945828 - Work around bug 945756

Status: RESOLVED FIXED

(Core :: XPConnect, defect)

Description

[Andy Wingo [:wingo]]

Bug 945756 is a case in which a local being allocated on the scope chain instead of the frame is creating some kind of cycle that is causing an xpcshell test to core at the end.  This only happens in debug mode, but still.

It's possible to get that local onto the scope chain with the code as-is, by aliasing it via a nested procedure, but bug 927782 will put it and all locals in all generators on the scope chain.  Therefore for generators work to proceed, we need to work around bug 945756.  The attached patch does so by avoiding making a local at all.  Gross.

Comment 1

[Andy Wingo [:wingo]]

Attachment: Work around bug 945756

Comment 2

[Andy Wingo [:wingo]]

Comment on attachment 8341812 [details] [diff] [review]
Work around bug 945756

Setting r? to luke, but really I have no idea who would be appropriate -- feel free to steal, I reckon

Comment 3

[Luke Wagner [:luke]]

Comment on attachment 8341812 [details] [diff] [review]
Work around bug 945756

Oh yeah, these bugs.  I hit like 3 during the aliasedvar work.

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/4cbb184bf6cb

Comment 5

[Ryan VanderMeulen [:RyanVM]]

Backed out for Linux64 debug xpcshell crashes.
https://hg.mozilla.org/integration/mozilla-inbound/rev/e98e65fd7203

https://tbpl.mozilla.org/php/getParsedLog.php?id=31403037&tree=Mozilla-Inbound

Comment 6

[Andy Wingo [:wingo]]

As mentioned in https://bugzilla.mozilla.org/show_bug.cgi?id=945813#c4, this was an erroneous backout.  Re-setting checkin-needed.

Comment 7

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/48ee35b92973

Comment 8

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/48ee35b92973