Datasets:

AndressaStefany

/

bug-reports

like 0

{'$oid': '52e9b48854dc1c25ebdb2075'}

DW (9/27/00 11:12:31 AM) Currently, the "type" of an object is shown in the variables view for its value. Only Strings and primitive types show concrete values. We should display the "toString()" of each object as its value. DW (10/23/00 11:22:02 AM) Comments from CM: First I need to define "thread" <g> because there are 2 of them: 1) when I say "thread", I mean a Java Thread, i.e. an object of type java.lang.Thread 2) when I say "vm thread", I mean the debugger's represenation of a thread in the VM, i.e. an object of type com.oti.defrogger.model.DbgThread So, when asked for an object's toString, the debugger creates a new thread to do the following: - lock the vm thread so that no other thread can invoke toString on that vm thread - set a flag on the vm thread: fIsInvokingToString = true - use INVOKE_SINGLE_THREADED to send toString() to the object in the vm thread - set fIsInvokingToString = false - unlock the vm thread The fIsInvokingToString flag is checked on a breakpoint, watchpoint, or exception, so that we don't stop on "debugger-invoked" toStrings. The debugger will stop if the USER runs through a toString method in the ordinary manner (if a breakpoint, watchpoint, or exception is encountered). DS (10/23/2000 4:14:16 PM) 3 days EG (11/8/00 4:54:47 PM) I've worked with toString enabled for some time in VAME. It introduces both performance and run-time behaviour problems. I had to switch it off eventually. DW (12/7/00 5:14:30 PM) Created simple implementation that is not robust. Needs to account for breakpoints, exceptions , and timeouts. DW (12/7/00 7:18:38 PM) Added breakpoint support, but did not release implementation into plugin file. Issue: If an infinite loop is hit, how can we stop it? It is a client thread - we cannot kill it. A "evaluation server thread" would be cool, but I don't see how we can create one without adding code to the target - which cannot always do. DW (12/8/00 9:26:18 AM) Found that LeapFrog debugger does not handle infinite loops. For example: public class Junky { public static void main(String[] args) { Junky j = new Junky(); j.doit(); } public void doit() { System.out.println(this.toString()); } public String toString() { int i = 0; boolean yes = true; while (yes) { i++; } return "never"; } } DW (12/8/00 9:40:38 AM) Locking not handled either: public class Lock { public synchronized void lock() { System.out.println("While in this method, 'this' is locked"); int i = 0; boolean yes = true; while (yes) { i++; } } public synchronized String toString() { return "un-locked"; } } public class Locking { public static void main(String[] args) { (new Locking()).doit(); } public void doit() { final Lock lock = new Lock(); Runnable r1 = new Runnable() { public void run() { lock.lock(); } }; Thread t = new Thread(r1); t.start(); lock.hashCode(); System.out.println(lock.toString()); } } DW (12/8/00 12:13:39 PM) Moved to inactive, as there is no safe way to do evaluation. DW (10/2/01 1:16:46 PM) Use pop-up text widget to display toString()

enhancement

RESOLVED

DUPLICATE

{'$oid': '52e9b48854dc1c25ebdb2076'}

DS (11/14/00 2:59:11 PM) You can set a breakpoint that is in an invalid location for example: public class Inner implements Enumeration { private int count; private Object[] contents; The breakpoint creator allows breakpoints to be positioned on the two variable declaration lines. These breakpoints are not hit. DW (1/2/01 2:07:02 PM) This seems like it should be a problem for the Java parser or complier, and hidden in the Java breakpoint creator. There is a BreakpointLocationVerifier for jdt.debug. It could be improved.

enhancement

RESOLVED

DUPLICATE

{'$oid': '52e9b48854dc1c25ebdb2077'}

DW (10/23/00 11:22:46 AM) Support to show which threads are waiting on which monitors. DS (10/23/2000 4:30:48 PM) 3 days EG (1/4/01 6:00:41 PM) this could be a simple list with all the active monitors. more advanced would be to integrate deadlock analysis. DS (7/9/01 11:12:19 AM) JDI: ThreadReference#currentContendedMonitor() Returns an ObjectReference for the monitor, if any, for which this thread is currently waiting JDI: ThreadReference#ownedMonitors() Returns a List containing an ObjectReference for each monitor owned by the thread.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48854dc1c25ebdb2078'}

DW (10/23/00 11:08:55 AM) We currently show the instruction pointer by highlighting code. We should also use an instruction pointer marker as in VAME (red arrow). DS (10/23/2000 4:06:12 PM) 0.5 day DW (11/29/00 10:17:47 AM) Issue: the types of markers is defined by the platform. We would need a new type of marker. DW (12/6/00 2:09:36 PM) Deferred until we get new marker support. JGS (1/17/01 10:51:04 AM) With the new desktop, it seems there are still only 6 predefined marker types. However, could we use type OTHER and the "getAdapter()" method from the IAdaptable interface to allow an InstructionPointerMarker to provide an icon (and anything else required)? DW (1/23/01 2:59:59 PM) Waiting on new marker support. DW (3/8/01 4:55:11 PM) Issues: How will editors know how to render the instruction pointer? What will the marker be associated with? I.e. how will it work with non-resources? DW (3/23/01 9:47:32 AM) On Hold - Waiting for Workbench support for marker rendering/registering. DW (3/30/01 9:18:40 AM) Support is coming - URL attribute on the marker to refer to an image. DW (4/20/01 8:51:01 AM) For now use existing instuction pointer that we use to highlight source lines. Give it an image, and the Java editor will add it as a known marker type. DW (4/23/01 6:21:24 PM) Changed to use a instruction pointer marker, stored with IWorkspaceRoot. DS (4/25/01 11:08:56 AM) On hold waiting for the image. Look at IMarkerImageProvider DW (4/30/01 11:25:56 AM) Replace with real image when received. Define an extension point for IMarkerImageProvider, using a class that dynamically returns either the top stack frame image (solid arrow), or other stack frame image (hollow). DW (4/30/01 3:16:00 PM) Added marker image support to our IP marker. Available in 102. Marker type is "org.eclipse.debug.ui.instructionPointer" Moved to JUI. Java Editor should display this marker type. KUM (5/11/01 4:45:51 PM) The instruction pointer is not displayed. The reason for that is that the DebugView creates the instruction pointer marker which is associated with the workspace root. An editor only displays those markers which are associated with its input element. Thus, the instruction pointer is never displayed as it is not associated with any editor input. Moved back to inbox. DW (5/11/01 11:00:02 AM) Problem is knowing what resource to associated the marker with - i.e. the debugger UI only knows of editor inputs, not resources. DW (5/11/01 2:34:52 PM) Suggest to not implement instruction pointer feature. DW (5/16/01 12:21:08 PM) Has been removed for GA.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48854dc1c25ebdb2079'}

I want to write code like this in a snippet file import com.ibm.swt.widgets.* Shell shell= new Shell(); shell.open(); NOTES: EG (11/13/00 2:36:42 PM) moving to ITPJCORE PM (11/13/00 11:08:57 AM) Import statements are not supposed to be declared inside snippets, however there is a mechanism to attach them to the page, along with global variables. We would need a new parser in order to process imports amongst regular statements+expressions, and it does not fit along with the intent of copying the snippet and paste it inside some method body, and it still works. Moving back to ITPJUI. EG (11/17/00 12:35:31 PM) the "attach" mechanism isn't useful enough to surface in the UI. >>with the intent of copying the snippet and paste it inside some method body, and it still works this is not a very stron argument. The goals of snippet support is to experiment with code not to write it so that it can be pasted into a method. Moving to Active Issues.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48854dc1c25ebdb207a'}

Can start the following snippet two times without terminating the Snippet-VM com.ibm.swt.widgets.Shell shell= new com.ibm.swt.widgets.Shell(); shell.setBounds(10, 10, 100, 100); shell.setText("Shell"); shell.open (); com.ibm.swt.widgets.Display display = shell.getDisplay(); while (!shell.isDisposed ()) { if (!display.readAndDispatch ()) display.sleep (); } NOTES:

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48854dc1c25ebdb207b'}

jkca (11/10/2000 5:21:39 PM) The user should be able to control whether run/debug sessions started in the desktop are terminated when the desktop exits.

enhancement

RESOLVED

WONTFIX

{'$oid': '52e9b48854dc1c25ebdb207c'}

Coloring the snippet output in java style results in funny result. For example an output of the form //text in rendered in red since it is a Java comment. NOTES:

enhancement

RESOLVED

WONTFIX

{'$oid': '52e9b48954dc1c25ebdb207d'}

DM (11/13/00 11:56:46 AM) The normal behavior of Eclipse when selecting a file is to - do nothing if the file is not yet open - go to the file if it is open The "Breakpoints" view of the Debugger Perspective does not go to an already opened file: a double-click is always needed. NOTES: EG (11/13/00 12:17:18 PM) "The normal behavior of Eclipse when selecting a file is to..." This is only the case when the Editor Linking preference is set. When editor linking is enabled than the single click behaviour from above should be implemented. Not for Nov15. Moving to ITPDUI DW (11/30/00 1:44:11 PM) Waiting for new desktop - not clear how we get at the preference via API. possibly: get the IDesktop, get the plugin class for it, cast it to an AbstractUIPlugin, and then get the preference store. EG (1/4/01 12:27:19 PM) There should be an official API to get at the this preference setting. See 1G70YNA: ITPUI:WIN2000 - "Editor Linking" preference issue. DS (1/22/2001 8:56:23 AM) I am assumming that the DebugView should follow the same behavior? Framework in place but waiting on above pr DW (2/26/01 11:51:24 AM) The debug view should not have the same behaviour - i.e. we always show source when a stack frame is selected. DS (3/28/2001 12:56:35 PM) Still waiting on official API to get at the workbench preference for editor linking. DW (6/1/01 11:28:50 AM) See also 1GEN72P: ITPDUI:ALL - Single select of breakpoint prompts for source

enhancement

CLOSED

WONTFIX

{'$oid': '52e9b48954dc1c25ebdb207e'}

EG (1/24/01 9:46:25 AM) "Open on Selection" support in Console View. The console can show references to resources in the case of a walkback. An open on selection command would allow to open the referenced resource. java.util.MissingResourceException: Can't find resource for bundle java.util.PropertyResourceBundle, key Editor.GotoLine.dialog.title at com.ibm.eclipse.ui.editors.text.GotoLineAction.actionPerformed (>>>>>GotoLineAction.java:69<<<<<<) Open On Selection with "GotoLineAction.java:69" will open an editor on GotoLineAction. NOTES: EG (1/24/01 9:57:02 AM) low priority DW (3/15/01 10:50:59 AM) Could be done with an #editorAction that is dynamically contributed by Java Debug UI. JGS (9/25/01 1:51:22 PM) The implementation of OpenOnSelectionAction depends on internal classes in JDT core via the ICodeAssist interface (which only ClassFile & CompilationUnit implement). Thus there is no obvious path to add this functionality to the ConsoleViewer. JGS (10/4/01 10:38:46 AM) Response from EG: I wouldn't stick to the OpenOnSelectionAction it is based on code select and code select is only helpful inside Java code. Let's assume that this is the output in the console: package.SomeType.someMethod(SomeType.java:89). what does the user have to select to do a Go To (file)? Is it enough to put the cursor on the line and you parse the "SomeType:89" or do you want to use the qualifed method name for navigation. If you go for the file name then you can use the workspace API to find the resource with the name. If the file name is ambigous show a dialog and open the editor via the workbench API. Finally use the ITextEditor API to reveal the line. If you go via the fully qualified type name you can use the search engine and look for a type with the given name (see AllTypesSearchEngine for an example) find the matching package and open the IJavaElement. You can give the search engine the package as a scope to constrain the search.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48954dc1c25ebdb207f'}

DW (12/11/00 1:39:04 PM) We should contribute "Run" and "Debug" actions to the pop-up menu for the resource navigator and packages views. DW (12/11/00 1:53:04 PM) Waiting for new desktop. EG (1/4/01 2:32:39 PM) I agree that having run and debug in the pop-up menu would be handy. However, we have to be very careful to not overpopulate the Navigator pop-up menu. It is already large, in particular since run and debug are not context sensitive they will always be in the menu. In addition there is no good support to position a pop-menu contribution. Until there is better support or clear guidelines from the desktop regarding pop-up menu contributions to the navigator we should hold back from doing this. DW (9/6/01 1:14:55 PM) We need to make our "run" and "debug" actions API.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48954dc1c25ebdb2080'}

