igor%mir2.org
0659e0b714
Committing E4X parser and runtime proxies: see enhancement report 242805.
...
This is based on code contributed to Rhino by AgileDelta, Inc, www.agiledelta.com and in particular by
Ethan Hugg
Terry Lucas
Milen Nankov
John Schneider
Thanks!
2004-07-28 21:11:02 +00:00
igor%mir2.org
ef41e24058
Added ScriptableObject.equivalentValues to allow for custom equality operator support.
2004-07-28 15:23:11 +00:00
igor%mir2.org
d3d45e59da
Added Reference.has() for future use.
2004-07-28 15:21:02 +00:00
igor%mir2.org
8799f9f62b
Ignore any exception during script class initialization unless it is an instance of RhinoException. In this way lazily class loading or incompatible with the current JVMs jars will simply hide the script variables.
2004-07-28 15:20:08 +00:00
igor%mir2.org
e47e67719d
Always call Scriptable.has(name, start) as obj.has(name, obj) since the start argument is useless in Scriptable.has and in this way it is possible to add temporary assertions to the code checking this.
2004-07-28 13:27:39 +00:00
igor%mir2.org
cf9b20185b
Redesigning IdScriptable implementation.
...
The new logic introduces 2 independent id namespaces, instance ids and prototype ids. Typically there are very few instance properties and instance ids corresponds exactly to them. The prototype ids allows to define functions and other constant objects that typically found in prototype objects.
The new implementation takes advantage of this difference to provide very different API to implement fast mapping both for instance and prototype ids.
2004-07-28 11:40:38 +00:00
igor%mir2.org
f02c491af8
Introduction of defaultParentScope and defaultPrototype in ScriptableObject for lazily initialization of scope and prototype.
2004-07-28 09:52:29 +00:00
igor%mir2.org
a87bb4edcd
Making ScriptRuntime.typeError* public
2004-07-28 09:51:02 +00:00
igor%mir2.org
9ad83535e6
Allow to subclass NativeWith
2004-07-28 09:50:07 +00:00
igor%mir2.org
cbb0e19a28
Favour same type converssions during LiveConnect for better support of java methods taking instances of Scriptable
2004-07-28 09:49:19 +00:00
igor%mir2.org
5207d03135
1. Optimizing getClassPrototype and getDefaultValue in ScriptableObject.
...
2. Start of the changes for simpler to use IdScriptable implementation: new utility functions in IdFunction and removal of Context argument from IdScriptable.fillConstructorProperties.
2004-07-28 07:12:46 +00:00
igor%mir2.org
1a1653605f
Context.getArrayElements implementation is moved to ScriptRuntime which also defined now isArrayObject
2004-07-28 07:06:35 +00:00
igor%mir2.org
fa889e8215
Making next release to be 1.6 release 1
2004-07-12 14:26:30 +00:00
igor%mir2.org
2c3f8b4844
Renaming ObjArray.setReadOnly() to ObjArray.seal() to follow the rest of Rhino code.
2004-07-09 13:28:48 +00:00
igor%mir2.org
fe6620df87
Fixing comments text to reflect real state of code
2004-07-07 14:47:46 +00:00
igor%mir2.org
faa2cc87fe
Almost reverting the prevois commit: it was too preliminary.
2004-07-07 14:46:58 +00:00
igor%mir2.org
4588cccd39
Changing the way Java getter and setter are stored: a new special class ComplexValue is used to indicate a controllable value that may have associated getter/setter methods. In this way getter can be added after slot creation without the need to replace slots.
2004-07-06 14:37:56 +00:00
igor%mir2.org
80b96958c9
Token.ZERO and Token.ONE are used only by interpreter and moved to Interpreter as corresponding icode constants.
2004-07-05 15:08:34 +00:00
igor%mir2.org
3ffab68246
Removal of no-longer applicable comments refering to SM code
2004-07-05 15:07:25 +00:00
igor%mir2.org
cfec98bb95
Removal of Token.DOT in NodeTransformer: it is always processed in IRFactory.
2004-07-05 15:06:18 +00:00
igor%mir2.org
b0e98cbca8
When support for Context.FEATURE_MEMBER_EXPR_AS_FUNCTION_NAME is enables,
...
decompiler now does not turn them into assignment statements but keeps
original code with extension when printing function source. It allows for
better emulation of this JScript extension
2004-07-04 17:26:50 +00:00
igor%mir2.org
a084aaa65f
Simplified handling of function expression statements: parser no longer
...
generate synthetic expression statement nodes for them.
2004-07-04 14:13:47 +00:00
igor%mir2.org
15426336fd
Normilizing numeration of Token constants after removal of few no longer used
...
constants in the previous patches
2004-07-04 14:12:01 +00:00
igor%mir2.org
53864e64b8
Splitting code generation in interpreter and class compiler into separated statement-only and expression only cases to add more checks.
2004-07-02 15:36:13 +00:00
igor%mir2.org
b91ef7e8a2
Replacing generic Token.EXPRSTMT by Token.EXPR_VOID and Token.EXPR_RESULT and removal of Token.POP and Token.POPV. Now parser creates appropriate expression statements itself so there is no need to replace the time in NodeTransformer.
2004-07-02 13:11:06 +00:00
igor%mir2.org
4ad826a35e
Fixing bug 249471:
...
Fixing parseFloat logic and Context usage cleanup.
I changed number start detection code in parseFloat to check properly for lone
+/- while making the code IMO more redable.
In addtion patch removes unused Context parameter from various functions in
NativeGlobal including parseFloat which was remainder from pre IdFunction
code.
2004-07-02 06:55:21 +00:00
igor%mir2.org
4035bd2c91
Removal of Tiken.UNDEFINED: it was not used outside Interpreter
2004-07-01 16:40:44 +00:00
igor%mir2.org
f0ca03fe7c
Replacing state flags in TokenStream by separated boolean values to simplify
...
code and to remove flags that were set/cleared but never queried.
2004-07-01 14:52:22 +00:00
igor%mir2.org
21a1489b59
Change IRFactory to return/accept Node, not Object as the type for parse tree
...
nodes. Parser already assumes in a few places that such Node and it is not
worth to pretend otherwise.
2004-07-01 10:39:28 +00:00
igor%mir2.org
05a6149911
Fixing JavaDoc: the previous changes removed Context(ContextFactory)
...
constructor as factory is associated automatically with Context but JavaDoc
continied to refer to this constructor form.
