Community
Participate
Working Groups
I am running Eclipse 3.1 RC but but noticed this behaviour in previous versions too. From time to time, the ANT view is no longer showing my single build.xml Script. Unfortunately I am not able to post a scenario how this happens. It is always that I am working some time on my different projects then shutting down Eclipse and the next day opening Eclipse and the Ant view, the script is no longer added. I am not closing any project. So is there another reason why an Ant Script could be removed from the Ant view? Ben
Ant buildfiles should only be removed when the user removes the enclosing project from Eclipse (closing the project included) or explictly removes the buildfile from the view. Any information in the .log file that could pertain?
So the ".log" file is showing when going to "About Eclipse Platform" > "Configuration Details" > "View Error Log"? Or is there another place. Ben
I use Window>Show View>Other>Pde Runtime>Error Log
Allright, I will have a look as soon as the problem re-occurs. Ben
Please reopen when more information is available. Thanks.
FYI I can reproduce this issue every day, both at home and from work with RC3. The projects are completly different. I am not seeing any Error in the Logs so far. Will see if I can nail it down to some reproducable steps. Ben
A reproduceable test case would be greatly appreciated.
Version: 3.1.0 Build id: I20050617-1618 Steps to reproduce: 1.) Start Eclipse 2.) Open ANT and have it as Fast View 3.) Add a build.xml file to the View 4.) Exit Eclipse 5.) Start Eclipse 6.) Directly exit Eclipse, without opening the ANT view 7.) Start Eclipse again Actual Results: The build.xml is no longer added to the view Ben
Reproducable here at work using same Eclipse version, but different Projects / ANT Scripts. Ben
Also reproducable using a clean Workspace. No difference if "Toggle Auto Reconcile" is activated or not. Ben
Thanks Ben
*** Bug 100210 has been marked as a duplicate of this bug. ***
Deferred to be fixed post 3.1
Oh, sad :| Ben
Reopening for post 3.1 work
*** Bug 103137 has been marked as a duplicate of this bug. ***
This appears to be a general problem with views that are not fully realized in an Eclipse session. I can reproduce the problem with the debug Display view as well. The view is never initalized nor is the state saved on shutdown via the memento. Therefore on the next Eclipse session when the view is used by the user, the memento for the view is null and the old session state cannot be restored. Will need to check with the UI guys to see if this is working by design and if so what, if any means are provided for this kind of support.
This is the same reason the Breakpoints view will lose the users settings of breakpoint organizers if the view is not activated by the user in an Eclipse session.
Moving to UI for comment.
I see. Not sure, but could'nt this problem be fixed using IDialogSettings for storing settings? They should exist all the time, even if the View was not opened. The PackageExplorerPart is using IDialogSettings to store settings. Ben
Paul, can you please investigate for 3.1.1? This must have been a regression late in 3.1. We previously held onto the memento for lazy views, and re-saved them if the view had not been materialized during the session.
Just chatted with DarinS, who agrees this should be a 3.1.1 candidate.
This has also a side-effect on the JUnit view when set as a fast view. It loses its orientation settings if not used in a session. On the next start, the view is initialized using the default. I always use the vertical orientation and I need to set it myself again if I don't run JUnit tests in a session.
*** Bug 106667 has been marked as a duplicate of this bug. ***
Created attachment 26208 [details] Patch unopened view state v01 The view's last session state was loaded into the ViewFactory, and then moved into the ViewReference as it was created. To save the state, if the ViewPart was instantiated it is used. If it wasn't instantiated, the ViewFactory needed to put the last session state (as saved in the ViewReference) back into the IMemento. PW
Kim, could you submit to 3.1.1 and HEAD (with Tod's help, if you need it). Thanx PW
Do you think we could get a session test that could confirm this behavior?
I am not sure what you are asking for...a test case? See comment #8.
(In reply to comment #28) > I am not sure what you are asking for...a test case? See comment #8. I think that's for me ... I'm going to try and put together a session test case tomorrow. PW
Awesome. I think our WorkbenchSessionTest class will need to be changed in order to run a multistage test. I have some ideas on how we can do this if you want to drag me in on it.
Created attachment 26225 [details] Session tests for the view state Added some session tests: session 1. open my view, set it's state, and then open a second view. session 2. check that my view hasn't been activated. session 3. activate my view and check the state. PW
Patches applied to 3.2. 3.1.1 forthcoming...
Submitted into 3.1.1.
*** Bug 108462 has been marked as a duplicate of this bug. ***
*** Bug 109122 has been marked as a duplicate of this bug. ***
verified in 3.2