DW (12/8/00 4:03:27 PM) The old VAME debugger displayed variables that changed on the last step in a different color. We cannot control color but could use a special icon to present this. JGS (8/8/01 1:55:20 PM) Waiting to hear from VI regarding styling TreeItems. JGS (8/14/01 11:42:53 AM) Example code from VI for bolding items in a tree. public static void main (String [] args) { final Display display = new Display (); Shell shell = new Shell (display); final Tree tree = new Tree(shell, SWT.NONE); tree.setBounds(10, 10, 200, 200); for (int i = 0; i < 10; i++) { TreeItem item = new TreeItem(tree, SWT.NONE); item.setText("item "+i); } final TreeItem treeItem = new TreeItem(tree, SWT.NONE); treeItem.setText("accented item "); final Canvas accent = new Canvas(tree, SWT.NONE); accent.setBackground(tree.getBackground()); FontData[] data = tree.getFont().getFontData(); data[0].setStyle(SWT.BOLD); Font font = new Font(display, data[0]); accent.setFont(font); accent.addListener(SWT.Paint, new Listener() { public void handleEvent(Event e) { e.gc.drawText( treeItem.getText(), 1, 1); } }); accent.addListener(SWT.FocusIn, new Listener() { public void handleEvent(Event e) { accent.setBackground(display.getSystemColor (SWT.COLOR_LIST_SELECTION)); accent.setForeground(display.getSystemColor (SWT.COLOR_LIST_SELECTION_TEXT)); tree.setSelection(new TreeItem[] {treeItem}); } }); accent.addListener(SWT.FocusOut, new Listener() { public void handleEvent(Event e) { accent.setBackground(tree.getBackground()); accent.setForeground(tree.getForeground()); } }); TreeEditor editor = new TreeEditor(tree); editor.horizontalAlignment = SWT.LEFT; GC gc = new GC(accent); editor.minimumWidth = gc.textExtent( treeItem.getText()).x; gc.dispose(); editor.grabVertical = true; editor.setEditor(accent, treeItem); shell.open (); while (!shell.isDisposed ()) { if (!display.readAndDispatch ()) display.sleep (); } font.dispose(); display.dispose (); } JGS (8/14/01 11:43:22 AM) The trick is to do this enhancement at a language independent level. This really means something like bolding or coloring variable tree items, since changing text or icons is model presentation specific. But since TreeItems don't support any styling options, we'd have to extend TreeViewer and do some fairly ugly work to get bolding behavior as outlined above. If we had support for styling individual tree items, this could be easier. Filed PR [1GIGJK4: ITPUI:ALL - Feature request: styled tree items] to track this request. Moving to Inactive.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48954dc1c25ebdb2081'}

EG (1/4/01 3:34:24 PM) JDI supports debug on demand. http://192.168.5.9/projects/debaffer/jdi/doc/conninv.html One comment from user testing was: · Debug - Must explicitly run in debug mode for breakpoints to hit. I'm so used to the way VAJ works (no explicit debug mode) that I ended up running the program first, realizing my breakpoints weren't hit, and then running it again in debug mode. With debug on demand breakpoints will not be hit but it brings it closer to the VA/java model. NOTES: DW (1/10/01 10:27:17 AM) To implement this, the launch should work as follows: * A launcher will need to launch the VM using a command line that specifies the --Xrunjdwp options "transport=dt_socket ,address=HOST:PORT,onuncaught=y,suspend=n" (server=n, by default). * A listening connector must be created with a map specifying a port number of PORT. * The launcher will create a DebugTarget with the JDI Factory method that takes an IProcess and a configured ListeningConnector. The debug target will use the listening connector to wait for a connection (if ever). It will be the debug target's repsonsibility to use the #startListening, #accept, and #stopListening methods on the connector. We will have to implement a "JDI proxy debug target" that simply waits for a connection (and has no threads, etc., unless a connection is made). It will have the above specified associated system process to show console output. We can implement the example launcher to experiment with this. The caveats are (since not running in debug mode until an exception is hit): * Breakpoints will not be hit * caught exceptions will not be hit (although we may be able to handle this via the "onthrow" parameter) * You will not be able to manually suspend & inspect any threads It is yet unclear if this is signifigantly faster than simply running in full debug mode that will support all debugging features. It may not become clear until we self-host.

enhancement

CLOSED

WONTFIX

{'$oid': '52e9b48954dc1c25ebdb2082'}

jkca (7/24/2001 4:40:29 PM) While working on the JSP breakpoints, DW suggested I add an attribute to the JSP breakpoint named "classFilePattern". This is used by the JDIDebugTarget to install breakpoints in JSPs. It would be a useful optimization for all breakpoint factories to add this attribute as it would simplify the getTopLevelTypeName code. If we use this mechanism as the sole class file pattern provider, then each debug model could add it's own breakpoints without having to modify the JDIDebugTarget code. This enables us to debug many languages which eventually get compiled to java byte code without futher modifying the Debug Model.

normal

VERIFIED

FIXED

{'$oid': '52e9b48954dc1c25ebdb2083'}

JGS (12/11/00 2:39:52 PM) For involved debug sessions, it would be nice to have the ability to "snapshot" the contents of the variables view. This could be a toolbar button/context menu item that would save the current tree of variables in some format so that at a later time, perhaps on a different launch, the snapshot could be inspected and perhaps compared to the current variables view. In addition, a "diff" ability would be handy to allow two different snapshots to be quickly compared for differences. This could look and function like the catchup browser in VAME. I think users tend to be cautious about debug sessions, trying to make sure that everything is just right before launching them, and moving slowly within them so that they don't miss anything. The ability to snapshot variables might reassure them so that if for instance they forgot to check to see if the "foo" variable was null just before their program blew up, they didn't have to redo a painful debug session.

enhancement

CLOSED

WONTFIX

{'$oid': '52e9b48954dc1c25ebdb2084'}

DW (1/26/01 4:05:15 PM) Suggestion from JJ - be able to add variables (fields/locals) to the watch list before running/debugging a program. Also, the ability to add expressions.

enhancement

RESOLVED

DUPLICATE

{'$oid': '52e9b48954dc1c25ebdb2085'}

DS (1/11/2001 1:41:08 PM) The debugger grids to a halt with big stacks. Investigate optimizations to handle this case. Hit pause and the step return until you can step over "loop()", press pause - the UI wil lock up... possible due to update callbacks on the creation of the stack frames. Sample program: package com.ibm.itp.dui.tests; public class InfiniteLoop2 { public static void main(String[] args) { (new InfiniteLoop2()).loop(); } public void loop() { int i = 0; while (true) { System.out.println("Looping " + i); i++; loop(); try { Thread.sleep(1000); } catch (InterruptedException e) { } } } } DW (1/16/01 5:19:54 PM) We no longer fire creation/terminate events for stack frames or variables. So this might be working better now - it depends on how efficient the viewer.refresh() is. DS (1/23/2001 2:17:44 PM) The refresh with a big stack is still slow and could use some busy cursors but it does not lock up in the original test case. Probably best to revisit once the platform is more stable. DW (6/2/01 12:54:57 PM) Performance is not bad. Currently, we retrieve all frames on a thread suspend, and update variables lazily. However, could investigate JDI implementation to see if it is more efficient to retreive frames one at a time as/if required. The implementation would have to know the underlying stack frame is invalid and needs refreshing. Must investigate JDI code, as it might retrieve all frames anyway. DS (8/3/2001 3:13:06 PM) With our current implementation both the DebugUIPlugin and the DebugContentProvider are doing refresh on suspension we end up updating all of the labels in the stack twice. Reworking so that all of the work is done in one place (DebugUIPlugin), we could remove the spurious second update.

enhancement

RESOLVED

DUPLICATE

{'$oid': '52e9b48954dc1c25ebdb2086'}

JGS (2/8/01 11:41:12 AM) We should modify our I18N of messages so that integer arguments in formatted messages are indicated as such. For example, the following: Error occurred at line {0} in {1} would have to be changed to: Error occurred at line {0,number,integer} in {1} This also means changing the value sent to the MessageFormat object from a String to an Integer. This would allow for automatic conversion to the number rendering style appropriate for the current locale. DW (4/20/01 11:38:45 AM) E-mail from GA: Numbers & Money have different internationalized representations (e.g. whether to use "," or ".") . Typically an application would allow users to enter & printer-back numbers by respecting the locale specific format. Usually this tends to apply mostly to apps dealing seriously with numbers etc (e.g. excel). Tradionionally there are a bunch of number related places where we do not bother to "locale print" a number. For example N errors were detected << we typically don't both with pretty printing N I believe you do not need to NL the format of the case below unless for readability reasons you want to show a line number (for a brkpt) at 2,304 rather than 2034. If you decide this, then you should use the local info to properly format that large number. Thus, we will not format integers.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48954dc1c25ebdb2087'}

DS (1/31/2001 2:43:09 PM) Provide support to drag and drop variables from the variables view to the inspector.

enhancement

CLOSED

WONTFIX

{'$oid': '52e9b48954dc1c25ebdb2088'}

DW (3/2/01 10:38:40 AM) The "propertiesAction" and the "viewFilterAction" extensions should be merged into one extension point called "debugAction". A debug action is an implementor of IAction & ISelectionChangedListener. The plugin xml should specify attributes for which views the action is in, and if the action should be in a menu and/or toolbar. We also need API on the action to determine when it should be visible. DW (3/15/01 9:56:40 AM) Leverage the workbench extensions #viewActions, #editorActions, #popupMenu. The "propertiesAction" extension point is subsumed with the #popupMenu extension defined by the eclipse ui. The "viewFilterAction" cannot be directly converted as we have extra functionallity which is required - i.e. dynamic contribution based on which debug model is currently being debugged. To accomodate the dynamic behavior, we can add a "IDebugViewAdapter" that provides this interface: #addDebugViewListener(IDebugViewListener) : provides access to the current debug model in a view, and when it changes, such that buttons can be added/removed dynamically. #getViewer() -> StructuredViewer : allows access to the viewer to add/remove filters DW (3/19/01 3:27:02 PM) Replaced "propertiesAction" with #popupMenu extension. Made menu group constants and debug views API. @see IDebugViewConstants & IDebugActionConstants. Replaced "viewFilterActions" with #viewAction. No allowance for dynamic appearance yet. Should add manual way to add/remove actions from toolbar - i.e. a "configure tool bar" action that presents a dialog similar to the action sets dialog in the workbench. DW (3/28/01 12:18:00 PM) Sent mail to EG & JW to ask if toolbar configuration is required. DW (3/29/01 5:52:20 PM) EG has agreed to hold off on toolbar configuration until after June. Workbench may provide more support. DW (6/19/01 9:44:24 AM) API request for workbench - allow dynamic configuration of toolbars for views - like action sets in the workbench. Could allow contribution of action sets to views. DS (8/3/2001 1:30:08 PM) @see 1GHW7P7: ITPUI:ALL - Should have action sets for views as well

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48954dc1c25ebdb2089'}

TM (13.03.2001 12:19:27) ITPUI 80/platform 33 1) launch junit.ui.TestRunner under the debugger (JDK 1.2.2) 2) suspend the AWTEventQueue thread 3) observe: The source file for java.lang.Object is opened, but wait(long) is not selected DW (3/14/01 9:03:35 AM) I believe that native methods do not provide line numbers for debug attributes. Need to investigate. DW (3/14/01 10:08:13 AM) When we ask a native stack frame's location for it's line number - it returns -1. We would have to do some special case handling and use the Java Model to do better (i.e. JDI does not help us). Deferred post March 16 drop.

enhancement

CLOSED

WONTFIX

{'$oid': '52e9b48954dc1c25ebdb208a'}

AK (3/13/01 6:56:07 PM) 1. create a snippet file 2. open the file in the editor 3. hit the 'run in package' button note: there are folders on the list NOTES: EG (24.07.2001 23:09:10) should provide option to filter them out.

minor

VERIFIED

FIXED

{'$oid': '52e9b48954dc1c25ebdb208b'}

DW (3/13/01 4:06:25 PM) We cannot create a ThreadDeathInstance until the VM is stopped by a user request. If there are no breakpoints/classloads before the user suspends a thread, they are unable to terminate a thread. If no more classes get loaded, they still are unable to terminate a thread. DW (6/19/01 9:45:35 AM) Investigate thread termination. Is there a better way? Can we terminate a thread on demand, rather than making an instance of ThreadDeath pro-actively?

enhancement

RESOLVED

DUPLICATE

{'$oid': '52e9b48954dc1c25ebdb208c'}

DW (3/15/01 3:34:15 PM) New feature: Breakpoint groups line breakpoints exception function storage load address DW (6/2/01 1:55:35 PM) Groups should be dynamic/extensible as breakpoints are extensible. Can we derive it from the set of defined breakpoints? Is a group an extension point? Must all breakpoints specify a group attribute? The presentation should allow to show/hide groups.

