1364051 - Barriered pointer wrappers currently get a copy constructor by default

Status: RESOLVED FIXED

(Core :: JavaScript: GC, enhancement)

Description

[Jon Coppeard (:jonco)]

Attachment: barrier-copy-constructors

These wrapper classes may need to add barriers to this operation so we don't want a compiler generated copy constructor from the superclass being used if they forget to add an their copy constructor.  

For example, HeapSlot currently gains a copy constructor this way.  This is not a problem but it was surprising to me.

This patch forces subclasses to opt-in if they want a copy constructor.

Comment 1

[Steve Fink [:sfink] [:s:]]

Comment on attachment 8866743 [details] [diff] [review]
barrier-copy-constructors

Review of attachment 8866743 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/gc/Barrier.h
@@ +326,5 @@
>      explicit BarrieredBase(const T& v) : value(v) {}
>  
> +    // BarrieredBase subclasses cannot be copy constructed by default.
> +    BarrieredBase(const BarrieredBase<T>& other) = default;
> +

How does this work? I was expecting to see = delete. Is this some dark magic?

Comment 2

[Jon Coppeard (:jonco)]

(In reply to Steve Fink [:sfink] [:s:] from comment #1)
This gives BarrieredBase the default copy constructor but makes it protected.

Comment 3

[Steve Fink [:sfink] [:s:]]

Comment on attachment 8866743 [details] [diff] [review]
barrier-copy-constructors

Review of attachment 8866743 [details] [diff] [review]:
-----------------------------------------------------------------

Ah, ok. Thanks.

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/a64e7f9efef2
Make WriteBarrieredBase copy constructor protected r=sfink

Comment 5

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/a64e7f9efef2