1185527 - nsScriptErrorWithStack shadows the refcount in nsScriptError

Status: RESOLVED FIXED

(Core :: XPConnect, defect)

Comment 1

[Bobby Holley (:bholley)]

Sorry for missing this one too. :-(

Comment 2

[Andrew McCreight [:mccr8]]

It isn't a big deal. The virtual AddRef/Release ensure we only use the right one. This is also the sort of tedious non-local property that should be caught by static analysis (as we should get with bug 602122).

The fix for this would be something like change nsScriptError into a nsScriptErrorBase class that does not implement refcounting, but is otherwise the same as nsScriptError. nsScriptErrorWithStack stays the same, except that it inherits from nsScriptErrorBase instead. Then you define a new nsScriptError class which inherits from nsScriptErrorBase, but doesn't do anything except implement refcounting. An example of this is nsArrayBase, nsArray, and nsArrayCC.

Comment 3

[Andrew McCreight [:mccr8]]

I can fix this unless Michael wants to.

Comment 4

[Nika Layzell [:nika] (ni? for response)]

(In reply to Andrew McCreight [:mccr8] from comment #3)
> I can fix this unless Michael wants to.

Go ahead.

Comment 5

[Andrew McCreight [:mccr8]]

Attachment: Split out a refcountless base class for nsScriptError.

This avoids shadowing the refcount if nsScriptErrorWithScript inherited from nsScriptError.

Mostly just a lot of search and replace.

try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=85a79af5458b

Comment 6

[Gabor Krizsanits (INACTIVE)]

Comment on attachment 8639107 [details] [diff] [review]
Split out a refcountless base class for nsScriptError.

Review of attachment 8639107 [details] [diff] [review]:
-----------------------------------------------------------------

LGTM

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/20afa46d73c1

Comment 8

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/20afa46d73c1