219055 - Delay creation of script object for EcmaError until execution of script catch

Status: VERIFIED FIXED

(Rhino Graveyard :: Core, enhancement)

Description

[Igor Bukanov]

Currently Rhino always creates script object representing EcmaError for each
EcmaError it throws.  But such object is only necessary if script contains a
catch statements that catches the exception. Moreover, if creation of such
object throws an exception on its own which is quite possible if script mutates
Error constructors, then Java application will get the secondary exception which
is rather inconvenient.

In addition, such delayed creation of script object would allow to simplify
internal API since it removes need to pass a scope object argument if such
object currently is only necessary to construct script error instances.  Such
situation is present currently in code that implements script compilation and
various ScriptRuntime routines.

Comment 1

[Igor Bukanov]

Attachment: Implementation

The essential patch part moves creation of script object represented by
NativeError to ScriptRuntime.getCatchObject and modifies EcmaError constructor
to take explicit error name and message that will be used when constructing
NativeError. The old form is preserved for compatibility but marked as
deprecated. 

The rest of the patch updates code not to pass scope argument when throwing
EcmaError instances which allowed to simplify many usages of the error related
routines. In addition patch moved functions to create EcmaError instances from
NativeGlobal to ScriptRuntime to remove dependence on that class from optimizer
and regexps packages which could allow at some point in future (probably a
distant point due to backward compatibility) to make the class package private.

Comment 2

[Igor Bukanov]

I committed the changes.

Comment 3

[Phil Schwartau]

Rubber-stamp vrfy

Comment 4

[Igor Bukanov]

Trageting as resolved against 1.5R5