2004-06-30 15:28:04 +00:00
igor%mir2.org
f32fceb8a1
Deprecation of ContextListener in favour of ContextFactory.Listener:
...
the later is stored in the factory and not in static variables.
For compatibility an instance of ContextFactory stored in static variables
is used to provide support for old Context.addContextListener etc. API.
The static instance of the factory is also used by Context.enter to create
Context instances providing simple way to customize Context: one just need
to initialize the global static factory.
2004-06-30 15:23:39 +00:00
igor%mir2.org
012fd05901
Fix menu enable/disable misbehavior.
2004-06-30 13:29:41 +00:00
igor%mir2.org
4cc8c9cd28
Fixing JDK 1.3 compatibility: TableModel.getRot() should not return null there.
2004-06-30 09:38:54 +00:00
igor%mir2.org
c1a3a327c2
Fixing scope problems with IdFunction when IdScriptable is top scope itself:
...
scope for IdFunction instances is determined through getTopScope,
not getParentScope() which can be null.
To prevent null scope initialization problems in future IdFunction constructor
now throws an exception if scope argument is null.
2004-06-29 15:43:01 +00:00
igor%mir2.org
d9fd2cc2b9
Make EvaluatorException(String) public for compatibility
2004-06-15 04:42:59 +00:00
igor%mir2.org
7f56cb9f8f
Fixing regression after previous commit for bug 245882: when used as top scope ImporterTopLevel should not define constructor property not to hide Object.prototype.constructor.
2004-06-09 17:01:01 +00:00
igor%mir2.org
204d41ff43
1. IdFunction.methodId is no longer public as some old Java compilers have troubles with it and accessor method has to be used instead.
...
2. IdFunction constructor now takes scope argument as well to prevent initialization bugs.
2004-06-09 12:54:04 +00:00
igor%mir2.org
45dd15a64f
The last commit made Rhino uncompilable! Fixing that.
2004-06-09 08:25:59 +00:00
igor%mir2.org
407b7697c2
Implementing bug 245882: now ImporterTopLevel can be constructed from JavaScript using JavaImporter to allow to have class and package collections without namespace polluting.
2004-06-09 07:36:31 +00:00
igor%mir2.org
f0b3c4b63f
Work on bug 245882: check that scope is not this when setting scope chain
...
to prevent scope chain loops.
2004-06-08 23:30:30 +00:00
igor%mir2.org
ec1dd303f9
Default implemetation of toSource now lives in ScriptRuntime to allow its
...
reuse in othr objects.
2004-06-08 23:28:32 +00:00
igor%mir2.org
e1d29463ec
Yet another change in IdFunction: now its constructor takes arity argument
...
which removes the need to have separated methodArity in IdFunctionMaster
intertface. The method, however, has to be preseved in IdScriptable since it
is used to find which arity should be used during IdFunction construction.
2004-06-08 22:25:57 +00:00
igor%mir2.org
f066506500
Using IdFunction in place of JIFunction in NativeJavaTopPackage
2004-06-08 15:29:34 +00:00
igor%mir2.org
8f28fc868a
Changin IdFunction/IdFunctionMaster iteraction:
...
1. execMethod and methodArity in IdFunctionMaster do not receive methodId explicitly but rather through the IdFunction instance. To ensure fastest access to method id it is made pubic final field.
2. IdFunction allows to associate an internal tag with itself that allows not to worry about id clashes and simply check for object tag before doing method dispatch.
2004-06-08 14:23:35 +00:00
igor%mir2.org
f5bd7a5a11
JIFunction was wrong idea at the end: it is better to stick with IdFunction and teach it missed functionality. The patch changes JavaAdapter to use IdFunction for its constructor call.
2004-06-08 12:38:20 +00:00
igor%mir2.org
e412382826
API to set global SecurityController is moved to SecurityController not to clatter Context.
2004-06-08 11:09:40 +00:00
igor%mir2.org
b538849b5b
Work on bug 244492: RhinoException is the new common base class for all exceptions.
2004-06-07 16:40:28 +00:00
igor%mir2.org
be6494c229
Switch to using Context.javaToJS instead of Context.toObject
2004-06-07 10:35:22 +00:00
igor%mir2.org
c613550feb
setGlobalSecurityController should be static
2004-06-07 10:34:22 +00:00
igor%mir2.org
daf6b75edb
Introduction of ContextAction interface as a safe alternative for Context.enter/Context.exit. All usages of Context.enter/Context.exit in Rhino are replaced by
...
Context.call(new ContextAction() {
public Object run(Context cx) { ... }
});
or similar alternatives.
It prevents bugs with missed Context.exit and provides room for optimization of Context association with current thread and simplify implementation of safe environment for execution of untrusted applets and scripts.
In addition, Rhino now supports global SecurityController that take precedence over any per/Context controllers so the application can setup the controller once and for all.
2004-06-06 15:33:59 +00:00
igor%mir2.org
70af322f7a
ContextFactory is a new API class
2004-06-06 09:28:18 +00:00
igor%mir2.org
172c10e986
added ContextFactory which is used to create Context instances in JavaAdapter when there is no Context associated with the current thread.
2004-06-04 17:54:42 +00:00
igor%mir2.org
c3ac073e09
Better JavaDoc comments for FEATURE_ constants
2004-06-04 17:51:27 +00:00
igor%mir2.org
4fe37bb798
Allow to disable special treatment of __proto__ and __parent__ via overriding Context.hasFeature().
2004-06-04 14:57:31 +00:00
igor%mir2.org
7048fd0cd3
Use consistent naming for property manipulation functions in ScriptRuntime and drop no longer used scope argument for ScriptRuntime.setProp that acts on strings.
2004-06-04 06:58:03 +00:00
igor%mir2.org
633428122c
Code to create references to manipulate with __proto__ and __parent__ is moved to ScriptRuntime.specialReference not to expose too much implementation details
2004-06-02 14:56:25 +00:00
igor%mir2.org
20af7b885d
Changing Refrence.set(Object) to return Object to return the real stored value.
2004-06-02 11:33:09 +00:00
igor%mir2.org
1a9a123c14
Pass __proto__ and __parent__ special properties to code generation as strings to allow to support application-specific special properties without altering code generation.
2004-06-02 09:13:12 +00:00
igor%mir2.org
ee2f62aef9
Allow Rhino embeddings to support function as left-hand-side extension via overriding BaseFunction.specialCall().