enhancement

RESOLVED

DUPLICATE

{'$oid': '52e9b48954dc1c25ebdb208d'}

DW (3/23/01 10:03:32 AM) Our deployment of class files for HCR and Evaluation does not support remote targets. DW (4/25/01 2:03:16 PM) Could provide a pluggable "deployer" onto a java debug target to handle. We only provide the local implementation. DW (7/31/01 3:33:59 PM) The JDK 1.4 spec uses the approach of replacing types over the wire. A map of (old) reference types and a collection of bytes that should be loaded as the new class are provided at the JDI layer. This takes care of the remote case (i.e. JDI handles it). This seems like a good design to follow. The part missing from the Sun 1.4 spec is a way to distinguish versions of types. DW (8/3/01 9:36:14 AM) In the case of evaluataion, we need to deploy new class files. In the case of HCR, we need to deploy if remote, otherwise deployment is already done by the builder. The launcher knows how to deploy class files - it knows where the target is and the classpath of the target. Thus, it should supply the deployment policy. The new JDI API for HotSwap takes care of loading (but not deployment) - i.e. bytes over the wire. For evaluation we need a deployment policy since the types are new. IDeploymentPolicy - public void deploy(byte[][] classFileBytes, String[][] typeNames) throws DebugException. The IJavaDebugTarget will support a #setDeploymentPolicy(IDeploymentPolicy), which the launcher will provide. For hotswap, we could still deploy files, and then also use JDI API's to notify target of updates. DW (9/17/01 11:32:05 AM) Using AST evaluation would not require file deployment.

normal

VERIFIED

FIXED

{'$oid': '52e9b48954dc1c25ebdb208e'}

DW (3/26/01 5:23:39 PM) If a subclass defines an inst var with the same name as in a supertype, they are difficult to distinguish in the UI - we should show the defining type name in this, when the "qualified" button is pressed.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48954dc1c25ebdb208f'}

DW (3/30/01 10:37:39 AM) Embedded users would like extensible launch modes to be able to define "profile" & "analyse" modes. DW (3/30/01 2:55:01 PM) For June release we will not provide extensible launch modes. The launch modes are: 1) debug - starts a debug session - i.e. results in a debug target that is under "debug" control 2) run - runs a program that cannot be debugged. Profiling, analyzing, tracing can fit into the above story. For example, if tracing does not result in a debuggable target, then it is "run" mode. If it is possible to "debug" and trace at the same time, then it is debug mode. Thus, launchers can be written in the current framework to support tracing/profiling/etc.

enhancement

RESOLVED

WONTFIX

{'$oid': '52e9b48954dc1c25ebdb2090'}

AK (4/5/01 4:27:00 PM) 1. start debugging 2. open another workbench window to java debug 3. when you step through yr program both windows get in synch and show the same thing. e.g. in both you see the same line selected in the same editor etc. this is not what i want as a user - i want the other window be at my disposal - i could browse code, do other stuff. it's hard to understand why stepping into a method in one window opens an editor in the other. or why the focus changes randomly between these windows when you step through the code. NOTES: EG (4/5/2001 11:58:04 AM) Serious issue when user debugs 2 programs and wants to step through them at the same time DW (4/5/01 4:10:14 PM) Perspective switching is also troublesome - i.e. a launcher specifies the java perspective as a debug layout. When I manually switch to the normal debug layout and then step, I end up back in the java perspective. DW (4/6/01 11:21:06 AM) No action on > 1 window. This is a fundamental model/view problem that we do not intend to solve. Multiple views will synch.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48a54dc1c25ebdb2091'}

DW (4/5/01 11:04:06 AM) When no launcher can be determined, we present all launchers to the user to choose from. This list could be reduced if we had a project nature to to filter the list - i.e. a C project only cares about C launchers. DW (5/1/01 11:19:13 AM) Another proposal was to allow a launcher to be attached to any resource - not just the project. DW (5/10/01 10:56:53 AM) Should consider this PR at the same time - they are related issues: 1GC2T2D: ITPDUI:ALL - Launching from Outliner DW (6/2/01 12:24:29 PM) Launching from outliner has been fixed via IProject adapter. DW (6/2/01 12:25:20 PM) Question: What is a launcher be bound to? Current Implementation: A launcher is bound to a project. Issue: Should/Can we allow more flexibility/specification here? I.e. can we support project natures? arbitrary resource? is there a need for anything more?

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48a54dc1c25ebdb2092'}

jkca (4/11/2001 12:02:47 PM) jre-sdk 043 I attempted to change the value of a char to be a newline. I tried both \n and '\n' and neither of them worked. I suspect there are problems with tabs and other escaped characters. Also, there should be support for (char)86. DS (4/11/2001 2:52:48 PM) We should handle the escape chars. (char)86 support is on hold. Need the valid set of escape chars. DS (4/17/2001 1:58:55 PM) Test with '\n'. DS (4/17/2001 3:30:46 PM) Implemented the escape chars. Verify and then move to inactive. DW (4/18/01 4:42:53 PM) Verified. We might also want to support unicode sequences. DW (6/2/01 12:46:40 PM) Possible enhancements: For characters, allow (char)86 unicode For numbers, allow hex input For Strings allow specification of "null"?

normal

RESOLVED

WORKSFORME

{'$oid': '52e9b48a54dc1c25ebdb2093'}

DS (4/11/2001 12:10:08 PM) Method entry breakpoints do not show in the breakpoints ruler for source types. Method entry breakpoints do not work for source types...the signatures are different e.g. ScopeTest#go(String) IMethod#getSignature --> (QString;)V Method#signature --> (Ljava/lang/String;)V DS (4/11/2001 2:39:25 PM) Temp fix was to disable the action for methods that are not binary. DW (4/16/01 1:53:54 PM) Could be re-enabled in future.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48a54dc1c25ebdb2094'}

(4/12/2001 11:43:12 AM) AB: When I step into a class that doesn't have source in the "Project" an empty source window is opened with the title set to "<classname>.class" This window is blank. I would expect to be asked where the source can be found. It is not an option to bring all of the source into a project. I may not even have the source. Something similar to "Step Debug" would be good. This would step until source was found. I avoid stepping into the class loader/security check etc. NOTES:

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48a54dc1c25ebdb2095'}

jkca (4/11/2001 11:47:37 AM) I added one method entry breakpoint to PingPong.java on the pong() method. When I ran, the program halted at System.arraycopy(). See stack trace below. When I hit resume, it halted on pong(), but the next time I hit resume it landed on the System.arraycopy() again. Continuing to hit resume I get pong(), arraycopy(), pong(), arraycopy(), etc. Thread [main] (Suspended) System.arraycopy(Object, int, Object, int, int) String.getChars(int, int, char[], int) line: 542 BufferedWriter.write(String, int, int) line: 205 BufferedWriter(Writer).write(String) line: 132 PrintStream.write(String) line: 271 PrintStream.print(String) line: 411 PrintStream.println(String) line: 534 PingPong.pong() line: 27 PingPong.ping() line: 21 PingPong.loop() line: 8 PingPong.main(String[]) line: 4 DS (4/11/2001 4:17:49 PM) This is a problem with source method entry breakpoints. Currently fixed by not enabling method entry breakpoints action on source methods. jkca (4/12/2001 10:42:06 AM) This does not occur if running against sun jdk1.2.2-005. It does occur with the VM that ships with the Eclipse SDK 043.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48a54dc1c25ebdb2096'}

(4/12/2001 2:00:04 PM) AB: I have added breakpoints on the line that contains the method declaration e.g. public int myMethod() { The breakpoints take but when I run the application in debug they don't get hit. Since the only way to stop a program is by breakpoints it might be good to alert the user that certain breakpoints were removed/ not installed otherwise they may not know why their program didn't stop. NOTES:

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48a54dc1c25ebdb2097'}

TM (4/20/2001 15:31:45) DBG 46 1) put a breakpoint in a java program (VectorTest) and debug it wiht J9. 2) when you hit the breakpoint, suspend the thread called "DebugResponseServer" 3) "Suspend" is still enabled for the "DebugResponseServer" thread. 4) click on different stack frames. 5) observe: you wait a long time until the UI follows. In the variable window, you see: "this = deallocated". I see two problems here: 1) I don't know whether "this" has been deallocated, but I figure we have a timeout problem. We should announce this to the user 2) We need a busy cursor when clicking around. Also, it would be nice to have a "don't know" or "waiting to be suspended" state for threads. DW (4/20/01 9:12:44 AM) We should likely not allow suspending the debug server thread. However, it looks like the request (suspend) should throw an exception and be caught by the UI? Further attempts to terminate threads & Vm also fail but do not report an exceptions. DS (4/23/2001 4:28:05 PM) The exception is "expected"...TimeoutException. We call jdiLogException and silently fail for this exception. This will require some deep thought. DW (5/4/01 4:43:39 PM) The exception is properly propogated to the UI now - Exception occurred suspending thread. DW (5/8/01 12:57:46 PM) The workbench does not display a busy cursor when getting children. The UI locks up while getting children in a non-responding debug target. I got stack frames to show as "<not responding>", but the threads and target remained with old names & state since their state & names are cached. When attempting to resume/suspend a thead in non-responding target, the error message just says "Exception occurred suspending thread". However, it does not specify what exception occurred or if it was a timeout. DW (5/10/01 10:58:35 AM) The only optimization I can see here is to have a debug target/model bail after a few timeouts so we do not lock up the UI. DW (5/10/01 11:24:57 AM) Just make a read-me note for this as a known problem. DW (6/15/01 10:30:03 AM) Added to README.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48a54dc1c25ebdb2098'}

DW (4/27/01 9:53:54 AM) 100 1) Run com.ibm.itp.dui.tests.InfiniteLoop with breakpoint on line 11 2) When program suspends, try to terminate thread 3) instead of terminating, thread resumes, and re-suspends at breakpoint DW (4/27/01 10:11:25 AM) Reproduceable - when I debugged there was no indication of why it is failing. The thread death instance was valid, and the thread was asked to stop. Resuming the thread so it could die just resumed without death. Same problem occurrs with running thread - does not terminate. I am able to terminate the VM. DW (5/11/01 2:30:29 PM) Make this into a readme note. DW (6/15/01 10:34:05 AM) Added to README.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48a54dc1c25ebdb2099'}

DW (4/30/01 12:26:07 PM) Show when a stack frame's source is in/out of synch with the running target. Queued until May 7. DW (5/9/01 2:58:01 PM) Proposal - let the java editor, in the debugger, display whether the source is in/out of synch. Reason - showing on a stack frame basis is less efficient, since more look-ups must be performed. For a stack frame, a type must be resolved from the stack frame's delcaring type name (which the source locator is already doing). From the type, bytes must be resolved (output folder of type's project, or if a binary type, the bytes are already available). From the bytes, a CRC is computed. Here's the sequence of events: 1) The debugger UI asks the source locator to map a stack frame to an object to be displayed. At this point, the source locator could cache a (runtime) CRC for the stack frame's associated type (we will add #getCRC to IJavaStackFrame). The development time CRC can also be cached, if not already. This cache must be cleared when a resource delta indicates the type may have changed. 2) The debugger UI asks the model presentation for an editor input for the object returned in (1). The editor input created can have a special name & icon if CRC are out of synch. This is more efficient since we only display sync info for type that we display source for, rather than every stack frame. Type lookup is minimized. DW (5/10/01 3:38:59 PM) Above story not sufficient. The updating of the editor must be done dynamically with each save/modification. Thus: * source lookup could generate special "debug editor inputs", by returning a special object that the editor utility uses to build "debug editor inputs". * When a java editor has a special debug editor input it must know to update dynamically. * A debug target could be queried for the CRC of a type (given a fully qualified name) * the java ui plugin would maintain a cache of type CRC's in the ide. Problems: * When > 1 debug target, which taget is querried (or do we show source out of synch with at least one target?). * Do we attempt to show in relation to "selected" target? DW (8/2/01 1:27:36 PM) I experimented with calculating CRC's. I was able to use Java model API to get from a CU to its corresponding class file (and calc a CRC with java.util.CRC32). I had to use non-public API to get the bytes for a class file in a zip to calc the CRC. I did not cache the CRCs - just computed on demand (which is less efficient). However, it showed that we will be able to tell what is in/out of synch. Unclear if we want to show in-synch info in the target & stack frames (debug view), or in the editor.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48a54dc1c25ebdb209a'}

DW (5/1/01 1:28:47 PM) Thread groups could be displayed for threads - see what other debuggers do. Perhaps we only show when "show qualified" is on?

