Comment 1 Dani Megert 2004-07-21 13:24:12 EDT

-Which build?
-Did you install additional plug-ins?
-Is this repeatable?
-Are there other errors in .log?
-You say a stack overflow happened the second time - what happened the first
time? -Dialog? Anything in .log?
-What kind of file did you try to open? Can you attach the file?

Comment 2 David Williams 2004-07-21 14:07:20 EDT

I would also ask which VM and if there's any special VM arguments specified 
that might effect stack size? I know I routinely open 300K java files (as part 
of a test suite) and I've never seen this. So, unlikely to be just file size. 
I'm sure there's something there, and would be nice to know what ... could it 
be related to number of warnings (or annotations) in file?

Comment 3 Mark Rzepka 2004-07-22 02:14:43 EDT

Created attachment 13508 [details]
The file that caused the the stack overflow

Comment 4 Mark Rzepka 2004-07-22 02:31:26 EDT

- I try to open the file: Dialog: Cannot open default editor on FrTest.java.
Editor could not be initialized.
- I try second time: Dialog: A stack overflow error has occured. ... Do you want
to exit the workbench?
- Here are the corresponding .log entries:
!SESSION Jul 22, 2004 07:46:25.78 ----------------------------------------------
eclipse.buildId=I200406251208
java.version=1.4.2_04
java.vendor=Sun Microsystems Inc.
BootLoader constants: OS=win32, ARCH=x86, WS=win32, NL=de_DE

!ENTRY org.eclipse.core.runtime 4 2 Jul 22, 2004 08:01:46.671
!MESSAGE An internal error occurred during: "Launching".
!STACK 0
java.lang.StackOverflowError

!ENTRY org.eclipse.ui 4 4 Jul 22, 2004 08:08:10.359
!MESSAGE Unhandled event loop exception

!ENTRY org.eclipse.ui 4 0 Jul 22, 2004 08:08:10.359
!MESSAGE java.lang.StackOverflowError
!STACK 0
java.lang.StackOverflowError

- plugins:
exmples, CDT, EMF-SDO-XSD

- the bug is repeatable with the attached FrTest.java file

- I start eclipse with:
C:\eclipse\eclipse.exe -vmargs    -Djava.endorsed.dirs=C:\java\xalan-j_2_6_0\bin\

- There are no warnings in file

Comment 5 Mark Rzepka 2004-07-22 02:32:53 EDT

Created attachment 13509 [details]
The superclass of FrTest.java

Comment 6 Mark Rzepka 2004-07-22 02:33:45 EDT

Created attachment 13510 [details]
A needed class for FrTest.java

Comment 7 Dani Megert 2004-07-22 02:39:54 EDT

The log entries you attached seem to be only parts. Please attach the whole .log
file (zip if too large). Also, the following entry rather looks like you
launched and not opened the editor:

!ENTRY org.eclipse.core.runtime 4 2 Jul 22, 2004 08:01:46.671
!MESSAGE An internal error occurred during: "Launching".
!STACK 0
java.lang.StackOverflowError

How do you open the editor exactly? Is the file part of a Java project?

Comment 8 Mark Rzepka 2004-07-22 02:54:03 EDT

I dubble click on the file in the package explorer.
The file is part of Java project in the default package.

Comment 9 Mark Rzepka 2004-07-22 02:57:33 EDT

Created attachment 13511 [details]
The whole log

Comment 10 Dani Megert 2004-07-22 03:22:18 EDT

Sorry, I don't have rar installed. Can you use simple zip or attach it
uncompressed. Thanks.

Another thing: can you try
1) open the file using Open With > "Java Editor"
2) try with a fresh Eclipse SDK?

Comment 11 Mark Rzepka 2004-07-22 03:38:35 EDT

Created attachment 13512 [details]
The whole log zipped

Comment 12 Mark Rzepka 2004-07-22 03:45:17 EDT

open the file using Open With > "Java Editor"

>>> the same behaviour

Comment 13 Dani Megert 2004-07-22 03:48:44 EDT

Too me the log looks like all the stackoverflows come from Debug when you try to
launch.
>at
org.eclipse.debug.ui.actions.OpenLaunchDialogAction.run(OpenLaunchDialogAction.java:80)

I also found errros caused by XMLBuddy but in comment 4 you only wrote:
- plugins: exmples, CDT, EMF-SDO-XSD

How is that possible?

Does it really fail when you simply open that file? Can you clear the .log and
simply open that file? Also, please try with a fresh install.

Comment 14 Mark Rzepka 2004-07-22 03:55:14 EDT

Created attachment 13513 [details]
Whole log of a fresh install

Comment 15 Mark Rzepka 2004-07-22 03:59:55 EDT

I have removed XMLbuddy and the behaviour was the same.

I have tried with fresh build (see new log)
The bahaviour ist the same.