2004-06-01 16:04:18 +00:00
igor%mir2.org
94302e5018
Making invalid left hand of assignment runtime reference error instead of syntax error as dictated by ECMAScript. Now interpreter/class compiler generates code that calls ScriptRuntime.genericReference() that in turn throws the error.
2004-05-31 15:49:19 +00:00
igor%mir2.org
6806c3160e
Work on bug 243057:
...
Introduction of support for calling function in reference context so with native functions code like f(args) = something or x.y(args)++ would be allowed.
The changes added necessary changes to parser and code generation but at runtime currently any such usage would throw an exception.
2004-05-31 14:51:20 +00:00
igor%mir2.org
cb3a4fef51
For prefix and postfix ++/-- use mask flags instead of enumeration to denote postfix/prefix and ++/-- operation.
2004-05-31 13:03:16 +00:00
igor%mir2.org
8f9ce34d52
Optimize ++name and --name similarly to name++ and name-- via special runtime support
2004-05-31 12:09:39 +00:00
igor%mir2.org
ad65f9094a
Optimizing ++x.y and --x.y in the same way as x.y++ and x.y-- via special runtime support
2004-05-31 08:18:38 +00:00
igor%mir2.org
2f00186bd2
Introduction of generic Reference class to support latter constructions like f(args) = y and f(args)++. For now it is used to implement __proto__ and __parent__ special properties so x.__proto__++ now works
2004-05-31 07:18:52 +00:00
igor%mir2.org
feca58e25b
++x[y] is optimized in the same way as x[y]++ via special runtime support
2004-05-30 15:54:21 +00:00
igor%mir2.org
35a14e8364
Using negative nimbers for interpreter-specific tokens to allow for 256 bytecodes without using 0xFF & conversion
2004-05-30 14:21:07 +00:00
igor%mir2.org
7d94a4273b
Added JavaDocs for Kit.initCauseOrIgnore
2004-05-29 09:37:48 +00:00
igor%mir2.org
5a542b76b3
Make code to call Throwable.initCause via reflection previously located in WrappedException publically available in Kit.java
2004-05-29 08:55:41 +00:00
igor%mir2.org
8d03a328c9
Follow conventinal code practics of throwing IllegalArgumentException, IllegalStateException directly instead of using additional helper methods for it.
2004-05-29 08:54:24 +00:00
igor%mir2.org
6dcea14f54
Patch from Merten Schumman <Merten.Schumman@asg.com> to add Main.mainEmbedded(String) which is a version of Main.main without System.in/err/out redirection. It simplifies life for embeddings.
2004-05-28 14:19:11 +00:00
igor%mir2.org
402fd02406
Final implementation of GUI/debug implementation internals separation. Now the newly introduced Dim.java (Dim: Debug IMplementation) depends only on GuiCallback and contains all iteractions with Rhino.
2004-05-27 11:26:59 +00:00
igor%mir2.org
7fe01ad6bc
exitAction is moved to DebugGui
2004-05-27 10:34:44 +00:00
igor%mir2.org
cd63019db8
Prepare to remove dependancy of core debugger implementation on GUI: introduction of GuiCallback interface that DebugGui implements.
2004-05-27 10:21:24 +00:00
igor%mir2.org
b85ecb8c02
Work on 244492: JavaScriptException extends RuntimeException
2004-05-26 11:54:10 +00:00
igor%mir2.org
fa34067a88
Make sure that Context.getSourcePositionFromStack never returns negative numbers and uses 0 to indicate absence of line information. It accounts for differences between JVMs when dealing with incomplete debug information.
2004-05-26 11:03:03 +00:00
igor%mir2.org
63596c3f2b
In JavaScriptException constructor do not call ScriptableObjec.toString(value) since it may potentially trigger script execution for objects defining toString method which in turn may throw exceptions.
...
Instead for Scriptable arguments prints their [Object className] form which is provided by the new ScriptRuntime.defaultObjectToString(Scriptable) method.
2004-05-25 15:39:28 +00:00
igor%mir2.org
7341c1f1b5
Added JavaDoc link to a better way of wrapping exceptions then using WrappedException constructor
2004-05-24 10:11:23 +00:00
igor%mir2.org
8804a5c10d
Restoring public constructor for WrappedException for compatibility with Batik.
2004-05-24 07:53:06 +00:00
igor%mir2.org
2283ed2681
Change Arguments.getIds to include argument indexes when getAll is true. It allows not to have a tailored getIds in the debugger.
2004-05-23 17:34:44 +00:00
igor%mir2.org
3130100ea6
Fixing JavaDoc comments
2004-05-23 17:32:32 +00:00
igor%mir2.org
1e8e9be624
1. Reorganize debugger to move all code that interacts with the engine into Main.java while all GUI logic now presents in DebugGui
...
2. Move downloaded files for swing tree table into downloaded directory and adjust build files accordingly.
2004-05-23 16:15:09 +00:00
igor%mir2.org
cfe01facb2
Make sure that for Rhino-generated exceptions Throwable.getMessage() contains script file and line information.
2004-05-23 14:32:07 +00:00
igor%mir2.org
e431bd1a66
Make EcmaError to follow the same pattern as JavaScriptException
...
and EvaluatorException for getMessage() methods: it provides error message
with script source/line information.
2004-05-23 09:50:20 +00:00
igor%mir2.org
30e9aee832
Switch to line info extraction through single top script processing
2004-05-22 23:19:56 +00:00
igor%mir2.org
677ac1cd76
Splitting Main.java into Main.java and GUI-related DebugGui.java
2004-05-21 15:06:36 +00:00
igor%mir2.org
7ceadc8e69
Split Main class into Main and DebugGui to have better debugger/gui separation
2004-05-21 14:25:21 +00:00
igor%mir2.org
1ad8dc2197
Move all logic to open source window at the end of compilation to single update action.
2004-05-21 12:04:47 +00:00
igor%mir2.org
3265deef44
1. Fixing reporting for calls to undefined: stringReg is not suitable for function names if function contains arguments. So Token.NEW/Token.CALL comes with index of potential function name.
...
2. If debug information is not generated, Context.interpreterLineIndex could contain garbage from previous function invocation leading to ArrayIndexOfBound. Now Context.interpreterLineIndex is initialized at the interpreter entrance and getSourcePositionFromStack checks against absent line numbers.
