572494 - removal of JSObjectOps::defaultValue

Status: RESOLVED FIXED

(Core :: JavaScript Engine, enhancement)

Description

[Igor Bukanov]

See 572411 comment 0 and comment 3 for details.

Comment 1

[Igor Bukanov]

Attachment: v1

The patch removes JSObjectOps::defaultValue and moves the logic behind the single custom implementation from xml_defaultProperty into xml_convert. The patch also fixes potentially problematic code in TraceRecorder::guardNativeConversion where for a non-native object a guard added only to check if JSObjectOps::defaultValue is default one without checking that JSClass::convert is a stub one.

Ideally it would be nice to allow for JSClass.convert to take a role of JSObjectOps::defaultValue. The idea is to make the JS_ConvertStub to call toString(), not valueOf("string") for a string convert hint so the switch over the convert hint in DefaultValue would be necessary only of JSClass::convert fails to produce a primitive value. But that is for another bug and for a discussion in the newsgroup.

Comment 2

[Jason Orendorff [:jorendorff]]

Comment on attachment 455214 [details] [diff] [review]
v1

Looks great.

Comment 3

[Brendan Eich [:brendan]]

(In reply to comment #1)
> Ideally it would be nice to allow for JSClass.convert to take a role of
> JSObjectOps::defaultValue. The idea is to make the JS_ConvertStub to call
> toString(), not valueOf("string") for a string convert hint so the switch over
> the convert hint in DefaultValue would be necessary only of JSClass::convert
> fails to produce a primitive value. But that is for another bug and for a
> discussion in the newsgroup.

This sounds great, but the newsgroups have been spammed and I may have missed your post. Also, bug on file yet?

/be

Comment 4

[Igor Bukanov]

http://hg.mozilla.org/tracemonkey/rev/ab518f91a9bd

Comment 5

[Robert Sayre]

http://hg.mozilla.org/mozilla-central/rev/ab518f91a9bd