1355533 - Try to avoid loading AboutPocket.jsm in the content process at startup

Status: RESOLVED WORKSFORME

(Firefox :: Pocket, enhancement)

Description

[Andrew McCreight [:mccr8]]

I'm not entirely sure what this means, but Gijs seemed to think it would be possible to stop loading this .jsm in the content process, which can save memory.

(In reply to :Gijs from comment #8)
> I'm confused, actually, it doesn't look like the panel in which we load the
> pocket iframe is remote, so that would allow removing this (ie all it does
> is register about:pocket-signup and about:pocket-saved, which are never used
> in the content process AIUI...) - but it probably *should* ideally be
> remote, except that would probably require its own set of work.
> 
> Regardless, the only reason this seems to need a jsm right now is to
> facilitate loading its own content pages. It could (and should) just use
> other URLs for that, they don't need to be about: URLs at all, I think. Per
> bug 1236755 they shouldn't be chrome URLs, but resource: URLs should be
> unprivileged enough for what we want here, maybe/hopefully?

Comment 1

[:Gijs (he/him)]

Mark, you wrote the patch in bug 1291641 that copied stuff to the content process here. Can you elaborate on why this was necessary?

Comment 2

[Mark Banner (:standard8)]

The about: pages have to be registered in both processes if you want to use it in the content process. See https://developer.mozilla.org/Firefox/Multiprocess_Firefox/Which_URIs_load_where for more information.

I can't speak to if using resource URLs would be equivalent on the security front.

Comment 3

[:Gijs (he/him)]

(In reply to Mark Banner (:standard8) from comment #2)
> The about: pages have to be registered in both processes if you want to use
> it in the content process. See
> https://developer.mozilla.org/Firefox/Multiprocess_Firefox/
> Which_URIs_load_where for more information.

I understand that. I don't think we load the about: URIs in the content process today, though. Did I miss something?

Comment 4

[Mark Banner (:standard8)]

(In reply to :Gijs from comment #3)
> I understand that. I don't think we load the about: URIs in the content
> process today, though. Did I miss something?

Hmm, assuming iframes default to the chrome process (which it looks like they might do), then it does indeed look like we're loading it in the chrome process:

https://dxr.mozilla.org/mozilla-central/rev/f914d40a48009c5acd1093e9939cc0ec035696dd/browser/extensions/pocket/content/main.js#540

I probably didn't check too deeply when I wrote the patch originally, as it was a "standard" set of conversions.

However, the question could be raised of should we be loading it in the content process...

Comment 5

[:Gijs (he/him)]

We fixed this, probably in bug 1514724