2004-05-21 11:43:38 +00:00
igor%mir2.org
e58d2a671b
Move logic to show exception dialog into EnterInterrupt action
2004-05-21 11:27:59 +00:00
igor%mir2.org
7706f20db2
Test file
2004-05-21 10:53:19 +00:00
igor%mir2.org
5c690d13af
I changed StackFrame.onLineChange to access breakFlag directly which allowed to remove global context hashtable
2004-05-21 01:20:48 +00:00
igor%mir2.org
261666150e
Merge various GUI initializations during breakpoint hit into single action
2004-05-21 01:10:04 +00:00
igor%mir2.org
0ec2c999de
Removal of manThread: breakFlag can do its job just fine
2004-05-21 00:17:45 +00:00
igor%mir2.org
9f73c42235
Replace Context switching during eval for interrupted threads by thread communication so all evals happens on the original thread.
2004-05-21 00:11:24 +00:00
igor%mir2.org
013271de02
To make variable access faster the form of getvar/setvar with var index
...
encoded in bytecode is reintroduced.
2004-05-18 22:33:41 +00:00
igor%mir2.org
c23cadde22
Added more fast index load instructions and removed 0xFF & when extracting bytecode as all byte codes are less then 128.
2004-05-18 15:02:26 +00:00
igor%mir2.org
f9804a0ac7
In the interpreter loop comparision operations are represented by separated optimized do_cmp.
2004-05-18 12:19:25 +00:00
igor%mir2.org
10f3c4233e
To make loading of strings into string register faster I added Icode_REG_STR_C[01023] that directly loads the string at the given index.
2004-05-18 11:19:30 +00:00
igor%mir2.org
351dcb728c
Removal of 32K restriction on the length of the maximum jump offset.
...
Now interpreter uses UintMap to store jump targets spanning more then 32K
of bytecode.
2004-05-17 21:36:47 +00:00
igor%mir2.org
a6a8297e30
Reorganization in jumping code to make the following patch to drop 32K
...
jump length restriction smaller
2004-05-17 19:19:18 +00:00
igor%mir2.org
7e1c77a956
Remove 256 limit on number of variable using index prefix for all variable
...
operations. To offset small performance loss for variable access new
index prefixes operation REG_IND_C[0123] are added to load 0,1,2,3 as index
value with one bytecode.
2004-05-17 17:57:43 +00:00
igor%mir2.org
bf289c6354
For GETPROP/SETPROP use explicit name prefix to load property name instead
...
of pusshing property to stack
2004-05-17 17:29:12 +00:00
igor%mir2.org
5b81311a84
Remove 256 limit on number of local slots using REG_IND* prefixes to load the
...
slot index
2004-05-17 16:27:39 +00:00
igor%mir2.org
c226bf5cde
Splitting name_and_this bytecode into name_fast_this and name_slow_this to
...
avoid flag reading during simple function calls
2004-05-17 15:58:48 +00:00
igor%mir2.org
0681989924
Use the same of specal index prefix byte code for [] and {} literals
2004-05-17 14:55:45 +00:00
igor%mir2.org
e88881f3a6
Elimination of 64K restriction limit for various index-based opcodes. Now
...
they are prefixed with special Icode_REG_IND* that load the index of proper
size.
2004-05-17 14:34:29 +00:00
igor%mir2.org
696129bcb0
Removal of 64K limit of total number of strings per script/function: now
...
string operands are preloaded with special LOAD_STR opcodes that handles full
int indexes
2004-05-17 13:07:49 +00:00
igor%mir2.org
9a6e1fe15d
In the interpreter loop avoid code duplication for calling instruction
...
observer after each jump: now all tsuch code blocks are merged into generic
code that follows the interpreter switch
2004-05-17 11:44:42 +00:00
igor%mir2.org
34e9dc56b7
Another small modifications to Interpreter to make the following patches
...
smaller: replacing all brek in the main switch of the interpreter with
"continue Loop"
2004-05-16 21:01:00 +00:00
igor%mir2.org
774204ecef
In the interpreter loop increment pc immediately after icode read to avoid
...
ever adding +1 in post byte codes with arguments
2004-05-16 14:02:26 +00:00
igor%mir2.org
daaa920276
Inside the interpreter loop merger many almost identical icode
...
implementations into common block that use additional ifs to distinguish
between icode. It allows to shrink the ever growing Interpreter.interpret()
2004-05-16 13:30:53 +00:00
igor%mir2.org
aae93331f3
New implementation of ARRAYLIT and OBJECTLIT that does not requires stack
...
space to hold all literals values before they are copied to a temporary array
to pass to ScriptRuntime functions. The new code populates array during values
calculations.
2004-05-15 17:28:22 +00:00
igor%mir2.org
b4d0f698c9
Quote property names in object literals produced by toSource() if they
...
coincide with reserved keywords.
2004-05-15 17:23:06 +00:00
igor%mir2.org
37be8f9cfb
Use special parse nodes for array and object literals to generate more compact
...
bytecode both for interpreter and compiler to have faster [] and {} literals.
It also fixes the bug of accessing local variables Array/Object for the
literals instead of top-scope objects.
2004-05-14 22:47:44 +00:00
igor%mir2.org
2494dc182a
Use code similar to optimizer when generating call sequences to avoid tree mutations during code generation
2004-05-12 22:39:40 +00:00
igor%mir2.org
dbaff764ea
1. Fixing bug of forgotten updates to itsWithNesting
...
2. Optimization for calls like name() which similar to the optimizer code
2004-05-12 20:03:23 +00:00
igor%mir2.org
49aabab156
New function ScriptRuntime.notFoundError to throw error on missed properties
...
instead of scattered duplicated code
2004-05-12 19:59:57 +00:00
igor%mir2.org
8423b14b48
Removal of Token.GETBASE, Token.GETTHIS, Token.PARENT, Token.NEWTEMP, Token.USETEMP, Node.TEMP_PROP, Node.FIXUPS_PROP, Node.USES_PROP constants which are no longer used after the last changes
2004-05-12 15:00:41 +00:00
igor%mir2.org
92f45d1054
Instead of transforming parser tree to apply special treatment for call nodes, do that only during code generation .
2004-05-12 13:02:40 +00:00
igor%mir2.org
992ee898f2
Refactoring Codegen.visitCall: preprae for the following switch to generate call without transformer modifications
2004-05-12 10:34:46 +00:00
igor%mir2.org
aaf9cdfddc
Improve error messages for the optimizer when accessing undefined values
2004-05-11 10:06:51 +00:00
igor%mir2.org
b9f058dfe3
Cleanup of hasVarsInRegs
2004-05-10 16:45:02 +00:00
igor%mir2.org
5d1aacca6a
Apply direct call optimization only for functions that do not need activation.
