Bug 37152 - java.lang.VerifyError:
Summary: java.lang.VerifyError:
Status: RESOLVED FIXED
Alias: None
Product: AspectJ
Classification: Tools
Component: Compiler (show other bugs)
Version: unspecified   Edit
Hardware: Other Linux
: P1 blocker (vote)
Target Milestone: ---   Edit
Assignee: Jim Hugunin CLA
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2003-05-01 17:05 EDT by Arno CLA
Modified: 2003-05-04 16:58 EDT (History)
1 user (show)

See Also:


Attachments
soureces to reproduce the bug (2.49 KB, application/zip)
2003-05-01 17:19 EDT, Arno CLA
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Arno CLA 2003-05-01 17:05:12 EDT
whenever I want to run one of my testcases, the VM aborts with following    exception:    Failed to invoke suite():java.lang.VerifyError: (class:   org/schmidmeier/unittests/cache/TimeCacheTestsWorking, method: runTest   signature: ()V) Expecting to find integer on stack      Classes:
Comment 1 Arno CLA 2003-05-01 17:19:47 EDT
Created attachment 4770 [details]
soureces to reproduce the bug

AspectJ 1.1 rc 2
Junit as with eclipse 2.1
JDK, linux 1.4.1 rc 2
Comment 2 Jim Hugunin CLA 2003-05-01 17:51:14 EDT
Arno, is this a new bug since 1.1rc1, or is this code that only has worked in 
1.0.6?  Knowing this should help in identifying the problem.
Comment 3 Arno CLA 2003-05-01 18:35:35 EDT
Comment on attachment 4770 [details]
soureces to reproduce the bug

This bug was already in rc1.
I didn't report it then, because I expectat, that it was a duplicate of one of
the other verify bugs.

The same code runs fine, when compiled and executed with 1.0.6
Comment 4 Jim Hugunin CLA 2003-05-01 18:43:13 EDT
This is a variant of the VerifyError that Rich Price submitted, but it is 
sufficiently different that the fix for his bug doesn't also apply here.
Comment 5 Jim Hugunin CLA 2003-05-01 18:50:26 EDT
This bug has been isolated and a minimal test case constructed.  The issue is
with complicated around advice that returns while values are still on the 
stack (the code that makes that happen varies from compiler to compiler, but 
is certainly legal and must be handled).  This conflicts with the more 
agressive around-inling strategy in 1.1.
Comment 6 Jim Hugunin CLA 2003-05-04 16:58:45 EDT
this is fixed in the tree and includes a minimal test

Thanks for the small self-contained example that made the bug easy to 
reproduce.  However, if possible it is most valuable if you can submit a case 
that doesn't contain any proprietary code and that you can submit without a 
copyright notice (or with the CPL license).  This will enable us to directly 
add your test case to the test suite rather than needing to isolate the bug 
and build a new test case for the same bug.