URIPatternList.parse mishandles * in domain

[mattmccutchen]

IIUC, the following line of noscriptService.js is responsible for * in force-HTTPS glob patterns:

Code: Select all

return '^' + p.replace(/^([^\/:]+:\/*)\*/, "$1[^/]*").replace(/\*/g, '.*?');

The first replacement appears to make the * in *.mattmccutchen.net not match slashes. But this doesn't work, because its output is affected by the second replacement. E.g., with *.mattmccutchen.net in the force-HTTPS list, http://a/.mattmccutchen.net is forced to HTTPS. Reversing the order of the replacements should fix the problem.

[Giorgio Maone]

Ouch!

Reversing the order of the replacements should fix the problem.

It's not enough, actually.
Changed into

Code: Select all

return '^' + p.replace(/\*/g, '.*?').replace(/^([^\/:]+:\/*)\.\*/, "$1[^/]*");

(notice the extra dot being elided).
Fixed in latest development build, thanks.

[GµårÐïåñ]

I had missed this, wow such a subtle but incredible difference. Sweet.