2004-05-10 15:28:32 +00:00
igor%mir2.org
d2ab61c277
Determine functions that require activation during parsing and not in NodeTransformer
2004-05-10 13:39:14 +00:00
igor%mir2.org
27e2f6d328
Removal of checks for unitialized lVal.getJRegister(): the registers are always initialized in prelude.
2004-05-10 05:24:29 +00:00
igor%mir2.org
44deec93c2
Removal of JavaVariable interface in favor of simple method in ClassFileWriter. It allowed to shrink OptLocalVariable and made other code simpler.
2004-05-09 22:09:03 +00:00
igor%mir2.org
ed03829d75
Clenup the usage of getJRegister/assignJRegister
2004-05-09 21:07:08 +00:00
igor%mir2.org
5a48e9dc3a
Move all logic regarding establishing type of variable into Block.java
2004-05-09 21:06:20 +00:00
igor%mir2.org
b7fec72bfc
Removal of no longer used LASTUSE property and corresponding optimizer code
2004-05-09 12:38:45 +00:00
igor%mir2.org
0be4188b87
Remove code for CSE optimisation: it was used only to merger this.property calls but this optimisation changes program semantic if property is controlled by native getter.
2004-05-09 10:14:01 +00:00
igor%mir2.org
316271b0a9
OptLocalVariable cleanup: under optimization GETVAR/SETVAR nodes always has a reference to them
2004-05-08 22:24:02 +00:00
igor%mir2.org
07a54c7b74
Remove markLiveAcrossCall functionality that was never used for any optimizations
2004-05-08 21:05:00 +00:00
igor%mir2.org
1059cf02ce
Merge addToken/addString into single addString
2004-05-08 19:21:37 +00:00
igor%mir2.org
749dfe5b5d
Optimization of ++/-- and merge separated postIncrement/postDecrement to single postIncrDecr to remove too much code duplication
2004-05-07 15:05:11 +00:00
igor%mir2.org
fd2bfff2a7
FatBlock is moved to Block as a private class
2004-05-07 15:03:39 +00:00
igor%mir2.org
09b1d12d63
Added addPush(boolean)
2004-05-07 15:00:17 +00:00
igor%mir2.org
a06c84f791
Extend optimization ==, !=, ===, !== against null to check null presence on any side of the test.
2004-04-21 12:00:27 +00:00
igor%mir2.org
5a691b58fa
Fix max stack calculations in dcpLoadAsObject/dcpLoadAsNunber and optimize visitIfJumpRelOp to take advantage of the fact that left and right numerical operands force numeric context.
2004-04-20 13:00:08 +00:00
igor%mir2.org
c39f4108d7
Reuse code to generate optimized if jumps for boolean operations that return Boolean instance
2004-04-18 20:29:41 +00:00
igor%mir2.org
8231a39a23
Cleanups:
...
1. Added addObjectToDouble() to replace explicit code generation for object to double convertion.
2. nodeIsDirectCallParameter returns dcp register to avoid double look up of OptVariable
2004-04-18 16:29:12 +00:00
igor%mir2.org
2957c844ca
Layout cosmetics
2004-04-18 16:26:32 +00:00
igor%mir2.org
97dc2f98e4
Trust JVM to implement "<=" and "<" according to JVM specs and replace code like
...
d1 == d1 && d2 == d2 && d1 < d2
with simple
d1 < d2
That in turn allows to simplify code generation and remove OptRuntime.cmp_ functions.
2004-04-18 14:53:34 +00:00
igor%mir2.org
b9219ca745
Trust JVM to implement "<=" and "<" according to JVM specs and replace code like
...
d1 == d1 && d2 == d2 && d1 < d2
with simple
d1 < d2
2004-04-18 14:52:22 +00:00
igor%mir2.org
d6576c5892
Remove commented out code to treat integer specially: it is not applicable any longer in any case.
2004-04-18 14:49:50 +00:00
igor%mir2.org
d62ad77af5
Work on bug 238649: no more support for ClassNameHelper and friends
2004-04-18 12:15:07 +00:00
igor%mir2.org
b71427a6ae
Worj on bug 238649: removal of NotAFunctionException
2004-04-18 09:30:30 +00:00
igor%mir2.org
534c7d51d6
1. Change cmp_* functions to return boolean, not int, to simplify code.
...
2. Move optimizer-only functions from ScriptRuntime.java to optimizer/OptRuntime.java
3. Add ScriptRuntime.wrapBoolean to wrap boolean value and use it in the optimizer
2004-04-16 07:38:20 +00:00
igor%mir2.org
81470d5283
1. Throw IllegalArgumentException instead of EvaluatorException when argument does not belong to JS runtime types.
...
2. Optimize eq and shallowEq in ScriptRuntime to share code with optimized versions in Interpreter.
2004-04-14 14:50:24 +00:00
igor%mir2.org
cb3ccb29b3
Layout cosmetics
2004-04-14 14:46:58 +00:00
igor%mir2.org
153f7be2a2
Support for Date.now()
2004-04-14 11:04:55 +00:00
igor%mir2.org
c685edc692
Fixing bug 58118 : long overdue commit of patch from Mike McCabe, mike+mozilla@meer.net
2004-03-30 12:48:29 +00:00
igor%mir2.org
701ea21521
Since new x in Java never returns null, the check for null in date_format was redundant.
2004-03-30 10:35:42 +00:00
igor%mir2.org
eef0796b8a
Faster implementation of MonthFromTime and DateFromTime:
...
1. Use day / 30 as month estimate with the following adjustment via switch()
2. Reuse year from day calculations in IsLeapYear
2004-03-29 15:19:04 +00:00
igor%mir2.org
021b376688
Fixing bug 239068: proper initialization of parent scope for functions of constructors.