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48a54dc1c25ebdb209b'}

MA (5/2/01 6:42:39 PM) 1. create your own plugin. In the startup method set a breakpoint 2. create a preference page in the plugin in the preference page contsructor set a break point 3. start eclipse, go to the preference page, select your plugin's preference page. 4. observe the break point in the plugins startup is not reached. The constructor is invoked but it's plugin has not been started NOTES: JM (5/6/2001 1:17:16 AM) The mechanism for starting a plugin when a class is loaded is clearly working as it is this mechanism which creates and initializes the underlying workspace. I suspect some sort of configuration problem with your environment. Need more information. For example: - what is the execution environment - what is the development environment - possible code example MA (07.05.01 16:44:15) Windows, started from Vame or Eclipse (SH) 1. start with empty workspace 2. Set a breakpoint in ImportOrganizePreferencePage-Constructor (at the super call) and one in initDefaults initDefaults is called from JavaPlugin.startup 3. When started up, Windows->Preferences->Java->ImportOrder (do not look at any other pages) -> first class from Java plugin 4. The constructor breakpoint is reached first JavaPlugin.startup is only reach due to the usage of the Java Plugin in the ImportOrganizePreferencePage-Constructor JM (6/5/2001 11:30:21 PM) Investigate JM (6/8/2001 1:45:17 PM) Moving to DJH for investitgation DJH (6/12/01 4:53:02 PM) Deferred JohnA (10/3/2001 10:11:44 AM) Weird! This seems to me like a debugger problem. If I add a System.out.println in the plugin startup() and the preference page constructor, they seem to be run in the correct order, but it never hits the breakpoint! I tried this with the readme plugin example. Here are the exact steps I followed: 1) Added a no-arg constructor to ReadmePreferencePage 2) Added a println statement in the constructor, and put a breakpoint on the println. 3) Added a startup() method to ReadmePlugin, with a println and a breakpoint 4) Ran the workbench, went to the preference page, and selected "Readme Example" 5) Both printlns occur, then it hits the breakpoint in the pref page constructor, but doesn't show any stack. Moving to ITPDUI for further investigation. DS (10/3/01 2:00:36 PM) Cannot reproduce the problem Did exactly the steps above except of course 4) should read "Debugged" instead of "Ran"

critical

RESOLVED

WORKSFORME

{'$oid': '52e9b48a54dc1c25ebdb209c'}

DS (5/3/01 10:38:38 AM) When a breakpoint is hit, we should select that breakpoint in the breakpoints view.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48a54dc1c25ebdb209d'}

DS (5/4/01 2:07:47 PM) The debug/run history needs to have the full qualified name of the launch element.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48a54dc1c25ebdb209e'}

Self hosting Core. Trying to track some problems in the core library, I added some printf's to get some information. The messages only show up in the console when I exit my workbench. NOTES: DW (5/9/01 7:47:28 PM) Note from RP: "I'm calling flush in Java after calling the native code." DS (5/11/01 3:23:47 PM) Behavior is identical to VAME. We are not getting any notification of the streams being appended for the printfs. Appears that this will be a known limitation specified in a README. DW (6/12/01 3:37:45 PM) Using "flushall()" in the native makes the printf's appear. DW (6/15/01 10:45:07 AM) Added to README.

normal

RESOLVED

INVALID

{'$oid': '52e9b48a54dc1c25ebdb209f'}

DS (5/8/01 10:24:28 AM) If you change the structure of a class, say by adding a instance variable, HCR fails but there is no indication of this to the user. DW (5/8/01 1:01:00 PM) For example: Internal error logged from JDI debug model: org.eclipse.debug.core.DebugException: Hot code replace failed. at org.eclipse.jdt.internal.debug.core.JDIDebugElement.targetRequestFailed(JDIDebugElement.java:284) at org.eclipse.jdt.internal.debug.core.JDIDebugTarget.typesHaveChanged(JDIDebugTarget.java:478) at org.eclipse.jdt.internal.debug.core.JavaHotCodeReplaceManager.notify(JavaHotCodeReplaceManager.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.JavaHotCodeReplaceManager.resourceChanged(JavaHotCodeReplaceManager.java:108) at org.eclipse.core.internal.events.NotificationManager$1.run(NotificationManager.java, Compiled Code) at org.eclipse.core.internal.runtime.InternalPlatform.run(InternalPlatform.java, Compiled Code) at org.eclipse.core.runtime.Platform.run(Platform.java, Compiled Code) at org.eclipse.core.internal.events.NotificationManager.notify(NotificationManager.java, Compiled Code) at org.eclipse.core.internal.events.NotificationManager.broadcastChanges(NotificationManager.java:34) at org.eclipse.core.internal.events.NotificationManager.broadcastChanges(NotificationManager.java, Compiled Code) at org.eclipse.core.internal.resources.Workspace.broadcastChanges(Workspace.java, Compiled Code) at org.eclipse.core.internal.resources.Workspace.endOperation(Workspace.java, Compiled Code) at org.eclipse.core.internal.resources.Workspace.run(Workspace.java:1092) at org.eclipse.ui.actions.WorkspaceModifyOperation.run(WorkspaceModifyOperation.java:79) at org.eclipse.ui.texteditor.AbstractTextEditor.doSave(AbstractTextEditor.java:486) at org.eclipse.jdt.internal.ui.javaeditor.CompilationUnitEditor.doSave(CompilationUnitEditor.java:190) at org.eclipse.ui.internal.EditorManager.saveEditor(EditorManager.java:677) at org.eclipse.ui.internal.WorkbenchPage.saveEditor(WorkbenchPage.java:859) at org.eclipse.ui.internal.SaveAction.run(SaveAction.java:33) at org.eclipse.jface.action.ActionContributionItem.handleWidgetSelection(ActionContributionItem.java:366) at org.eclipse.jface.action.ActionContributionItem.handleWidgetEvent(ActionContributionItem.java, Compiled Code) at org.eclipse.jface.action.ActionContributionItem$ActionListener.handleEvent(ActionContributionItem.java, Compiled Code) at org.eclipse.swt.widgets.EventTable.sendEvent(EventTable.java, Compiled Code) at org.eclipse.swt.widgets.Widget.notifyListeners(Widget.java, Compiled Code) at org.eclipse.swt.widgets.Display.runDeferredEvents(Display.java, Compiled Code) at org.eclipse.swt.widgets.Display.readAndDispatch(Display.java, Compiled Code) at org.eclipse.ui.internal.Workbench.runEventLoop(Workbench.java, Compiled Code) at org.eclipse.ui.internal.Workbench.run(Workbench.java:554) at org.eclipse.core.internal.boot.InternalBootLoader.run(InternalBootLoader.java:825) at org.eclipse.core.boot.BootLoader.run(BootLoader.java:280) at java.lang.reflect.Method.invoke(Native Method) at org.eclipse.core.launcher.Main.basicRun(Main.java:66) at org.eclipse.core.launcher.Main.run(Main.java:314) at org.eclipse.core.launcher.UIMain.main(UIMain.java:46) DW (5/8/01 4:31:20 PM) Currently we have no way to report the failure to the UI - we are in core debug support, in a resource callback, with no knoweldge of the UI. DW (5/9/01 7:01:02 PM) Could be addressed by showing "synch" info in UI. 1GD0ANF: ITPDUI:ALL - CRC's and in synch TM (5/23/2001 14:16:55) I just got this behaviour when adding this line: String sourceName= frame.getSourceName(); to BuildPathSourceLocator.getSourceElement(IStackFrame stackFrame). Do we understand why I'm not getting HCR in this case? DW (8/3/01 9:07:21 AM) Propose to a IHotSwapListener service to the Java debug mdoel, reporting failures to registered clients. - public void hotSwapFailed(IJavaDebugTarget target, String[] classNames, DebugException exception)

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48a54dc1c25ebdb20a0'}

- self hosting workspace. - break point in ExceptionAnalyzer - when break point is hit, Eclipse IDE doesn't come to front. NOTES: DS (5/4/01 2:45:44 PM) Does the Eclipse debugger come to the front of the development ide windows? Currently, the debugger will only come to the front of the active workbench windows of the development ide. It does not become the active "Windows" window. Our behavior is similar to VAME. DB (07.05.2001 09:48:47) Yes, the debugger comes to the front of the development ide windows. I also opt that the debugger becomes the active Windows window. DW (5/8/01 8:14:56 AM) Comment from EG: "the debugger is brought to the front of the windows in its application group and this is what I'd expect. Bringing the window to front even when the eclipse window group isn't active is problematic and feels anti-social to me. On some platforms (Mac, win2000) there is an alerting feature when a background application wants to get attention." Moved to inactive.

normal

VERIFIED

FIXED

{'$oid': '52e9b48a54dc1c25ebdb20a1'}

AK (5/8/01 7:02:07 PM) i'd like to be able to run programs from package view and from the outliner (if there's a main method ) from the outliner - it'd be also acceptable if the run/debug entries showed up only after selecting the main method NOTES: EG (5/9/2001 12:00:27 PM) DW you have looked into this earlier, pls comment. DW (5/10/01 8:23:57 AM) Launching from the packages view works. What is the problem with the package view? Launching from the outliner is problematic. @see 1GC2T2D: ITPDUI:ALL - Launching from Outliner AK (5/10/01 3:31:25 PM) misunderstanding i'm talking about the ui i'd like to have a popup menu entry in the package: run DW (5/10/01 10:57:32 AM) How can we/others contribute the run action to other menus? DS (5/10/01 3:29:31 PM) From IWorkbenchPartSite#registerContextMenu * Registers a pop-up menu with the default id for extension. * The default id is defined as the part id. * <p> * Within the workbench one plug-in may extend the pop-up menus for a view * or editor within another plug-in. In order to be eligible for extension, * the target part must publish each menu by calling <code>registerContextMenu</code>. * Once this has been done the workbench will automatically insert any action * extensions which exist. * </p> * <p> * A menu id must be provided for each registered menu. For consistency across * parts the following strategy should be adopted by all part implementors. * </p> * <ol> * <li>If the target part has only one context menu it should be registered * with <code>id == part id</code>. This can be done easily by calling * <code>registerContextMenu(MenuManager, ISelectionProvider). * <li>If the target part has more than one context menu a unique id should be * defined for each. Prefix each id with the view id and publish these * id's within the javadoc for the target part. Register each menu at * runtime by calling <code>registerContextMenu(String, MenuManager, * ISelectionProvider)</code>. </li> * </ol> * <p> * Any pop-up menu which is registered with the workbench should also define a * <code>GroupMarker</code> in the registered menu with id * <code>IWorkbenchActionConstants.MB_ADDITIONS</code>. Other plug-ins will use this * group as a reference point for insertion. The marker should be defined at an * appropriate location within the menu for insertion. * </p> So the menus we are interested in must be registered and then our plugin must specify the action in the plugin.xml? DW (5/10/01 4:49:29 PM) I believe this goes along with "viewerContribution" extension point. However, we cannot contribute to views that we do not know about (i.e. the packages view). EG (9/24/2001 8:32:54 AM) right, but the JUI doesn't want to add the run action to all its views. instead of a viewer contribution, we could consider an object contribution on IJavaElements. This requires that all views support object contributions. Rework should wait until we have a proposal for launch configurations.

enhancement

RESOLVED

DUPLICATE

{'$oid': '52e9b48a54dc1c25ebdb20a2'}

DW (5/9/01 1:24:17 PM) I did a full build with an HCR target active, and got this in the console. Internal error logged from JDI debug model: org.eclipse.debug.core.DebugException: Hot code replace failed. at org.eclipse.core.runtime.CoreException.<init>(CoreException.java:30) at org.eclipse.debug.core.DebugException.<init>(DebugException.java:36) at org.eclipse.jdt.internal.debug.core.JDIDebugElement.targetRequestFail ed(JDIDebugElement.java:284) at org.eclipse.jdt.internal.debug.core.JDIDebugTarget.typesHaveChanged(J DIDebugTarget.java:478) at org.eclipse.jdt.internal.debug.core.JavaHotCodeReplaceManager.notify( JavaHotCodeReplaceManager.java:145) at org.eclipse.jdt.internal.debug.core.JavaHotCodeReplaceManager.resourc eChanged(JavaHotCodeReplaceManager.java:108) at org.eclipse.core.internal.events.NotificationManager$1.run(Notificati onManager.java:92) at org.eclipse.core.internal.runtime.InternalPlatform.run(InternalPlatfo rm.java:579) at org.eclipse.core.runtime.Platform.run(Platform.java:388) at org.eclipse.core.internal.events.NotificationManager.notify(Notificat ionManager.java:103) at org.eclipse.core.internal.events.NotificationManager.broadcastChanges (NotificationManager.java:34) at org.eclipse.core.internal.events.NotificationManager.broadcastChanges (NotificationManager.java:48) at org.eclipse.core.internal.resources.Workspace.broadcastChanges(Worksp ace.java:99) at org.eclipse.core.internal.resources.Workspace.endOperation(Workspace. java(Compiled Code)) at org.eclipse.core.internal.resources.Workspace.build(Workspace.java:11 1) at org.eclipse.ui.internal.GlobalBuildAction.build(GlobalBuildAction.jav a:60) at org.eclipse.ui.internal.GlobalBuildAction$1.run(GlobalBuildAction.jav a:144) at org.eclipse.jface.operation.ModalContext$ModalContextThread.run(Modal Context.java:99) DW (5/9/01 1:34:08 PM) Is reproduceable with self host env, but when I try a smaller workspace example, it does not happen. DW (5/14/01 4:13:04 PM) Still reproduceable with 5/11/2001 J9.

