517749 - removal of weakRoots.newborn = null

Status: RESOLVED FIXED

(Core :: JavaScript Engine, enhancement)

Description

[Igor Bukanov]

Currently in a few places the code clears the newborn object on failures like in

cx->weakRoots.newborn[GCX_OBJECT] = NULL;

The idea behind this code is to prevent the GC to see partially initialized GC things like objects without the private data or without particular property. 

Since the invention the local roots that are populated in NewGCThing such zeroing of newborn is not enough and the code is updated to properly deal in finalizers the missing values. Thus such clearing of newborn just adds to code bloat and should be removed.

Comment 1

[Igor Bukanov]

Attachment: v1

The patch removes weakRoots.newborn = null assignments. It also converts few related JS_PUSH/POP_TEMP_ROOTS into JSAutoTempValueRooter.

Comment 2

[Jeff Walden [:Waldo]]

I'm all in favor of removing these mostly-unexplained lines, but going so far as to request wanted? on it seems a bit much to me.  :-)

Comment 3

[Igor Bukanov]

(In reply to comment #2)
> but going so far as to request wanted? on it seems a bit much to me.  :-)

This simplifies patching for bug 517199.

Comment 4

[Igor Bukanov]

Attachment: v2

synchronizing patch with the tip

Comment 5

[Blake Kaplan (:mrbkap) (inactive)]

Comment on attachment 402070 [details] [diff] [review]
v2

I'm assuming that returning true and false from JSBool-returning functions is OK for now.

Comment 6

[Brendan Eich [:brendan]]

Returning true and false from JSBool return-typed functions is good for now and always. The JSBool taint is there cuz of either (a) API compat; (b) ABI issues in nanojit.

Still, we can use true and false instead of JS_TRUE and JS_FALSE, so we should. The ugly part we can't avoid is unwinding from ABI or API downcall such that the compiler sees a JSBool -> bool return value conversion, which requires !! or equivalent.

/be

Comment 7

[Igor Bukanov]

https://hg.mozilla.org/tracemonkey/rev/1980589f09ad

Comment 8

[Robert Sayre]

http://hg.mozilla.org/mozilla-central/rev/1980589f09ad

Comment 9

[Robert Sayre]

http://hg.mozilla.org/releases/mozilla-1.9.2/rev/86f1efb05a93