Codecov Report

Merging #486 into master will decrease coverage by 0.01%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##           master     #486      +/-   ##
==========================================
- Coverage    89.9%   89.89%   -0.02%     
==========================================
  Files         134      134              
  Lines        5241     5244       +3     
  Branches      905      906       +1     
==========================================
+ Hits         4712     4714       +2     
- Misses        529      530       +1

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1d0f782...a07f816. Read the comment docs.

@jccr - What's the easiest way to test this with a standalone HTML page?

@robertknight
Give this HTML a shot: https://gist.github.com/JCCR/27f6dddf0f4341d68654409955361bb3
Press any key to "flip" the page, then you should be able to reproduce the issue.

Possibly related: #487

The problem here actually happened earlier before Adder#target was called. The input rect is supposed to be in coordinates relative to the document element, but the conversion from viewport coordinates to "document" coordinates in mapViewportRectToDocument is incorrect because it only takes into account the scrollLeft, scrollTop of the document element (via window.pageXOffset and window.pageYOffset) but not the top/left position of the document element.

Additionally, code in Adder#target then tries to adjust the adder to fit within the coordinates of the viewport but doesn't allow for the possibility of a document element with a top/left other than 0,0. In #487 I outline what I think will be the best solution which is to use viewport coordinates everywhere, but I need to investigate this in a little more detail. I'll get back to you on Monday.

I've investigated a couple of options for addressing the issue described in #486 (comment) and the related issue described in #487 and have created a follow-up PR in #493 which addresses both and adds some tests.

If you get a chance, could you please try out the PR (#493) and let me know if that works correctly with Readium?