305935 - allow Web authors to "say" that onunload, etc. handlers are fastback-safe

Status: NEW

(Core :: DOM: Navigation, defect)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

One of the things that we check for before saving a presentation (in particular,
nsDocument::CanSavePresentation calls
nsEventListenerManager::HasUnloadListeners) is that the page doesn't have any
onunload or onbeforeunload listeners.  This could provide a very frustrating
experience for Web developers who:
 * need to use onunload or onbeforeunload for some reason, and
 * care about the performance of their pages in Gecko,
since they would be unable to make their pages get the benefits of fastback.  We
should care about helping Web developers who care about Gecko, and we certainly
don't want to make them mad at us.  This is a little risky, though, since it
introduces a situation that authors aren't likely to test very well, so I have
some mixed feelings about whether we should do it at all.

One possible mechanism for doing this is to consider the presence of onpagehide
and/or onpageshow handlers to be a statement from the author that the
onunload/onbeforeunload handlers don't make the page unusable.  However, we
should make sure that such handlers coming from extensions/chrome don't cause
this change (at the very least by ensuring that there's a way for
extensions/chrome to add listeners that doesn't affect this test).

(I thought we were also checking for onsubmit, but we don't seem to be.)