critical

VERIFIED

FIXED

{'$oid': '52e9b48a54dc1c25ebdb20a3'}

I set a breakpoint on the first line of the method below (i.e., on "LaunchInfo li") and ran the program. The breakpoint was not hit even though the method was run. private static LaunchInfo restoreProfile(URL base) { LaunchInfo li; URL info; try { info = new URL(base,INSTALL_INFO_DIR+LAUNCH_PROFILE_CHKPT); NOTES: DW (5/14/01 8:41:39 AM) See what LF does. DS (5/14/01 9:07:47 AM) Behavior is identical to VAME...you are not on an executable line of code. Proposed feature for post-JUNE.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48a54dc1c25ebdb20a4'}

DW (5/14/01 10:19:36 AM) Readme note: Breakpoints can be set on un-executable lines. For example: Object x; DW (6/15/01 9:22:41 AM) Added to README.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48a54dc1c25ebdb20a5'}

After inspecting new Main(), I would like to send foo() to the result. Currently this is not possible. NOTES: JW (5/11/01 12:26:21 PM) Not a show stopper, but a nice improvement. DW (5/11/01 2:32:56 PM) I believe this would require a new evaluation context - i.e. we have a stack frame context and a global context, but we would need a way to tell the compiler about the inspected result, and how to resolve/refer to it. Would also need a UI to allow the snippet to be associated with the inspect item - i.e. like a VAJ or Smalltalk inspector.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20a6'}

DW (5/16/01 1:57:33 PM) ISession might be a better term for an ILaunch. DW (7/30/01 1:31:42 PM) Currently, a launcher launches a debug session, system process, or connects to an already running program. In the future there could be other launchers that may do things such as start a tracing session, or start profiling session, etc. The term "Launch" (as a noun, i.e. the result of launching) feels awkward and could be better described as a session. Propose to rename ILaunch to ISession ILaunchManager to ISessionManager ILaunchListener to ISessionListener Launch to Session Change DebugPlugin accessor #getLaunchManager -> #getSessionManager ISession would be defined as: [ package org.eclipse.debug.core; /* * (c) Copyright IBM Corp. 2000, 2001. * All Rights Reserved. */ import org.eclipse.debug.core.model.*; import org.eclipse.core.runtime.IAdaptable; /** * A session is the result of launching a program in debug mode, connecting * to a program that is already running in debug mode, and/or * starting one or more system processes (which may or may not be in debug mode). * <p> * This interface is not intended to be implemented by clients. Clients * should create instances of this interface by using the implementation * provided by the class <code>Session</code>. * </p> * <p> * <b>Note:</b> This class/interface is part of an interim API that is still under development and expected to * change significantly before reaching stability. It is being made available at this early stage to solicit feedback * from pioneering adopters on the understanding that any code that uses this API will almost certainly be broken * (repeatedly) as the API evolves. * </p> * @see Session */ public interface ISession extends ITerminate, IAdaptable { /** * Returns the debug target associated with this session, or <code>null</code> * if no debug target is associated with this session. * * @return the debug target associated with this session, or <code>null</code> */ IDebugTarget getDebugTarget(); /** * Returns the object used to create this session - for example the program that * was launched. Cannot return <code>null</code>. * * @return the launched object */ Object getElement(); /** * Returns the launcher that was used to create this session. Cannot return <code>null</code>. * * @return the launcher */ ILauncher getLauncher(); /** * Returns the processes associated with this session, * or an empty collection if no processes are associated with * this session. * * @return array of processes */ IProcess[] getProcesses(); /** * Returns the source locator to use when locating source elements associated * with the debug target of this session, or <code>null</code> * if source mapping is not supported. * * @return the source locator */ ISourceLocator getSourceLocator(); /** * Returns the mode this session was launched in - one of the mode constants defined by * the session manager. * * @return the launch mode * @see ISessionManager */ public String getMode(); } ] An ISessionListener would be defined as: [ package org.eclipse.debug.core; /* * (c) Copyright IBM Corp. 2000, 2001. * All Rights Reserved. */ /** * A session listener is notified of sessions as they * are added to and removed from the session manager. * <p> * Clients may implement this interface. * </p> * <p> * <b>Note:</b> This class/interface is part of an interim API that is still under development and expected to * change significantly before reaching stability. It is being made available at this early stage to solicit feedback * from pioneering adopters on the understanding that any code that uses this API will almost certainly be broken * (repeatedly) as the API evolves. * </p> * @see ISession */ public interface ISessionListener { /** * Notifies this listener that the specified session has been removed * from the session manager. * * @param session the deregistered session */ void sessionRemoved(ISession session); /** * Notifies this listener that the specified session has been added to * the session manager. * * @param session the registered session */ void sessionAdded(ISession session); } ] ISessionManager would be defined as: [ package org.eclipse.debug.core; /* * (c) Copyright IBM Corp. 2000, 2001. * All Rights Reserved. */ import org.eclipse.debug.core.model.IDebugTarget; import org.eclipse.debug.core.model.IProcess; import org.eclipse.core.resources.IProject; import org.eclipse.core.runtime.CoreException; import org.eclipse.core.runtime.IAdaptable; /** * The session manager manages the set of registered sessions, maintaining * a collection of active processes and debug targets. Clients interested * in session notification may register with the session manager. * <p> * For convenience, a default launcher may be associated with a project. * The preference is stored as a persistent property with the project. * The debug UI plug-in uses this preference to help determine which * launcher to invoke when the the run or debug button is pressed, based on * the selection in the workbench. If a launcher cannot be determined, * the user is prompted to choose a launcher. * </p> * <p> * Clients are not intended to implement this interface. * </p> * <p> * <b>Note:</b> This class/interface is part of an interim API that is still under development and expected to * change significantly before reaching stability. It is being made available at this early stage to solicit feedback * from pioneering adopters on the understanding that any code that uses this API will almost certainly be broken * (repeatedly) as the API evolves. * </p> * @see ISession * @see ISessionListener */ public interface ISessionManager extends IAdaptable { /** * A session in a normal, non-debug mode(value <code>"run"</code>). */ public static final String RUN_MODE= "run"; /** * A session in a special debug mode (value <code>"debug"</code>). */ public static final String DEBUG_MODE= "debug"; /** * Adds the given listener to the collection of registered session listeners. * Has no effect if an identical listener is already registerd. * * @param listener the listener to register */ void addSessionListener(ISessionListener listener); /** * Removes the specified session and notifies listeners. Has no effect * if an identical session is not already registered. * * @param session the session to remove */ void removeSession(ISession session); /** * Returns the session the given process is contained in, or <code>null</code> * no registered session contains the process. * * @param process the process for which to find a session * @return the session containing the process, or <code>null</code> if none */ ISession findSession(IProcess process); /** * Returns the session the given debug target is contained * in, or <code>null</code> no registered session contains the debug target. * * @param target the debug target for which to find a session * @return the session containing the debug target, or <code>null</code> if none */ ISession findSession(IDebugTarget target); /** * Returns the collection of debug targets currently registered with this * session manager. * * @return an array of debug targets */ IDebugTarget[] getDebugTargets(); /** * Returns the default launcher for the given project, * or <code>null</code> if no default launcher has been set. * The default launcher is stored as a persistent property * with a project. * * @param project the project for which to retrieve a default launcher * @return the defualt launcher, or <code>null</code> if none * @exception CoreException if an error occurs accessing the * persistent property */ ILauncher getDefaultLauncher(IProject project) throws CoreException; /** * Returns the collection of registered launchers that can operate in the * specified mode - run or debug. * * @return an array of launchers */ ILauncher[] getLaunchers(String mode); /** * Returns the collection of registered launchers. * * @return an array of launchers */ ILauncher[] getLaunchers(); /** * Returns the collection of sessions currently registered * with this session manager. * * @return an array of sessions */ ISession[] getSessions(); /** * Returns the collection of processes currently registered with this * session manager. * * @return an array of processes */ IProcess[] getProcesses(); /** * Adds the specified session to the session manager. * Has no effect if an identical session is already registered. * * @param session the session to register */ void addSession(ISession session); /** * Removes the given listener from the collection of registered session listeners. * Has no effect if an identical listener is not already registerd. * * @param listener the listener to deregister */ void removeSessionListener(ISessionListener listener); /** * Sets the default launcher for the given project as a persistent property. * * @param project the project for which to set the preference * @param launcher the launcher preference * @exception CoreException if an error occurs setting the persistent property */ void setDefaultLauncher(IProject project, ILauncher launcher) throws CoreException; } ]

enhancement

RESOLVED

WONTFIX

{'$oid': '52e9b48b54dc1c25ebdb20a7'}

JGS (5/10/01 7:03:19 PM) The icons for variables underneath InspectItems in the Inspector look squished. They don't seem to get as much horizontal space as in the variables view, and consequently look different. JGS (5/16/01 7:22:24 PM) In fact, the whole inspector tree looks like things are squeezed together more than in the variables view tree. JGS (5/17/01 11:50:37 AM) The reduced space in the inspector also results in the 'S' for static variables getting chopped off. JGS (5/18/01 10:47:58 AM) The problem is that the icons of children of an InspectItem are OverlayIcons, which are 22 pixels wide, whereas the icon for an InspectItem is a regular icon, which is only 16 pixels wide. Apparently, the Tree decides how much space to allow for icons for TreeItems based on how much space the parent item requires (or it could be how much space the *first* TreeItem requires). Because of this, 22 pixel wide icons were getting squeezed into 16 pixels. There are a number of ways to fix. We could make InspectItem public API, so that JDIModelPresentation could handle its icon, thus allowing it to become an OverlayIcon. But for now, this is fixed by simply widening the InspectItem icon ("expression_obj.gif") to be 22 pixels wide. This doesn't require any code changes. After verifying, move this PR to Inactive for reconsideration after first release? DW (5/21/01 6:46:53 PM) Verified.

minor

RESOLVED

WORKSFORME

{'$oid': '52e9b48b54dc1c25ebdb20a8'}

DW (5/16/01 11:26:07 AM) It would be good if the model presentation were also the content provider for a view. This would allow the presentation to handle sorting/filtering on its own. DW (8/10/01 11:15:47 AM) Currently, the debug UI implements content providers for the debug views. This means that it controls the sorting of elements in the view. The JDI debug model has to do sorting for variables - i.e. args, locals, statics, etc. It would be more desirable to have sorting handled in the presentation layer. DW (9/20/2001 10:13:38 AM) We use actions to add/remove filter (eg. final & static filters). There is no simple way for a presentation to add a sorter to a view. Currerntly the model implements sorting. However, the only view where sorting is applicable is in the "variable" view. Order is defined in the debug/launch view (eg. stack frames are top-down). In views where there are elements from different debug models (breakpoints view), we would have to have a top level sorter that first grouped elements (by debug model), and then sorted within groups. This would require an API on the presentation to "get a sorter for a specific view". Deferred to P3.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48b54dc1c25ebdb20a9'}

DW (5/16/01 11:28:45 AM) The toggles that we contribute - qualified names & show types - could have their state persisted across workbench invocations. JMB (8/17/2001 9:50:50 AM) Deferred. Need ActionContributionItems to properly set their ids. They don't yet.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20aa'}

Steps to see the problem: 1) Set a breakpoint in some code, and run to it. 2) Notice that the "current" line of code is selected 3) Now click somewhere in the editor 4) The "current" line is no longer selected (because the "selection" folows the caret) Would be nice if the user always knew where the current line was, even if they select somewhere else in the text editor. You could do this by setting the background color in the StyledText widget.... (maybe) Or you could do it the "VAME" way, by having a little red arrow in the "breakpoint bar" to the left of the text. NOTES: DS (5/17/01 2:05:51 PM) Duplicate of 1G3A7CG: ITPDUI:ALL - Displaying Instruction pointer DW (5/17/01 5:48:23 PM) Deferred to post June.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48b54dc1c25ebdb20ab'}

DW (5/18/01 8:58:12 AM) Can we re-use editors? And just re-set the input? Can we look at editor IDs? DW (5/18/01 1:37:06 PM) EditorPart.setInput(IEditorInput) is not part of the API - just the implementation. DW (6/19/01 9:37:28 AM) Requires API request/change from workbench. DS (8/3/2001 3:39:38 PM) @see 1GHWE4M: ITPUI:ALL - Reuse of editors: changing the input

enhancement

RESOLVED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20ac'}

DW (5/18/01 12:43:36 PM) We only install BPs in the first class with a given name - not all. And not any others that get loaded with the same name. DW (5/18/01 2:54:42 PM) This could cause trouble with snippet editor that reloads classes on each evaluation. DW (6/15/01 9:34:52 AM) Added to README. DW (8/7/01 3:13:56 PM) Should be re-worked when breakpoints are re-done as objects. DW (9/20/2001 10:09:08 AM) Fixed with breakpoint re-work.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20ad'}

DS (5/17/01 4:23:29 PM) The remove all terminated launches, deregisters the launches one at a time. This causes flash and is inefficient. Would require changes to IlaunchManager#deregisterLaunches or something similar DW (5/17/01 4:51:19 PM) Deferred.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20ae'}