2004-03-29 15:15:34 +00:00
igor%mir2.org
3a93fa2e8e
DaysInYear is replaced by IsLeapYear since the former was used only to determine if a particular was leap or not
2004-03-29 14:16:23 +00:00
igor%mir2.org
ba8f69355a
Cosmetics: layout fixes
2004-03-29 12:30:24 +00:00
igor%mir2.org
dfdd5ab46c
Starting 1.5R5.1 development
2004-03-29 12:29:11 +00:00
igor%mir2.org
40db418d07
Fixing 238823 : throw explicit IllegalArgumentException when JS source for Context.compileFunction does not contain single JS function statement instead of producing silent empty functions or throwing obscure NullPointerExceptions
2004-03-27 09:35:22 +00:00
igor%mir2.org
e229fd6ab4
Better formating
2004-03-26 13:01:17 +00:00
igor%mir2.org
78ec796300
Fixing bug 238699 : refactoring to expose simpler code generation API caused to select wrong code path when compiling functions
2004-03-26 12:42:00 +00:00
igor%mir2.org
0292a753a1
Explicit date in Context.getImplementationVersion() for 1.5R5 release
2004-03-25 20:37:32 +00:00
igor%mir2.org
1aafe09612
Reflecting 1.5R5 release
2004-03-25 19:45:56 +00:00
igor%mir2.org
cbce2b4745
Starting 1.5R6
2004-03-25 17:58:03 +00:00
igor%mir2.org
420074b3a1
Removal of macbuild support: its Mac OS X time after all
2004-03-25 17:39:21 +00:00
igor%mir2.org
afe5a6cd9d
Info about commercial support
2004-03-25 15:46:40 +00:00
igor%mir2.org
abf0a3a1ac
*** empty log message ***
2004-03-25 11:04:04 +00:00
igor%mir2.org
37c396db64
Preparations for 1.5R5
2004-03-24 15:52:55 +00:00
igor%mir2.org
533e9f8489
Updates to reflect new extension to allow to pass function to Java method expecting interface: now interface with multiple methods are allowed as long as all methods has the same signature
2004-03-24 15:44:19 +00:00
igor%mir2.org
ae3df9d02a
Finalizing 223435 : function can be converted to Java interface with more then one method as long as all methods has the same signature.
2004-03-24 14:15:37 +00:00
igor%mir2.org
76316f18dd
More work on bug 223435 : pass method name for the interface as the last parameter to JS function so it can use for debug purposes
2004-03-23 16:24:32 +00:00
igor%mir2.org
f4ff06f911
Making BaseFunction.decompile package private: any new public function will always find its users even if the usage is broken. It was public to allow calls from debugger but simple call to Context.toString() would do the same job.
2004-03-18 13:37:29 +00:00
igor%mir2.org
0cbfdabe29
RUnit Software now includes a link on their page about JS (Declaimer: I work there!)
2004-03-18 13:32:52 +00:00
igor%mir2.org
f040fa2436
Added RUnit Software
2004-03-17 18:36:57 +00:00
igor%mir2.org
46abef99b9
Fixing bug 237771 : allow to transfer toSource implementation to different objects
2004-03-17 18:36:39 +00:00
igor%mir2.org
6b70acca26
Documenting new Context seal API and fixing obvious English mistakes.
2004-03-09 23:34:30 +00:00
igor%mir2.org
ecfd54c02f
Updated file with license info from Eugene Aresteanu :
...
> Hi,
>
> Does it look right now?
>
> Regards,
> Eugene
>
> -----Original Message-----
> From: Igor Bukanov [mailto:igor@fastmail.fm]
> Sent: Friday, March 05, 2004 10:07 AM
> To: eugene aresteanu
> Subject: Re: Rhino and license for Messages_fr.properties
>
>
> eugene aresteanu wrote:
>
>>Hello,
>>
>>Here it is. I hope it is right.
>>Why do you say the default Messages.properties lacks notices?
>>I used the header of Messages.properties as a template.
>
>
> Sorry for not been clear, but what I wanted to say is that did not have
> triple MPL/GPL/LGPL but rather the old NPL/GPL pair. Could you update
> the file one more time using
> http://www.mozilla.org/MPL/boilerplate-1.1/mpl-tri-license-sh as template?
>
> Regards, Igor
2004-03-05 19:17:43 +00:00
igor%mir2.org
5586a28481
Resolving bug 236117 : new API to seal Context
2004-03-03 19:59:25 +00:00
igor%mir2.org
58db1d9fd5
Resolving bug 236193: require enetered Context for compilation
2004-03-03 11:20:33 +00:00
igor%mir2.org
e0a2c2536d
Making javaToJS static to match the rest of API
2004-03-01 19:19:47 +00:00
igor%mir2.org
40f5e70cee
Fixing JavaDoc
2004-02-24 07:20:49 +00:00
igor%mir2.org
a3d001e0e8
Making NativeScript package private: if this will be controversial, then the public status can be restored
2004-02-24 07:19:56 +00:00
nboyd%atg.com
98a70b8580
Remove "pre" from a candidate build.
2004-02-18 17:28:40 +00:00
igor%mir2.org
4b836891b2
Preparations for 1.5R5
2004-02-18 12:24:27 +00:00
igor%mir2.org
8f085f363e
Preparations for 15R5 release.
2004-02-13 18:50:51 +00:00
igor%mir2.org
adeccbc67c
Fixing NativeGlobal.encode: it was broken for chars beyond 0xFFFF as it sb.setLength(k) was called after k was increased to consume the second char from UTF-16 encoding to build UCS-4.
2004-02-12 19:05:47 +00:00
igor%mir2.org
1e3f7aaaed
Preparations for 1.5R5
2004-02-12 18:13:00 +00:00
igor%mir2.org
ec7de43908
Fixing bug 233274: for/in loop goes through array elements in wrong order
...
Fix: populate ids array in NativeArray with dense indexes first
Since array literals in Rhino creates instances of NativeArray with the
internal dense array containing literal elements, the patch changes
NativeArray.getIds to return ids array with dense indexes coming first and
indexes for elements added later after that.
2004-02-06 20:16:36 +00:00
igor%mir2.org
3ba7b938c3
Adding backward compatibility to SecurityController so it can be used with Rhino 1.5R4.1 subclasses
2004-02-05 15:10:52 +00:00
igor%mir2.org
d3b31dac4f
Fixing bug 210605 again: for now re-throw Error instances as is without wrapping them into WrappedException so scripts would not be able to catch them.
2004-01-29 17:04:31 +00:00
igor%mir2.org
a7573a8a91
Removal support for deprecated getTargetClassFileName/setTargetClassFileName: they throw exception instead of calling a special proxy to avoid having probably unused any more class. If this is not true, then the methods should be restored.
