Community
Participate
Working Groups
We are using AspectJ on a large java project (ca 6000 classes) and have been running ajc 1.1 steadily for nearly a year - with no serious problems. When testing out the new 1.2 rc1 candidate, the following NPE ocurred: ABORT Exception thrown from AspectJ 1.2rc1 This might be logged as a bug already -- find current bugs at http://bugs.eclipse.org/bugs/buglist.cgi?product=AspectJ&component=Compiler Bugs for exceptions thrown have titles File:line from the top stack, e.g., "SomeFile.java:243" If you don't find the exception below in a bug, please add a new bug at http://bugs.eclipse.org/bugs/enter_bug.cgi?product=AspectJ To make the bug a priority, please include a test program that can reproduce this exception. null java.lang.NullPointerException at org.eclipse.jdt.internal.compiler.CompilationResult.computePriority (CompilationResult.java:110) at org.eclipse.jdt.internal.compiler.CompilationResult.quickPrioritize (CompilationResult.java:336) at org.eclipse.jdt.internal.compiler.CompilationResult.getProblems (CompilationResult.java:230) at org.eclipse.jdt.internal.compiler.CompilationResult.getAllProblems (CompilationResult.java:118) at org.aspectj.ajdt.internal.core.builder.AjBuildManager$4.acceptResult (AjBuildManager.java:661) at org.aspectj.ajdt.internal.compiler.AjCompilerAdapter.acceptResult (AjCompilerAdapter.java:178) at org.aspectj.ajdt.internal.compiler.WeaverAdapter.finishedWith (WeaverAdapter.java:203) at org.aspectj.ajdt.internal.compiler.WeaverAdapter.weaveCompleted (WeaverAdapter.java:167) at org.aspectj.weaver.bcel.BcelWeaver.weave(BcelWeaver.java:570) at org.aspectj.ajdt.internal.compiler.AjCompilerAdapter.weave (AjCompilerAdapter.java:239) at org.aspectj.ajdt.internal.compiler.AjCompilerAdapter.afterCompiling (AjCompilerAdapter.java:114) at org.eclipse.jdt.internal.compiler.Compiler.compile(Compiler.java:376) at org.aspectj.ajdt.internal.core.builder.AjBuildManager.performCompilation (AjBuildManager.java:600) at org.aspectj.ajdt.internal.core.builder.AjBuildManager.doBuild (AjBuildManager.java:160) at org.aspectj.ajdt.internal.core.builder.AjBuildManager.batchBuild (AjBuildManager.java:94) at org.aspectj.ajdt.ajc.AjdtCommand.doCommand(AjdtCommand.java:102) at org.aspectj.ajdt.ajc.AjdtCommand.runCommand(AjdtCommand.java:53) at org.aspectj.tools.ajc.Main.run(Main.java:280) at org.aspectj.tools.ajc.Main.runMain(Main.java:217) at org.aspectj.tools.ajc.Main.main(Main.java:79) 1 fail|abort
This is likely to be related to bug 57432. If it is what I think it is then the problem is that you are having >100 errors/warnings (possibly declare errors/warnings) for some input files - is that possible? (And is this a binary weave?)
Correct. There are probably > 100 warnings generated. And it is a binary weave.
phew... so I do know what I'm doing :) We go into the code that NPEs if there are >100 errors - the code is intended to throw away all but the most important 100 errors. There are two options for 1.2 final: - Disturb the codebase quite a lot adding support for binary source reference contexts. This will make the code that determines the top 100 errors work successfully. - Increase the number from 100 to something like 100000 - that will mean it is highly unlikely we'll ever trigger the prioritization logic. We'll also leave the bug open and fix it properly in the next release.
I like your idea of increasing the max errors given the fact that this is coming after 1.2rc1. You should increase it to Integer.MAX or something similar to make sure it really won't show up in practice.
Ok, I've made the change, instead of the limit being 'compiler.options.maxProblemsPerunit', it is 'Integer.MAX_VALUE'. The line I fixed is in AjCompilerAdapter.getBinarySourcesFrom(). Its marked XXX and with this bug number. All the tests have passed. I could respin a whole build and upload it somewhere for Per to try but as its such a localized change and I'm 99.99999999% confident it is the fix for this bug, I'm about to attach a zip file to this bug report that contains the fix (a single .class file) - can you try this for me Per? Easiest way is to put patch.zip ahead of aspectjtools.zip in the ajc.bat file and then attempting your binary weave. Let me know if it works for you and we'll decide whether to cut a new RC containing the fix.
Created attachment 9681 [details] Contains one class file that 'fixes' this bug.
I have created a new testcase for this. A program that is subject to a binary weave and has 101 declare errors attached to it. In AspectJ without the fix, the testcase fails with an NPE. I have also checked in the fix to up the limit to Integer.MAX_VALUE. We should now change this bug report to an enhancement. The future enhancement would be to correctly support binary reference contexts for 'binary source' files (then the prioritization logic would not NPE). I've marked the place in the code where Integer.MAX_VALUE is used - search for this bug number in the source.
*** This bug has been marked as a duplicate of 57432 ***