Most of the time (but not allways) I don't get a + sign in front of UIMain when I debug Eclipse using J9. Also hitting a breakpoint stops the program execution but doesn't reveal the corresponding method on the stack frame. Zipped my workspace if it helps. NOTES: DW (5/21/01 8:02:00 PM) Sounds like unable to retrieve threads from VM on startup. Try J9 on Win2000. DW (5/22/01 3:35:13 PM) Unable to reproduce - asked DB to use debug options to look for error messages. DB (23.05.2001 10:18:24) Got the following on my console when I started J9 in debug mode. Had no plus in front Internal error logged from JDI debug model: org.eclipse.jdi.TimeoutException at org.eclipse.jdi.internal.connect.PacketReceiveManager.getReply(PacketReceiveManager.java, Compiled Code) at org.eclipse.jdi.internal.connect.PacketReceiveManager.getReply(PacketReceiveManager.java:127) at org.eclipse.jdi.internal.MirrorImpl.requestVM(MirrorImpl.java:152) at org.eclipse.jdi.internal.MirrorImpl.requestVM(MirrorImpl.java:208) at org.eclipse.jdi.internal.VirtualMachineImpl.getIDSizes(VirtualMachineImpl.java:704) at org.eclipse.jdi.internal.VirtualMachineImpl.packetSendManager(VirtualMachineImpl.java:231) at org.eclipse.jdi.internal.MirrorImpl.requestVM(MirrorImpl.java:148) at org.eclipse.jdi.internal.MirrorImpl.requestVM(MirrorImpl.java:170) at org.eclipse.jdi.internal.request.EventRequestImpl.enable(EventRequestImpl.java:200) at org.eclipse.jdt.internal.debug.core.JDIDebugTarget.initializeRequests(JDIDebugTarget.java:233) at org.eclipse.jdt.internal.debug.core.JDIDebugTarget.initialize(JDIDebugTarget.java:197) at org.eclipse.jdt.internal.debug.core.JDIDebugTarget.<init>(JDIDebugTarget.java:165) at org.eclipse.jdt.debug.core.JDIDebugModel$1.run(JDIDebugModel.java:144) at org.eclipse.core.internal.resources.Workspace.run(Workspace.java:1099) at org.eclipse.jdt.debug.core.JDIDebugModel.newDebugTarget(JDIDebugModel.java:148) at org.eclipse.jdt.internal.ui.launcher.J9DebugLauncher.run(J9DebugLauncher.java, Compiled Code) at org.eclipse.jdt.internal.ui.launcher.JavaApplicationLauncher$1.run(JavaApplicationLauncher.java:190) at org.eclipse.jface.operation.ModalContext$ModalContextThread.run(ModalContext.java:102) DB (23.05.2001 10:30:01) BTW: the walkback doesn't show up in the normal .log file. DB (23.05.2001 10:30:57) After terminating the target I got the following on the console java.lang.NullPointerException at org.eclipse.jdt.internal.debug.core.JDIDebugTarget.removeAllChildren(JDIDebugTarget.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.JDIDebugTarget.terminate0(JDIDebugTarget.java:1212) at org.eclipse.jdt.internal.debug.core.JDIDebugTarget.handleVMDisconnect(JDIDebugTarget.java:706) at org.eclipse.jdt.internal.debug.core.EventDispatcher.dispatch(EventDispatcher.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.EventDispatcher$1.run(EventDispatcher.java:153) at org.eclipse.core.internal.resources.Workspace.run(Workspace.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.EventDispatcher.run(EventDispatcher.java, Compiled Code) at java.lang.Thread.run(Thread.java:479) DW (5/23/01 10:49:35 AM) Increasing the timeout fixes the problem. Fixed NPE by safeguarding the removeAllChildren with a null check. DW (5/23/01 1:46:51 PM) We could propogate creation/init failures by having the debug target creation methods throw a debug exception on failure - this way, the launch could report the failure to the user. Would require an API change. DW (5/25/01 9:08:29 AM) Add a readme note for this - increase timeout if the problem occurrs. DW (6/15/01 10:20:39 AM) Added to README.

critical

CLOSED

WONTFIX

{'$oid': '52e9b48b54dc1c25ebdb20af'}

DS (5/21/01 10:51:52 AM) When selecting a new context, the current context should be selected. DW (5/24/01 9:35:14 PM) Fixed. JGS (5/29/01 4:16:15 PM) This works for real packages and package fragments, but for default packages, if I select the default package for my project, then come back in, the default package for rt.jar is selected. DW (5/30/01 1:58:34 PM) Not easily improved. The interaction between the "run in package" action and the scrapbook is via package name - not package handles. Would require changing the interaction to use element handle, and persisting the handle id. DW (5/30/01 3:04:47 PM) Deferred to post-june. DW (8/10/01 11:09:22 AM) This is a non-breaking change that we can fix in ZRH code. JGS (8/15/01 3:22:09 PM) Made changes to RunInPackageAction and JavaSnippetEditor to persist the IPackageFragment handle identifier instead of just the package name. Outstanding problem is that the SelectionDialog returned by JavaUI.createPackageDialog() only seems to take package names in its setInitialSelection() method. We need a way to set the intial selection on the dialog with the actual IPackageFragment. Sent mail to EG for advice. JGS (8/16/01 5:13:15 PM) Changes made to JavaSnippetEditor: Changed member def: private String fPackageHandle= null; Changed method def: protected void doSetInput(IEditorInput input) throws CoreException { super.doSetInput(input); fPackageHandle = getPage().getPersistentProperty(new QualifiedName(JavaPlugin.getPluginId(), PACKAGE_CONTEXT)); } Changed methods def: /** * Set the handle identifier of the run-in package fragment */ public void setPackage(String packageHandle) { fPackageHandle= packageHandle; // persist try { getPage().setPersistentProperty(new QualifiedName(JavaPlugin.getPluginId(), PACKAGE_CONTEXT), packageHandle); } catch (CoreException e) { ErrorDialog.openError(getShell(), SnippetMessages.getString("SnippetEditor.error.packagecontext"), null, e.getStatus()); //$NON-NLS-1$ } } /** * Get the handle identifier of the run-in package fragment */ public String getPackage() { return fPackageHandle; } Changed method def: protected IEvaluationContext getEvaluationContext() { if (fEvaluationContext == null) { IJavaProject project= getJavaProject(); fEvaluationContext= project.newEvaluationContext(); } if (fPackageHandle != null) { String packageName = JavaCore.create(fPackageHandle).getElementName(); fEvaluationContext.setPackageName(packageName); } return fEvaluationContext; } Changes made to RunInPackageAction: Changed methods def: /** * The user has invoked this action. */ public void run() { Shell s= fEditor.getSite().getShell(); IPackageFragment result= choosePackage(s); if (result != null) { fEditor.setPackage(result.getHandleIdentifier()); } } private IPackageFragment choosePackage(Shell shell) { try { IJavaProject p= fEditor.getJavaProject(); //fix for 1G472LK: ITPJUI:WIN2000 - Package selection dialog must qualify package names regarding source folders SelectionDialog dialog= JavaUI.createPackageDialog(shell, p, IJavaElementSearchConstants.CONSIDER_BINARIES); dialog.setTitle(SnippetMessages.getString("RunInPackage.dialog.title")); //$NON-NLS-1$ dialog.setMessage(SnippetMessages.getString("RunInPackage.dialog.message")); //$NON-NLS-1$ String pkgHandle= fEditor.getPackage(); if (pkgHandle != null) { //String pkgName = JavaCore.create(pkgHandle).getElementName(); IPackageFragment pf = (IPackageFragment)JavaCore.create(pkgHandle); dialog.setInitialSelections(new Object[] {pf}); } dialog.open(); Object[] res= dialog.getResult(); if (res != null && res.length > 0) return (IPackageFragment)res[0]; } catch (JavaModelException e) { } return null; } JGS (8/17/01 10:42:55 AM) Need to integrate above code, as well as whatever fix is found for setting initial selection on dialog. JGS (8/20/01 9:43:14 AM) New PR created to address inability to set initial selection. [1GIRGYV: ITPJUI:WIN2000 - SelectionDialog.setInitialSelection broken for all JDT dialogs]

minor

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20b0'}

DS (5/21/01 7:05:24 PM) If you evaluate a method that has "void" as its return type out of ThreadEvaluationContext#convertResult ReferenceType ref= fResultType.reflectedType(); String sig = ref.signature(); sig on JDK1.2.2, is "Lvoid;" sig on J9, is "V" Currently ThreadEvaluationContext is handling this difference. DW (5/22/01 12:41:16 PM) Not critical - we have a temp fix in to look for "Lvoid".

normal

CLOSED

WONTFIX

{'$oid': '52e9b48b54dc1c25ebdb20b1'}

JW (5/22/2001 7:25:22 AM) 108 I put a breakpoint in a loop. Everytime I resume, all of the state is flushed, and then immediately reconstructed. I believe you have already handled this case for step. DW (5/22/01 9:28:22 AM) We cannot assume anything about a resume - i.e. where we will end up on the next suspend. Thus we throw state away. When stepping we can assume that we will end up close to where we left off. Too risky of a change for now - would have to change incremental computation of new/removed stack frames. DW (9/27/2001 4:37:00 PM) This will be useful for evaluations too. An evaluation causes a thread to resume/suspend. This causes a UI flash and the variables view to collapse.

normal

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20b2'}

It is a bit odd that the "Step" actions on the Debugger tool bar are separated from the "Resume", "Suspend", "Terminate" actions by the "Disconnect" and "Remove all terminated" tools. It would seem more natural to have the "run" and "step" tools side-by-side. One configuration that makes more sense to me is: Step In, Step Over, Step Out || Resume, Suspend, Terminate || Remove all terminated, Disconnect || other... NOTES: CM (5/22/2001 11:46:27 PM) Note that I also swapped the order of Remove all terminated & Disconnect so that Terminate was immediately followed by Remove all terminated. (after the separator) DW (5/24/01 10:37:47 AM) Deferred.

normal

CLOSED

WONTFIX

{'$oid': '52e9b48b54dc1c25ebdb20b3'}

Not sure if this is desired behaviour. If you create a Scrapbook page, execute some code, you get an associated VM in the list of process'. If you rename the scrapbook page and execute some other code, you get a new process with the new Scrapbook name and the old one is still left running. Should it be terminated? NOTES: EG (5/23/2001 11:42:18 AM) it is not the desired behaviour. KUM what is the method to hook into for tracking whether an editor is renamed. DS (9/6/01 5:39:29 PM) Have added code to terminate the VM on a name change of the editor but 1GJNB52: ITPUI:ALL - ToolItems in EditorToolBarManager can get out of synch with the state of the IAction results in the tool item in the tool bar still being enabled (initially) when the editor is reactivated.

critical

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20b4'}

I went to debug a CU by selecting it and clicking the debug tool item. I was prompted to select a launcher and I also had to select the only available class on the second page. The second step shouldn't be necessary. The next button on the first page should be disabled. NOTES: DW (5/25/01 2:35:01 PM) The first time you launch something in your workspace, you must go through this pain...This is due to the debugger being pluggable for different lauguages. In this case, the launcher selection is generic debug support and the choosing of a class to launch is java specific debug support. To promote lazy plugin loading and to avoid launchers doing exhaustive searching for launchable targets, the "launcher selection page" does not poll the "pluggable launch page" to see if it can finish with the current selection. Once you have selected a defualt launcher for a project, the launcher selection page will not bother you again. Moved to inactive for post-June consideration.

enhancement

RESOLVED

WONTFIX

{'$oid': '52e9b48b54dc1c25ebdb20b5'}

The UI allows you to add breakpoints beside a method definition, yet we don't break on enter of the method. We either need to support break on enter (my preference <g>) or disallow setting them there. NOTES: DW (5/24/01 8:59:23 AM) This actually causes a "break on exit". Notice that if you step out of a method, the last line executed is the method declaration - this is an artifact of JDI (or java?). We support method entry breakpoints for .class files only (see pop-up menu in outliner). We do not intend to support method entry breakpoints for .java's for June. Will add a README note to describe "break on exit" behavior.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48b54dc1c25ebdb20b6'}

EG (5/24/2001 4:14:09 AM) The current API to set an exception breakpoint requires passing in an IType. This requires that the type is present in the workspace. Could the API be extended so that we can also add an exception breakpoint by passing in a String? This is a request for after June. NOTES: DW (5/24/01 10:08:20 AM) We need to create a marker - that is why we need the IType (i.e. to associate the marker with the IType or its project). We would need to change the API so we know where to create the marker. Moved to Inactive (after June).