2004-01-18 21:26:34 +00:00
igor%mir2.org
7d76224ac8
Fixing JavaDoc comments
2004-01-18 21:23:53 +00:00
igor%mir2.org
b224c3704c
EvaluatorException and WrappedException are public API now
2004-01-18 21:21:41 +00:00
igor%mir2.org
176a213396
Deprecating ClassNameHelper and ClassRepository in favor of new optimizer/ClassCompiler.
2004-01-17 21:52:30 +00:00
igor%mir2.org
b629e258d7
optimizer.ClassCompiler provides new simple API for generating of class files from scripts that does not require to create Context instances.
...
The jsc tool is updated to use the new interface instead of using ClassNamehelper and friends.
2004-01-17 21:02:32 +00:00
igor%mir2.org
134428fef2
1. optimizer/Codegen now just generates byte code classes, the instantiation of classes is done in a separated class optimizer/ClassCompiler. This is done to provide in future a simple interface to generate class files instead of complicated ClassNameHelper and friends implementation.
...
2. Parser now stores TokenStream instance in its field instead of having it present in each and every method signature. It allowed to simplify interaction between Parser and IRFactory where the later now uses Parser.reportError so parser van recover from syntax errors and provide better error reporting.
2004-01-16 18:20:22 +00:00
igor%mir2.org
0fc143dd7c
Added setters for properties of CompilerEnvirons.java so it can be initialized
...
without having Context object.
2004-01-16 16:09:07 +00:00
igor%mir2.org
34ebe5f4c1
Renaming Context.hrowAsUncheckedException to Context.throwAsScriptRuntimeEx to better reflect its purpose, providing better documentation and making sure it throws only instances of EcmaError and EvaluatorException.
2004-01-16 15:14:25 +00:00
igor%mir2.org
497a0f766c
Fixing bug 228336 :
...
This is a Rhino version of Brandan's patch from attachment 138844, bug 228087
plus code to fix overflow problem reported in bug 230216.
2004-01-16 15:11:31 +00:00
igor%mir2.org
a7050e5460
throwAsUncheckedException is moved from ScriptRuntime to Context as a useful API to use in external applications.
2004-01-15 20:41:57 +00:00
igor%mir2.org
f26418ea3f
Fixing typo in comments
2004-01-15 20:40:56 +00:00
igor%mir2.org
96ea5ea515
Patch from Alex Bitney to fix bug 230883
2004-01-15 20:10:54 +00:00
igor%mir2.org
59ae0034a1
Fixing bug 229571 : code refactoring for speed optimization
2004-01-06 13:35:51 +00:00
igor%mir2.org
332cc13ae5
Move details of parsing/code generation from Context to interpreter/optimizer to allow more pluggable compilation interface.
2004-01-05 13:22:40 +00:00
igor%mir2.org
75d4dec8ea
Cosmetics: removal of tabs left by the previous commit
2003-12-27 21:36:00 +00:00
igor%mir2.org
3cab5b3e26
Decompling parsing from code generation: compiler instances are created only when parsing tree is build.
2003-12-27 21:32:35 +00:00
igor%mir2.org
44e703307d
Initial patch to separate parsing from code generation: changing IRFactory so it can be created in Context.
2003-12-27 09:53:57 +00:00
igor%mir2.org
d4590c48fa
Renaming Kit.badArg() -> Kit.argBug() for consistency with Kit.codeBug()
2003-12-27 09:38:57 +00:00
igor%mir2.org
65aea23579
Moving parsing code from FunctionNode.finishParsing to IRFactory to have better overview of parsing code. The optimizer now calls explicitly OptFunctionNode initialization instead of overriding FunctionNode.finishParsing in OptFunctionNode.
2003-12-26 17:17:21 +00:00
igor%mir2.org
10e91b97ff
Functions to report synatx errors are moved from TokenStream to Interpreter to remove dependancy on TokenStream in NodeTransformer.
2003-12-26 14:01:21 +00:00
igor%mir2.org
32ebb1d38d
Dropping scope argument from Context.compileString/compileReader since it is no longer used as Script instances are scope independent.
...
For compatibility the old form of Context.compileReader is preserved but marked as deprecated.
2003-12-25 22:37:57 +00:00
nboyd%atg.com
8390b6a651
Contribution from earestea@scanpak.com.
2003-12-18 20:55:30 +00:00
nboyd%atg.com
3384c4c1f5
Add Aviva info.
2003-12-18 16:53:32 +00:00
igor%mir2.org
d7a98d84dc
Message about overcapacity
2003-12-09 19:23:25 +00:00
igor%mir2.org
aeafac2fcf
Cosmetics: removal of tabs
2003-12-09 19:22:58 +00:00
igor%mir2.org
6625e3229d
To avoid bad performance when joining array with few small strings, js_join precalculate the resulting string size.
2003-12-09 19:22:11 +00:00
igor%mir2.org
f53203c746
Replacing custom buffer management in replace code with the standard StringBuffer
2003-12-08 17:10:23 +00:00
igor%mir2.org
528efdfe62
Inlining LabelTable into Interpreter and ClassFileWriter: to support jumps spanning over 64K in Interpreter in future LabelTable should be modified significantly and complexity of shared code would not justify it. In addition such inlining actually decrese total code size.
2003-12-08 14:16:53 +00:00
igor%mir2.org
44fae2ad68
Merger ReplaceData and MatchData helper classes into GlobData to reduce code size and rename helper methods to follow SpiderMonkey closely.
2003-12-08 13:24:49 +00:00
igor%mir2.org
aed233007b
Fixing missed cleanup of old values stored in RegExpImpl.parens: now it has type SubString[] and reinitialized after running regexps and see bug 220367.
2003-12-07 21:56:29 +00:00
igor%mir2.org
dcd6522b88
Patch to fix bug 223451 :
...
Check for unbalanced ) and fix for ArrayIndexOutOfBoundsException
Check for unbalanced ')' is done differently then in SM since rhino uses
recursive version of SM code while fix for ArrayIndexOutOfBoundsException is
just a variation of my patch for bug 227705 in SM.
2003-12-07 16:00:38 +00:00
igor%mir2.org
21f15557ab
Fixing bug 225926 : patch from Roger Lawrence
2003-12-07 15:35:40 +00:00
igor%mir2.org
c97a96edca
Fixing bug 225366 : port of SM patch
2003-12-05 23:22:10 +00:00
igor%mir2.org
449bf5eadc
In Object.toSource implementation if property is not valid Java identifier, print it as 'escaped_property' so the property will read as proper Java string.
