harmony-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alexey Varlamov (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HARMONY-3575) [drlvm][eclipse] One class in Pydev fails classloading
Date Sat, 14 Apr 2007 13:58:15 GMT

    [ https://issues.apache.org/jira/browse/HARMONY-3575?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12488867
] 

Alexey Varlamov commented on HARMONY-3575:
------------------------------------------

Vera, as you did not report any progress, I tried to investigate these puzzles myself. 
Using CAFEBABE classfile browser [1], I see DRLVM correctly reports about real defects. So
why RI accepts these files but rejects the synthetic test classes? The answer appeared in
different classfile versions:
PyObject - 45.3
PydevPlugin - 48.0
locVar - 49.0

Indeed, after changing version of locVar to 48.0, RI accepted it fine! I looked at JVMS v1
(which documents v45.3) and it does not require any checks for LocalVariableTable content.
Also, passing -noverify makes RI ignoring duplicate vars.
So the fix should be in making these validations conditional. 

BTW, I noticed DRLVM gives misguiding message in case of truncated classfile, like " could
not parse bla-bla-bla", while RI plainly says "Truncated file". Really, this is widely used
pattern in Class_File_Loader.cpp, like 
if(!cfs.parse(&value)) {
  REPORT_FAILED_CLASS("could not parse bla-bla value");
  return false;
}
for each and every entity. Quite verbose and not really useful, we need to refactor this.

[1] http://switch.dl.sourceforge.net/sourceforge/cafebabe/cafebabe-main-1.4-bin.zip


> [drlvm][eclipse] One class in Pydev fails classloading
> ------------------------------------------------------
>
>                 Key: HARMONY-3575
>                 URL: https://issues.apache.org/jira/browse/HARMONY-3575
>             Project: Harmony
>          Issue Type: Bug
>          Components: App-Oriented Bug Reports, DRLVM
>         Environment: linux x86_64
>            Reporter: Santiago Gala
>         Assigned To: Alexey Varlamov
>         Attachments: clf.patch, local_var_table.ZIP, local_variable_table.patch, lvt_20070406.patch,
PydevPlugin.class, pydevplugin.txt, PyObject.class
>
>
> I was getting a problem with PyDev in eclipse, until I found the error in the log: 
> Root exception: 
> java.lang.ClassFormatError: org/python/pydev/plugin/PydevPlugin : Element: 0x413b1bb0
from LocalVariableTypeTable doesn't coincide with element from LocalVariableTable for method
createEditorInput(Lorg/eclipse/core/runtime/IPath;Z)Lorg/eclipse/ui/IEditorInput; 
> at java.lang.ClassLoader.defineClass0(ClassLoader.java) 
> at java.lang.ClassLoader.defineClass(ClassLoader.java:417) 
> at org.eclipse.osgi.internal.baseadaptor.DefaultClassLoader.defineClass(DefaultClassLoader.java:161)

> Removing the check under 
> // See specification 4.8.12 second paragraph. 
> in Class_File_Loader.cpp makes it load, though I'm not sure it is the proper solution.

> I'll attach the output of javap -v

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message