Don't interpret binary files as HTML

[jdm]

We want to avoid interpreting streams of binary bytes as HTML, since that's silly and easily breaks our layout code that is unprepared for such things. In the code that receives the HTTP headers for a page (ParserContext::headers_available) we should check for the Content-Type header application/octet-stream and ignore the forthcoming response body if the header matches. We can replace it with an informative message in the same way that we create a special document for images in that method.

Testcase: firefox.com and press the download button
Code: components/script/dom/servohtmlparser.rs

[SimonSapin]

Rather than a black list with application/octet-stream, we might want a white list of supported MIME type (and possibly some sniffing. We know how to deal with text/html, text/plain, some image MIME type, etc, and everything unknown should be treated like application/octet-stream.

[sgmenda]

is it possible a n00b like me can fix it

[jdm]

@s142857 I think you should give it a shot :)

[sgmenda]

could you guide me @jdm

[jdm]

First step is to find the headers_available method of ParserContext. Next step is to get the Content-Type header from the headers in the metadata value. Final step is to check its value and return a fake document if the content-type is not supported.

[jdm]

Fixed by #7956.