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20b7'}

DW (5/24/01 9:58:51 AM) I evaluated "VectorTests.main(new String[]{})" java.lang.NullPointerException at org.eclipse.jdt.internal.ui.snippeteditor.JavaSnippetEditor.evaluationComplete(JavaSnippetEditor.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.ThreadEvaluationContext.run(ThreadEvaluationContext.java:115) at java.lang.Thread.run(Thread.java:479) I think the thread died after the eval because the program does a "system.exit(0)". The page remained disabled and I could not do anymore evals. Had to close/re-open page. DW (5/24/01 10:01:32 AM) System.exit kills us because the target dies, and we have no more access to the value (result). DW (5/24/01 11:36:43 AM) A call to #evaluationEnds should be made when the VM terminates. DW (5/24/01 3:56:26 PM) Added call to #evaluationEnds when VM Terminates. Remaining problem is System.exit(0) & NPE. DW (5/25/01 2:18:19 PM) Fixed NPE. The user now sees a "VMDisconnect' stack dump in the snippet editor showing that the target disconnected. Could fix (with Security Manager?) or leave as a README note. DW (6/15/01 9:59:56 AM) Added to README.

critical

RESOLVED

WONTFIX

{'$oid': '52e9b48b54dc1c25ebdb20b8'}

DW (5/24/01 12:38:30 PM) Currently the launch API returns a boolean (launch success). However, it would be better to return the resulting launch, and have the debugger automatically register the launch. Rather than having the client register the launch. This would have avoided 1GDQJE7: ITPDUI:WINNT - Launch did not change perspective & console output DW (7/30/01 9:12:24 AM) Currently, an ILauncher returns a boolean indicating if a launch was successful, and is responsible for registering each successful launch with the launch manager. We propose that a launcher should instead return the session that resulted in the call to #launch, or null if the launch did not succeed, or was cancelled. It will be the debug plugin's responsibility to register the session returned from the launcher, thus relieving the clients of this responsibility (which is something every client would have to do, which should reduce duplication of functionality among clients).

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48b54dc1c25ebdb20b9'}

DW (5/24/01 1:14:15 PM) Placing a line breakpoint on a method declaration line causes execution to stop when exiting the method (not while entering). This is an artifact of JDI/Java. DW (6/15/01 9:42:12 AM) Addded to Readme.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20ba'}

JGS (5/29/01 12:14:31 PM) Swipe "InfiniteLoop.main(null);" in a scrapbook page and run. Once running, click main toolbar Terminate button. Consistently get following walkback: com.sun.jdi.VMDisconnectedException: Got IOException from Virtual Machine at org.eclipse.jdi.internal.connect.PacketReceiveManager.getReply(PacketReceiveManager.java, Compiled Code) at org.eclipse.jdi.internal.connect.PacketReceiveManager.getReply(PacketReceiveManager.java, Compiled Code) at org.eclipse.jdi.internal.MirrorImpl.requestVM(MirrorImpl.java, Compiled Code) at org.eclipse.jdi.internal.MirrorImpl.requestVM(MirrorImpl.java, Compiled Code) at org.eclipse.jdi.internal.ObjectReferenceImpl.invokeMethod(ObjectReferenceImpl.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.JDIThread.invokeMethod(JDIThread.java:426) at org.eclipse.jdt.internal.debug.core.ThreadEvaluationContext.runSnippet(ThreadEvaluationContext.java:264) at org.eclipse.jdt.internal.debug.core.ThreadEvaluationContext.acceptClassFiles(ThreadEvaluationContext.java:209) at org.eclipse.jdt.internal.core.eval.RequestorWrapper.acceptClassFiles(RequestorWrapper.java, Compiled Code) at org.eclipse.jdt.internal.eval.EvaluationContext.evaluate(EvaluationContext.java:221) at org.eclipse.jdt.internal.eval.EvaluationContext.evaluate(EvaluationContext.java:239) at org.eclipse.jdt.internal.core.eval.EvaluationContextWrapper.evaluateCodeSnippet(EvaluationContextWrapper.java:166) at org.eclipse.jdt.internal.debug.core.ThreadEvaluationContext.doEvaluation(ThreadEvaluationContext.java:128) at org.eclipse.jdt.internal.debug.core.ThreadEvaluationContext.run(ThreadEvaluationContext.java:114) at java.lang.Thread.run(Thread.java:479) DS (5/30/2001 11:23:48 AM) Walkback shows in scrapbook. Try to be more friendly in the future (usability) DW (6/15/01 10:02:14 AM) Added to README.

minor

RESOLVED

WORKSFORME

{'$oid': '52e9b48c54dc1c25ebdb20bb'}

If I create a simple Java class and set a breakpoint in the code, then move the file to another package then the breakpoint is lost. NOTES: eg (6/1/2001 3:39:19 AM) moving to DUI. preserving the breakpoint would be nice, but isn't critical. DW (6/1/01 9:20:23 AM) Deferred. This changes the type name etc, and is tricky to preserve.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48c54dc1c25ebdb20bc'}

DW (5/31/01 8:28:26 AM) Put a method entry breakpoint on ClassLoader.loadClassInternal(String). When the breakpoint is hit, press step over... (This is related to the step causing a resume...) com.sun.jdi.IncompatibleThreadStateException: Thread was not suspended. at org.eclipse.jdi.internal.ThreadReferenceImpl.frames(ThreadReferenceImpl.java, Compiled Code) at org.eclipse.jdi.internal.ThreadReferenceImpl.frames(ThreadReferenceImpl.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.JDIThread.getUnderlyingFrames(JDIThread.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.JDIThread.getChildren0(JDIThread.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.JDIThread.setRunning(JDIThread.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.JDIThread.handleStep(JDIThread.java:609) at org.eclipse.jdt.internal.debug.core.EventDispatcher.dispatchStepEvent(EventDispatcher.java:131) at org.eclipse.jdt.internal.debug.core.EventDispatcher.dispatch(EventDispatcher.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.EventDispatcher$1.run(EventDispatcher.java, Compiled Code) at org.eclipse.core.internal.resources.Workspace.run(Workspace.java, Compiled Code) at org.eclipse.jdt.internal.debug.core.EventDispatcher.run(EventDispatcher.java, Compiled Code) at java.lang.Thread.run(Thread.java:479) DW (5/31/01 4:50:10 PM) Could not reproduce after fixing the step/resume problem. DW (5/31/01 5:01:43 PM) However, placing an entry breakpoint in that class and then stepping twice causes: org.eclipse.debug.core.DebugException[5010]: com.sun.jdi.InvalidStackFrameException at org.eclipse.jdi.internal.MirrorImpl.defaultReplyErrorHandler(MirrorImpl.java(Compiled Code)) at org.eclipse.jdi.internal.StackFrameImpl.thisObject(StackFrameImpl.java:176) at org.eclipse.jdt.internal.debug.core.JDIStackFrame.getChildren0(JDIStackFrame.java:186)Internal error logged from UI: at org.eclipse.jdt.internal.debug.core.JDIDebugElement.getChildren(JDIDebugElement.java:151) at org.eclipse.debug.internal.ui.VariablesContentProvider.doGetChildren(VariablesContentProvider.java:29) at org.eclipse.debug.internal.ui.BasicContentProvider.getChildren(BasicContentProvider.java:95) at org.eclipse.debug.internal.ui.VariablesContentProvider.getElements(VariablesContentProvider.java:43) at org.eclipse.jface.viewers.AbstractTreeViewer.getRawChildren(AbstractTreeViewer.java(Compiled Code)) at org.eclipse.jface.viewers.AbstractTreeViewer.getRawChildren(AbstractTreeViewer.java(Compiled Code)) at org.eclipse.jface.viewers.StructuredViewer.getFilteredChildren(StructuredViewer.java(Compiled Code)) at org.eclipse.jface.viewers.StructuredViewer.getSortedChildren(StructuredViewer.java(Compiled Code)) at org.eclipse.jface.viewers.AbstractTreeViewer.createChildren(AbstractTreeViewer.java(Compiled Code)) at org.eclipse.jface.viewers.AbstractTreeViewer$3.run(AbstractTreeViewer.java:651) at org.eclipse.jface.viewers.StructuredViewer.preservingSelection(StructuredViewer.java:534) at org.eclipse.jface.viewers.AbstractTreeViewer.inputChanged(AbstractTreeViewer.java:643) at org.eclipse.jface.viewers.ContentViewer.setInput(ContentViewer.java:239) at org.eclipse.jface.viewers.StructuredViewer.setInput(StructuredViewer.java:647) at org.eclipse.debug.internal.ui.VariablesView.setViewerInput(VariablesView.java:73) at org.eclipse.debug.internal.ui.VariablesView.selectionChanged(VariablesView.java:52) at org.eclipse.debug.internal.ui.DebugUIPlugin.selectionChanged(DebugUIPlugin.java:599) at org.eclipse.jface.viewers.Viewer.fireSelectionChanged(Viewer.java(Compiled Code)) at org.eclipse.jface.viewers.StructuredViewer.updateSelection(StructuredViewer.java:891) at org.eclipse.jface.viewers.StructuredViewer.setSelection(StructuredViewer.java:671) at org.eclipse.debug.internal.ui.LaunchesViewer.setSelection(LaunchesViewer.java:77) at org.eclipse.debug.internal.ui.DebugView.autoExpand(DebugView.java:469) at org.eclipse.debug.internal.ui.DebugUIPlugin$3.run(DebugUIPlugin.java:304) at org.eclipse.swt.widgets.RunnableLock.run(RunnableLock.java(Compiled Code)) at org.eclipse.swt.widgets.Synchronizer.runAsyncMessages(Synchronizer.java(Compiled Code)) at org.eclipse.swt.widgets.Display.runAsyncMessages(Display.java(Compiled Code)) at org.eclipse.swt.widgets.Display.readAndDispatch(Display.java(Compiled Code)) at org.eclipse.ui.internal.Workbench.runEventLoop(Workbench.java(Compiled Code)) at org.eclipse.ui.internal.Workbench.run(Workbench.java:602) at org.eclipse.core.internal.boot.InternalBootLoader.run(InternalBootLoader.java:820) at org.eclipse.core.boot.BootLoader.run(BootLoader.java:285) at java.lang.reflect.Method.invoke(Native Method) at org.eclipse.core.launcher.Main.basicRun(Main.java:66) DW (5/31/01 5:20:41 PM) Problem only occurrs on JDK. See 1GEM3UQ: ITPDUI:WIN2000 - JDK ClassLoader.loadClassInternal(String) throws NativeMethodException Low Priority. DW (6/2/01 10:59:37 AM) It appears that the thread is resuming after the step completes, and then the method entry breakpoint is hit. This means our strack frames are stale. The thread is resuming and not telling us (i.e. it appears the VM is resuming the thread). There is something magical about this #loadClassInternal() method. No action at this time. DW (6/7/01 4:46:05 PM) Make into a README. DW (6/15/01 10:48:45 AM) Added to README.

normal

RESOLVED

WORKSFORME

{'$oid': '52e9b48c54dc1c25ebdb20bd'}

JGS (6/4/01 11:58:37 AM) If I am trying to set a bp on last line of method, but miss and actually click in ruler on the closing {, the bp actually gets set on the next executable line, even though it is 3 lines down, whereas the line I wanted is only 1 line above. Shouldn't the guess at bp location be bidirectional?

normal

VERIFIED

FIXED

{'$oid': '52e9b48c54dc1c25ebdb20be'}

DebugPlugin when it attempts to see if there is a previous launch entry doesn't check to see if there are no elements before getting the element. My suggested fix (see below) is to check for a null element first. According to the documentation launches are allowed to have 0 elements. Existing code (in 105): for (index= 0; index < history.length; index++) { ILaunch historyLaunch= history[index]; if (historyLaunch != null && historyLaunch.getElement().equals(launch.getElement()) && launcher.equals(historyLaunch.getLauncher())) { found= true; break; } } Suggested change (>>): for (index= 0; index < history.length; index++) { ILaunch historyLaunch= history[index]; if (historyLaunch != null && >> historyLaunch.getElement() != null && historyLaunch.getElement().equals(launch.getElement()) && launcher.equals(historyLaunch.getLauncher())) { found= true; break; } } NOTES: DS (5/29/01 4:38:23 PM) From ILaunch in 113 /** * Returns the object that was launched. Cannot return <code>null</code>. * * @return the launched object */ Object getElement(); No action required. AB (5/30/2001 11:01:48 AM) Part of the problem is that the javadoc for the "Launch" class constructor is missing the "launchedElement" parameter. Everything seemed to work without the element parameter except for the history processing. Perhaps I hadn't tested enough to find other problems. The launch that I create doesn't know which project or resource it will be associated with until the debugging starts. That is why I need to run the launch with no associated element. But, I still have a problem with requiring an element to do a launch. What if I want to debug an application where I don't have the main class in a project? Eventually the debuggee will call classes that are in a project. Do I have to pass in an arbitrary project to get things going? DW (5/31/01 10:37:26 AM) AB will use non-null placeholder for now. We will revisit post-june.

minor

RESOLVED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20bf'}

Try debugging: public class Test { public static void main(String[] args) { System.out.println(String.class); } } Turn on static fields... I don't think I should see synthetic static fields like class$0. NOTES: DW (5/31/01 4:44:20 PM) Deferred.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20c0'}

