Bug 57019 – JavaInstance should not use jvalue in its API

• Component: WebCore Misc.
• Status: RESOLVED FIXED

[Steve Block]

The JavaInstance API should represent a generic Java object and should be independent of JNI, to give embedders flexibility in how a Java object is provided. This means that the API should use a new JavaValue type to represent a Java value, rather than the JNI jvalue type.

[Steve Block]

See bug 55383.

[Steve Block]

This bug tracks fixing only the V8 version of JavaInstance.

The JSC version can not be fixed until Bug 57023 is fixed.

[Steve Block]

Created attachment 87296 [details]
Patch

[Andrei Popescu]

LGTM

View in context: https://bugs.webkit.org/attachment.cgi?id=87296&action=review

> Source/WebCore/bridge/jni/v8/JavaInstanceV8.cpp:76
> +    jvalue* jvalueArgs = new jvalue[numParams];

Is there a scoped_pointer for arrays?

> Source/WebCore/bridge/jni/v8/JavaInstanceV8.cpp:78
> +        jvalueArgs[i] = javaValueToJvalue(args[i]);

Maybe add a comment to say that args is guaranteed to have at least numParams elements.

[Steve Block]

> Is there a scoped_pointer for arrays?
Yes - OwnArrayPtr - fixed.

> Maybe add a comment to say that args is guaranteed to have at least numParams elements.
I've added a comment in the header. Note that this is not a new requirement, as previously the array was passed straight through to callJNIMethod().

[Steve Block]

Created attachment 87413 [details]
Patch

[Jeremy Orlow]

Comment on attachment 87413 [details]
Patch

r+ based on Andrei's review

[WebKit Commit Bot]

Comment on attachment 87413 [details]
Patch

Clearing flags on attachment: 87413

Committed r82361: <http://trac.webkit.org/changeset/82361>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.