IIRC Blink uses a different mechanism (called "separate worlds") to allow extensions to interact with content, whereas we use a separate global + xrays. So this likely will be a problem for WebExtensions, and we'll presumably need a sandboxOption to opt into the old behavior.
On Thu, Jan 11, 2018 at 8:16 AM, Gijs Kruitbosch <gijskruitbo...@gmail.com> wrote: > Based on what Cameron wrote, other browsers already return false if things > get mixed, so hopefully the WebExtensions side of the problem is still > limited? > > ~ Gijs > > On 11/01/2018 16:12, Tom Schuster wrote: > >> This could be an issue for WebExtensions as well. I think the >> contentscript >> sandbox runs in a different compartment. >> >> On Thu, Jan 11, 2018 at 3:58 PM, Gijs Kruitbosch < >> gijskruitbo...@gmail.com> >> wrote: >> >> On 11/01/2018 05:29, Cameron McCormack wrote: >>> >>> For use in the meantime, I just landed bug 1428531 on inbound, which adds >>>> a new chrome-only static method "isInstance" to Web IDL defined >>>> interfaces, >>>> so you can write for example: >>>> >>>> Document.isInstance(otherWindow.document) >>>> >>>> So that we don't have even more call sites we need to fix up once we >>>> start looking at bug 1360715, please try to use isInstance rather than >>>> instanceof if you need to do cross-context DOM object tests from chrome >>>> JS. >>>> >>>> >>> Assuming we're happy to update all sites (not only sites where we're sure >>> the DOM instance is sometimes cross-context), this type of thing is >>> relatively easy to automatically rewrite with an xpcshell script like the >>> ones Florian has been using to update other conventions (e.g. use of >>> Services.prefs instead of manually Cc[...].getService(Ci.nsIPrefBranch) >>> etc. ). If we rewrite the extant callsites, we can add eslint checks that >>> warn and can do similar rewrites for people, which will be better at >>> avoiding new callsites than just convention (which people are liable to >>> forget in a few days/weeks/months), and would make the transition >>> relatively "easy". >>> >>> ~ Gijs >>> >>> _______________________________________________ >>> dev-platform mailing list >>> dev-platform@lists.mozilla.org >>> https://lists.mozilla.org/listinfo/dev-platform >>> >>> > _______________________________________________ > dev-platform mailing list > dev-platform@lists.mozilla.org > https://lists.mozilla.org/listinfo/dev-platform > _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
