716069 - GC: missing barriers in NewObjectCache::newObjectFromHit

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Terrence Cole [:terrence]]

We should replace these tiny memcpys with direct assignment so that we trigger the barriers automatically through the operator= set in the HeapPtr<>.

Comment 1

[Terrence Cole [:terrence]]

Attachment: v0

Comment 2

[Terrence Cole [:terrence]]

Attachment: v1: a friendlier version

I had forgotten that I haven't landed the patch to put JSGC_GENERATIONAL in tree yet, so I missed the need to friend JSObject and NewObjectCache.

Comment 3

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

Comment on attachment 588283 [details] [diff] [review]
v1: a friendlier version

Review of attachment 588283 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/jsobj.h
@@ +1677,5 @@
>  
>    private:
>      inline bool lookup(Class *clasp, gc::Cell *key, gc::AllocKind kind, EntryIndex *pentry);
>      inline void fill(EntryIndex entry, Class *clasp, gc::Cell *key, gc::AllocKind kind, JSObject *obj);
> +    static inline void copyCachedToObject(JSObject *tgt, JSObject *src);

Please use dst for the name. It's more standard.

::: js/src/jsobjinlines.h
@@ +1590,5 @@
>      return fill(entry, clasp, type, kind, obj);
>  }
>  
> +inline void
> +NewObjectCache::copyCachedToObject(JSObject *tgt, JSObject *src) {

Use dst here too. Also, the { brace goes on its own line.

Comment 4

[Terrence Cole [:terrence]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/72c55496b6e9

Comment 5

[Marco Bonardo [:mak]]

https://hg.mozilla.org/mozilla-central/rev/72c55496b6e9