This function proceeds to push its cx and enters a compartment, so it can't be
depending on any compartment or callstack state of the cx it's using. The only
potential issue would then be reporting the error to the correct DOM window, but
this stuff is used only for chrome, where that doesn't matter. The safe JSContext
uses the same error reporter as JSMs and such, which is probably fine.
These things currently do a complicated refcounting dance to avoid destroying the
cx until all the consumers of it are gone. That stuff can mostly go away now that
we're just using the SafeJSContext, but DestroyCx also nulls out the global, so
we should make sure to keep that alive for anyone that might be using it.
This function proceeds to push its cx and enters a compartment, so it can't be
depending on any compartment or callstack state of the cx it's using. The only
potential issue would then be reporting the error to the correct DOM window, but
this stuff is used only for chrome, where that doesn't matter. The safe JSContext
uses the same error reporter as JSMs and such, which is probably fine.
These things currently do a complicated refcounting dance to avoid destroying the
cx until all the consumers of it are gone. That stuff can mostly go away now that
we're just using the SafeJSContext, but DestroyCx also nulls out the global, so
we should make sure to keep that alive for anyone that might be using it.
This doesn't do anything anymore. The compile options should generally carry
the right version through, with the exception of eval, which will end up using
the version of the running script anyway.
This test calls the version() shell command, and expects it to work like an
override, rather than the dumb compartment-mutator that I'm turning it into.
Let's just kill the test.
Sandboxes always default to JSVERSION_DEFAULT in the browser. But XPCShell sets
up a ContextCallback that does JS_SetVersion(cx, JSVERSION_LATEST) on every
context that gets created, including the ephemerial Sandbox JSContexts. Since
httpd.js runs in xpcshell and evaluates SJS in a sandbox, we've (somewhat
accidentally) supported JSVERSION_LATEST in SJS, which certain SJS files have
taken advantage of. Let's continue to support it explicitly.
Looks like cdleary added this back in bug 595691, as an equivalence to some
even-older-and-more-overly-cautious XBL code that was saving and restoring
the version across XBL calls. It doesn't seem like this should be an issue
anymore, and it's just a debugging aid to boot. Let's kill it.
This has lower precedence than 'overrides' and running script, and higher
precedence than the cx's version. We can migrate the API consumers who clearly
want something like this, which will eventually let us remove the override
mechanism.
A quick grep of Firebug indicates that it doesn't use this. And even if I
missed it, doing so from a debugger with the current semantics is a feature
we can't support going forward.
