This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
Reproduction: 1. Create a new file 2. Add a new method that returns i.e. String public String foo(){ return ""; } 3. Place cursor inside the method and invoke editor context menu 4. Invoke Tools/Correct JavaDoc and save the document immediately.
Created attachment 10032 [details] Deadlock full thread dump
Feel free to change the priority to P2 in accordance with http://qa.netbeans.org/processes/bug_priority_guidelines.html The reproductions steps are not very common
NB 3.5 beta, build 200304172350, jdk1.4.1 Now I have reproduced datalost using these repro steps. The document has not been saved complete, several lines from the end of the file was lost. Undo doesn't work.
From the thread dump it looks that two save actions were started on the document. Strange, but it is not cause of this issue. I've already seen this deadlock, but was never able to reproduce it. I'm sure there is a problem and it is there for some time (definitely >4 months). I'm not able to reproduce this test case as well. It works fine on my computer. I'm going to analyze source code. Lowering to P2.
It is strange that after "Correct javadoc" action (following by save action), the document is reloaded sometimes. Sometimes the confirmation dialog appears with the question "File x.y was modified externally. Reload it?" After pressing "Yes" the document is empty, all data are lost. The exception is thrown, I'll attach it. Now, I am not able to reproduce the deadlock such easy on the current build. Anyway, I think, that reproduction steps are very, very uncommon. The priority evaluation depends only on the threads analyzing. David, if you think, that problem cannot occur during normal circumstances, feel free to downgrade the priority.
Created attachment 10065 [details] BadLocationException
*** Issue 30309 has been marked as a duplicate of this issue. ***
I spent whole afternoon playing with this and could not reproduce it. Problem background: Document has Lines and Line has PositionRefs (reference to position in document) and PositionRef has Kind which can be PositionKind, OffsetKind, OutKind or LineKind. The Kind of the PositionRef *dynamically changes* according to document changes like document was opened in editor or document was closed. Somewhere there is synchronization problem causing that some positions are not correctly converted to the proper Kind (ie. OffsetKind was not converted properly). As a result when document is saved and Lines are updated the line with incorrect Kind starts opening of the document. Without having reproducible test case on which I could test a fix I do not dare to touch the code. QA, could you please try to reproduce it? So far this problem appeared quite rarely. If we do not find a testcase I propose to lower priority to P3 and TM=4.0. Hopefully will be killed by threading redesign of IDE.
It seems like it's unreproducible. Nobody can reproduce it (me, David, ... ) . I propose to not fix them for 3.5. David, please change milestone and if you think it's really hard to reproduce, decrease priority to P3. Thanks in advance.
Feel free to increase priority if it appears again. If reproducible case is found increase to P1!
*** Issue 32584 has been marked as a duplicate of this issue. ***
"future" TM is more appropriate. Without reproducible case I cannot promise to fix this for next version.
*** Issue 33556 has been marked as a duplicate of this issue. ***
Petr is responsible for the text package now.
*** Issue 37594 has been marked as a duplicate of this issue. ***
Maybe better reproducible as described in issue 37594 It may coincide somehow with background reload process.
*** Issue 39389 has been marked as a duplicate of this issue. ***
Should be prevented by fixing issue 40766
verified in [nb_dev](200406291800)