Comment 16 Dani Megert 2004-07-22 04:02:28 EDT

OK. I think I got it. You launch with default VM args i.e. don't give Eclipse
enough memory to hanlde this large file. Try 
eclipse.exe -vmargs -Xmx200M -Djava.endorsed.dirs=C:\java\xalan-j_2_6_0\bin\

Comment 17 Mark Rzepka 2004-07-22 04:15:43 EDT

I start with -Xmx200M
1. Eclipse don't want to start because FrTest.java was editing in last session.
So I must delete the file.
2. Open with -> Java Editor is OK!
3. Open with dubble click on the file -> old error

Comment 18 Dani Megert 2004-07-22 04:20:56 EDT

ok - a bit better now ;-)
Go to the preferences and check what's the default editor for *.java files.

Comment 19 Mark Rzepka 2004-07-22 04:26:08 EDT

There is no default editor associated with *.java files, but on dubble click in
the windows explorer the usual notepad is opened

Comment 20 Mark Rzepka 2004-07-22 04:44:53 EDT

reference to comment #7:
I cannot also launch the application with FrTest.java

Comment 21 Dani Megert 2004-07-22 06:50:00 EDT

In comment 18 I meant which one is associated in Eclipse:
Window > Preferences > Workbench > File Associtations

Comment 22 Mark Rzepka 2004-07-22 07:00:41 EDT

the default editor is Java Editor

Comment 23 Dani Megert 2004-07-22 08:04:34 EDT

Can reproduce the problem with the attached files.

Comment 24 Dani Megert 2004-07-22 08:25:57 EDT

Can be reproduced using SDK 1.4.2_03 (build 1.4.2_03-b02) but not using IBM J2SC
(with or without -XJ9 flag). Works if running in interpreted mode (-Xint).
==> JIT bug in VM.

Here's where the stack overflow happens:

BinaryExpression.resolveType(BlockScope scope) {

		boolean leftIsCast, rightIsCast;
		if ((leftIsCast = left instanceof CastExpression) == true) left.bits |=
IgnoreNeedForCastCheckMASK; // will check later on
		TypeBinding leftType = left.resolveType(scope);
                                            ^^^^^^^^^^^

Comment 25 Dani Megert 2004-07-22 08:42:10 EDT

Correction to last comment: it also does not work when using:
  build 1.4.2, J2RE 1.4.2 IBM Windows 32 build cndev-20040524 (JIT enabled: jitc)
without -Xj9. Works if JIT is disabled or -Xj9 is used.

Comment 26 Philipe Mulet 2005-04-07 07:26:55 EDT

David - can you still reproduce ?

Comment 27 David Audel 2005-04-07 09:33:19 EDT

I can reproduce the bug in I20050405-0800

Comment 28 David Audel 2005-04-07 10:35:55 EDT

The problem come from the resolution of an AST that represent a concatenation of
more than 2800 String.

Comment 29 Olivier Thomann 2005-04-07 11:27:37 EDT

The problem is that the string concatenation has not been optimized in the
compiler's tree in order to produce all the string literals for the DOM/AST. The
resulting issue is that the compiler cannot resolve the tree because of the
StackOverFlowError.

Comment 30 Olivier Thomann 2005-04-07 11:28:52 EDT

The only way I can see to fix it is to optimize the string literals during the
DOM/AST parsing and reparse the whole string literal to get the proper positions.

Comment 31 Olivier Thomann 2005-04-07 11:59:08 EDT

In fact on this specific source, even if I use the optimizeStringLiteral support
from the compiler it won't work.
The problem comes from the fact that the huge string concatenation starts with a
method invocation and then it is not optimized.
So the problem is really in the compiler that cannot compile this file.
javac 1.5 also fails with a StackOverFlowError.

Comment 32 Olivier Thomann 2005-04-08 11:04:11 EDT

Philippe,

This might require the addition of a new node in the compiler.
Jikes can successfully compile it. It is the only java compiler that
successfully compiles this code.

Comment 33 Olivier Thomann 2005-04-08 12:22:14 EDT

If the code is simply using a huge string concatenation, the conversion works fine.
Add regression test in
org.eclipse.jdt.core.tests.dom.ASTConverterTestAST3_2.test0602.

Comment 34 Philipe Mulet 2005-05-09 12:55:12 EDT

not for 3.1

Comment 35 Dani Megert 2006-06-16 10:35:58 EDT

What about 3.3 ;-)

Comment 36 Philipe Mulet 2007-04-02 07:44:23 EDT

reopening

Comment 37 Philipe Mulet 2007-04-02 07:44:58 EDT

Olivier - is this still an issue ?

Comment 38 Jerome Lanneluc 2007-06-04 05:03:12 EDT

If this is still an issue, please investigate for 3.4. Otherwise please close.

Comment 39 Olivier Thomann 2007-06-21 11:18:01 EDT

Could not reproduce the failure using 3.3RC4.
Closing as WORKSFORME.
Please reopen if you get it again.