for SFNT input testing.
> 
> Good, but that seems to be a different test. Its about conversion from sfnt to woff.
> 
> The quoted spec text is about conversion from woff (which does not explicitly store these values) to sfnt (so they need to be correctly calculated).

Ah, yes. I can't think of a way to provide test cases for this conversion. The only thing that I have come up with so far is to generate WOFFs from an SFNT that has one of the fields set incorrectly and has a head checkSumAdjustment that is calculated with the invalid field value. When a UA recreates the SFNT data from the WOFF, it would set the fields correctly and thereby create SFNT data that is a mismatch against the head checkSumAdjustment. But, we don't require UAs to check the head checkSumAdjustment so this won't work. Any other ideas?

Tal

Received on Tuesday, 29 November 2011 18:37:42 UTC