2003-12-04 15:47:03 +00:00
igor%mir2.org
b1ba2b408e
Fixing bug 225831 : use before_225831_merge to get code before the far reaching fix
2003-12-04 14:41:22 +00:00
igor%mir2.org
2997e9f7ee
Fixing http://bugzilla.mozilla.org/show_bug.cgi?id=226045 :
...
(new RegExp).toString/toSource() now gives /(?:)/ instead of // which is comments and does not work with uneval.
2003-11-24 15:51:44 +00:00
igor%mir2.org
46e97e6ab7
Fixing debug printout of icode: CALLSPECIAL prinout was not updated to reflect proper value of only 6 additional bytes for bytecode.
2003-11-17 18:31:55 +00:00
igor%mir2.org
a9bb9a44ac
To mark special properties __proto__ and __parent__ in parse tree use integer ids instead of strings so it is easier to check tree consitency.
2003-11-17 17:39:43 +00:00
igor%mir2.org
cbacb923d3
Finalizing toSource() support: do not put () arround {} for source representation of objects that are part of recursive toSource call.
...
For details, see http://bugzilla.mozilla.org/show_bug.cgi?id=225465
2003-11-17 17:37:21 +00:00
igor%mir2.org
02c0cb7a89
Changing wrapping of caught Java exceptions: now they are constructed with JavaException constructor, have fileName, lineNumber defined to point to script line that executed Java code triggering exception and have javaException property pointing to the wrapped exception object.
...
For details, see http://bugzilla.mozilla.org/show_bug.cgi?id=225817
2003-11-17 11:04:29 +00:00
igor%mir2.org
834efdbe71
Fixing http://bugzilla.mozilla.org/show_bug.cgi?id=225817 :
...
Proper support for fileName and lineNumber in Error objects.
2003-11-17 09:45:01 +00:00
igor%mir2.org
7c4e5d8424
Allow for scripts to catch instances of EvaluatorException
2003-11-14 20:38:06 +00:00
igor%mir2.org
8e546b9047
Transform StackOverflow during parsing into runtime exception so script can detect too deep recursion during parsing.
2003-11-14 20:33:43 +00:00
igor%mir2.org
7c0cff6574
Splitting Token.ASSIGN into Token.ASSIGN and Token.ASSIGNOP where the former represents only = while ASSIGNOP is used to denote +=, -= etc. Parse tree for the 2 operations is rather different and split allowed to simplify IRFactory code.
2003-11-14 16:03:31 +00:00
igor%mir2.org
2e75448c13
Implementing uneval/toSource functionality of SpiderMonkey. Foe details, see http://bugzilla.mozilla.org/show_bug.cgi?id=225465 .
2003-11-13 17:15:31 +00:00
igor%mir2.org
7d1e5eda47
Passing to decompiler information about function type so it can use it to customize its output in future.
2003-11-12 15:47:50 +00:00
igor%mir2.org
f2ba87fed2
Changing Decompiler.decompile to take just encoded source, flag argument controlling output format and property table so it would be possible to add more options to decompiler without changing the rest of code.
2003-11-12 12:22:12 +00:00
igor%mir2.org
e734cecce2
Changing return type of NativeFunction.getEncodedSource() to String from Object since the idea of having anything else but String would require changes in API in any case.
2003-11-12 10:23:53 +00:00
igor%mir2.org
b7d4dfaa28
In ClassFileWriter.addPush(double) use DNEG to generate more compact code for -0.0 and -1.0
2003-11-11 20:31:36 +00:00
igor%mir2.org
f260baaf11
Changing just introduced CodeBlock into Callable that define call method with exactly the same signature as Function.call and changing Function to extend Callable.
...
Now Context defines new method "call" that provide optimized version of the following code:
Context cx = Context.enter()
try {
callable.call(cx, scope, thisObj, args);
} finally {
Context.exit();
}
Since Function extends Callable, it can be passed to this method directly thus simplifying ScriptableObject.call method and Java adapter code.
The new interface is used in SecurityController implementation which allows to pass interpreted functions there directly as well removing the need to have intermediate class in Interpreter.
2003-11-11 20:30:08 +00:00
igor%mir2.org
dd1efd374f
Extend constant folding to cover few unary operations
2003-11-11 20:18:34 +00:00
igor%mir2.org
12c1d23434
1. Make sure that -0.0 from parser are treated properly: new constant folding code there may produce it.
...
2. Instead of generating code to call Double constructor use the new method OptRuntime.wrapDouble() that allows to cut amount of produced byte code.
2003-11-11 20:04:38 +00:00
igor%mir2.org
8313023fae
Patch from Dave Russo to wrap arguments array properly: the code has to allocate new Object[] array since Context.newObject needs instances of Object[], not SomeObjectSubclass[], to work properly.
2003-11-10 02:18:11 +00:00
igor%mir2.org
44c6535ad4
Check in Context.newArray(Scriptable scope, Object[] elements) that elements is instance of Object[] and not SomeObjectSubclass[] to expose broken usage of newArray as soon as possible.
2003-11-10 02:14:37 +00:00
igor%mir2.org
dc34ed3dc1
Changes in Security API: using CodeBlock interface instead of reusing Script to represent code that should be executed with different domain.
2003-11-07 22:29:59 +00:00
igor%mir2.org
49ffeac3df
New interface to refer to code executed with a particular context and arguments.
2003-11-07 22:28:22 +00:00
igor%mir2.org
6006898229
1. Removal of very recently introduced hideFromContextListeners: it was a wrong solution to the problem of running scripts under independent Context instances on the same thread.
...
2. Optimization of Context.enter()/Context.exit() when ThreadLocal is available:
Store Context[1], not Context in the variable so Context.enter()/Context.exit() would need to access ThreadVar only once and then update the reference.
2003-11-07 22:24:53 +00:00
igor%mir2.org
5b675f523e
Added setDouble, setString that are used in IRFactory during constant folding
2003-11-07 22:18:49 +00:00
igor%mir2.org
8eb8104ae7
Move all contant folding logic to IRFactory where it simpler to do and possible to reuse tree nodes during folding.
2003-11-07 22:17:54 +00:00
igor%mir2.org
7ed29ee5a5
Do constant folding during parsing since it is cheap to do that at this stage then in optimizer during separated phase so even the interpreter will take advantage of the folding.
2003-11-07 15:58:57 +00:00