DW (5/31/01 5:40:02 PM) Before putting in a safeguard to catch NativeMethodException when trying to get the location for a linenumber in ClassLoader.loadClassInternal(String), it threw a NativeMethodException. However, the method is not declared as native, and inspection revealed that it does not think it is native. Only happends in JDK. Related to 1GELEBF: ITPDUI:WIN2000 - Exception while stepping DW (6/1/01 9:37:38 AM) Low priority.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20c1'}

jkca (5/31/2001 8:22:55 PM) jre-sdk 116 This is an issue with the algorithm for determining which element to launch. To Duplicate: 1. Create Java Project 2. Create java file, Main.java, with a main method 3. Create java file, Other.java, without a main method 4. Select Other.java and hit run. 5. Click next in the launch wizard. 6. Nothing shows up in the list to run (but Main is right there!). Proposed algorithm: If the leaf of a resource tree cannot be launched, check it's parent node to see if it can be launched. If you only check the parent, and not up the entire tree, you can be fast and correct 98% of the time. Perhaps the correct way to change this is for the java launcher to do the parent check. This way you don't have to change the debug core right now, but the behaviour is much more user friendly. DW (6/1/01 7:29:12 AM) Yes. Suggest the fix is in the java launcher - i.e. if there are no targets found then go back up the hierarchy looking for a target - or jump right to the project to look for more targets. DW (6/1/01 9:02:16 AM) Deferred.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48c54dc1c25ebdb20c2'}

jkca (5/31/2001 10:06:41 PM) This is a nit. Inspecting anonymous types in a scrapbook gives a bizzare type name. For example, I inspected new Object() { int foo= 5; } and got as a name CodeSnippet_14$1 I suspect this is as good as we can do right now. It is unfortunate that it exposes the underlying infrastructure. DW (6/1/01 6:52:44 AM) Propose no action. PM (6/15/2001 11:50:36 PM) This is a debugger problem, it should not deploy the evaluation temporary file in a visible area. This file should probably go into the metadata. Moving to ITPDUI for documenting and plus later.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20c3'}

While debugging 1GEBP07: ITPDUI:WIN2000 - Debug view flicker when doing an inspect, I was using Display instead of Inspect. I found that it kept appending the result to the same line in the Display view, rather than giving each one its own line. NOTES: DW (6/1/01 12:11:49 PM) Moved to JUI. DS (10/10/2001 4:27:48 PM) Fixed in the 2.0 stream.

minor

VERIFIED

FIXED

{'$oid': '52e9b48c54dc1c25ebdb20c4'}

When debugging a program, it uses the bug icon for the program being debugged. This has a negative connotation for the user's program. Should indicate that it's a Java Application, not a Buggy Java Application. NOTES: DW (6/1/01 10:04:28 AM) Deferred.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20c5'}

jkca (6/1/2001 11:04:14 AM) jre-sdk 116 If you mess with the System class in the inspector, all following inspections (until you terminate the scrapbook) can be effected. This is not necessarily obvious to the end user. For example, if you inspect System.out and change autoFlush to false, all System.out.println calls following will appear to fail. Something to doc I guess. DW (6/15/01 10:06:27 AM) Added to README.

normal

RESOLVED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20c6'}

DS (6/1/01 3:35:32 PM) On 116 Test Case Have a program running and paused at a breakpoint Go to the display view With nothing selected, hit the Inpsect or Display actions An evaluation occurs on the empty String. NOTES: GDA (6/5/01 11:40:48 AM) Defer DS (10/10/2001 4:29:14 PM) This is still the current behaviour...but have proposal/prototype to select the entire current line of an empty selection. @see 1GJ6GR9: ITPDUI:ALL - Display window enhancement

major

VERIFIED

FIXED

{'$oid': '52e9b48c54dc1c25ebdb20c7'}

Debug perspective. In the processes view, a terminated process (top-level item in the tree) still has a running man instead of a standing man as its icon. Its children properly have the standing man as their icon. NOTES: EG (6/2/2001 7:51:02 AM) moving to DUI. Is this a linux only problem? jkca (6/2/2001 1:52:33 PM) likely not a linux problem as the motif 117 drop was missing the swt natives (read, does not launch) DW (6/2/01 2:24:00 PM) We do not change the icon of the top level item (launch) - only the debug target or process.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48c54dc1c25ebdb20c8'}

DW (6/2/01 2:00:35 PM) A Watch List to store expressions that are evaluated on each suspend. Issues: Binding expressions to locations in code. When to evaluate.

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48c54dc1c25ebdb20c9'}

DW (6/2/01 2:08:48 PM) If source lookup fails, can we somehow prompt for source? Is the best we can do to provide a resource navigator?

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48c54dc1c25ebdb20ca'}

JGS (6/4/01 5:33:40 PM) Rebooted, launched stand-alone Eclipse (118), then tried to debug simple program under J9. Blew up on OutOfMemoryErrors every time. JGS (6/4/01 5:42:01 PM) Actually, this only happens if Eclipse is launched with the -debug option. WIthout it, can debug on J9 fine.

normal

CLOSED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20cb'}

DS (6/4/2001 11:24:31 AM) In the debug menu, the relaunch last action is always enabled...this differs from all the other elements in the menu DW (8/7/01 9:34:10 AM) Fix in conjunction with 1GAHLUB: ITPDUI:WINNT - Feature: API Change: Persist debug/run history (However, not sure if possible to fix without more desktop support, as this is an action delegate). EG (9/24/2001 8:31:30 AM) this requires more workbench support. JUI needs this as well and it is on our wish list.

minor

CLOSED

WONTFIX

{'$oid': '52e9b48c54dc1c25ebdb20cc'}

DS (6/4/2001 3:43:48 PM) These actions are not friendly

normal

CLOSED

INVALID

{'$oid': '52e9b48c54dc1c25ebdb20cd'}

DS (6/4/2001 2:57:07 PM) Changing the font is immediate and retroactive Changing the colors is only on subsequent output.

normal

VERIFIED

FIXED

{'$oid': '52e9b48c54dc1c25ebdb20ce'}

Running Eclipse in Eclipse under debugger/J9, I get the dialog "Cannot connect to VM - time out" quite often. Is there something that can be done to improve this ? Why is it happening at all in fact ? NOTES: PM (6/5/2001 2:42:46 PM) I am guessing that when telling it to retry, the timeout could be lengthened a little, given the user explicitly chose to retry, so it is likely he wants to wait if necessary. DW (6/5/01 9:38:14 AM) This happens when it takes longer than the timeout preference setting to launch and connect to a VM. When this happens, the workaround is to increase the timeout on the preference page. You get to the preference fron the "edit" dialog from an installed VM. Will make this into a README. DW (6/15/01 10:51:29 AM) Added to README.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48c54dc1c25ebdb20cf'}

DS (6/4/2001 3:14:09 PM) In the top toolbar we display the actions to create Java elements. We do not list these same elements in the new menu pull down.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48c54dc1c25ebdb20d0'}

You have probably had this PR reported before, but just in case... When a breakpoint is hit, and the Debugger opens the editor, it should scroll the current line to the center of the screen. Also, when the stack frame is selected, the current line should be scrolled to the center. The current method of always showing the source editor scrolled so that the current line is at the top of the screen is very disconcerting, because it doesn't let the user see where they have come from. I find I always need to scroll a little bit after stepping - which makes stepping through something kind of slow... NOTES: CM (6/6/2001 1:17:40 PM) The code to do this is in the old VAME debugger. This is actually one of the more important - albeit non-critical - PRs, because it is annoying for every single line of code debugged in the debugger. We though about where to scroll to, and decided to go for the center, and everybody was much happier. Note though that you have to only scroll if the current line is not visible at all. If the user stepped and the new current line is visible, do not scroll.

normal

VERIFIED

FIXED

{'$oid': '52e9b48c54dc1c25ebdb20d1'}

If the debugger cannot start the VM properly, an indication of why the debugging VM will not start should be given. This was discovered on Solaris but I believe it might apply to all platforms. On the Solaris system, only the Java 1.3 JRE was installed, which does not include the shared libraries libdt_socket.so and libjdwp.so. These libraries are required when the VM is started with the switches: -Xdebug -Xnoagent -Djava.compiler=NONE -Xrunjdwp:transport=dt_socket,addesss=... In eclipse, if the debugger is started for an application, the process dialog for "Starting VM..." is displayed and after several seconds, the timeout dialog (Timeout while waiting for VM to connect. Retry?) appears but does not give any indication of what went wrong. I intercepted the Java command that eclipse used to start the VM and I tried it in an xterm (console) and the following error message was output: Can't load libjdwp.so Could not create the Java virtual machine. This error message should be caught by eclipse and displayed to the user, otherwise he has no indication of why eclipse could not connect to the debugging VM. NOTES:

normal

VERIFIED

FIXED

{'$oid': '52e9b48d54dc1c25ebdb20d2'}

1. Add a breakpoint for "java.lang.ArrayIndexOutOfBoundsException: caught and uncaught" 2. Run the code which follows 3. When the program stops at the illegal index access, select the Main.foo() method in the execution stack display 4. Go to the Display view and evaluate "aVector.size()" Rather than give the correct output (1), an exception is thrown: "java.lang.ClassNotFoundException" Test code: import java.util.*; public class Main { private Vector aVector; public static void main(String[] args) { new Main().foo(); } public Main() { aVector = new Vector(); } public void foo() { aVector.add(new Integer(1)); aVector.elementAt(-1); return; } } NOTES: DW (6/18/01 5:00:20 PM) I found that if the top stack frame is in a binary type, but the selected stack frame is in a source type, I get the same error. I tried Infinite Loop, and stepped into the System.out.println(). From the calling stack frame I did an evaluation of a constant and got a ClassNotFoundException. Not sure why this matters. DW (6/19/01 9:23:34 AM) Defer. OT (14/08/2001 4:10:38 PM) The problem seems to come from the stack frame used for the evaluation. I need to find out which stack frame is used. The top one seems to be a good guess and like this class is classes.zip the bootstrapclassloader is used instead of the classloader. The consequence of this that the code snippet classes are not found. OT (23/08/2001 2:29:52 PM) We should always use the stack frame in which the code is evaluated and not the top one. DW (8/28/01 10:20:10 AM) The stack frame comes from the UI selection in "EvaluateAction.#getContext()". This needs to be debugged. A quick look at the code looks like it should get the selected stack frame properly. DS (9/10/01 1:37:59 PM) The top stack frame is not being used...the stack frame is the one that the evaluation occurs in. DS (9/10/01 4:20:24 PM) Talking with JW, this is the known problem of not being able to perform evaluations in classes that have been loaded from the BootClassLoader as this loader will not find our snippet class.

minor

RESOLVED

DUPLICATE

{'$oid': '52e9b48d54dc1c25ebdb20d3'}

AK (6/5/01 2:38:29 PM) 1. debug a class with a String variable (let's say 's') 2. add s.value[2] (for instance ) to the inspector 3. you see sth like [2]=t no indication that [2] is a char NOTES: DW (6/5/01 12:03:22 PM) There is a usability problem. We do not show types in the RHS of expressions/variables. Top level expressions do not show types for primitive data types. NOTE: types appear if you "Display" rather than "Inspect".

enhancement

VERIFIED

FIXED

{'$oid': '52e9b48d54dc1c25ebdb20d4'}

- write a hello world example - launch it using J9 observe a dos console pops up and disappears when program has terminated. NOTES: DB (05.06.2001 15:49:16) In debug I get a dos console for j9.exe and j9proxy.exe EG (6/5/2001 4:02:52 AM) for JDK we can use javaw to launch a program without a dos box. There no analog for the J9. Not criticial, SDK does not ship with J9. Moving to DUI DW (6/5/01 9:39:14 AM) Should make this into a J9 request.

critical

RESOLVED

WONTFIX

{'$oid': '52e9b48d54dc1c25ebdb20d5'}

AK (6/5/01 4:02:47 PM) see title NOTES: EG (6/5/2001 4:59:19 AM) setting breakpoints on empty lines is not allowed and as a user I'd expect the same for comments. The JavaEditor knows about comment ranges (partitions) and this knowledge should be leveraged. Not critical, the user will see that the breakpoint didn't get installed from the icon.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48d54dc1c25ebdb20d6'}

DW (6/5/01 10:26:25 AM) Remote evaluation/inspect is not supported. DW (6/15/01 10:46:23 AM) Added to README.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48d54dc1c25ebdb20d7'}

AK (6/5/01 3:17:42 PM) 1. debug a program 2. modify the source in some way - like organize imports, format or sth 3. note: the debugger does not notice that and still shows you the previous line numbers. and selects wrong lines too. NOTES: EG (6/5/2001 3:20:29 AM) support for distinguishing between the source seen by the debugger and the source currently worked on in an editor didn't make it in for June. Moving to ITPDUI, should be mentioned in the Readme.

normal

RESOLVED

DUPLICATE

{'$oid': '52e9b48d54dc1c25ebdb20d8'}

When a breakpoint is hit, and the debugger is minimized, it doesn't come to front as VAJ's debugger does. NOTES:

normal

VERIFIED

FIXED