Sometimes, the viewer did not render for PDF files in an iframe,
because document.readyState not reaching "complete".
Deferring window.stop() until the root element is placed in the
document fixes the problem (typically a few ten milliseconds).
The Chrome extension activates PDF.js by inserting the script tags
in a document whose URL and location origin is identical to the PDF
file.
Because of this, the path './images/' was resolved relatively to the
location of the PDF file instead of the extension.
To fix this, the IMAGE_DIR constant is moved outside the local scope,
to allow extensions/chrome/insertviewer.js to override the value.
Originally, the IMAGE_DIR variable was a global variable, but commit
f8f4b3f45d moved the global variable
to the local scope, causing the extension to malfunction.
Impact: low, the only consequence is that some rarely used images
were not visible.
Trivial test:
At the center of page 2, the annotation icon
(images/annotation-comment.svg) should be visible:
http://linorg.usp.br/CTAN/macros/latex/contrib/pdfcomment/doc/pdfcomment.pdf
Full list feature changes in this commit:
- Support for iframes
- Switched to content-type (MIME) detection instead of hard-coding a
case-sensitive check for the .PDF extension
- The PDF's original URL is visible in the omnibox
- Support for incognito mode
Note: PDF viewer is disabled for the file:// + incognito
combination, because it's currently impossible to get the combination
to work.
See https://github.com/mozilla/pdf.js/pull/3017#issuecomment-15693432
