Request for review: JDK-8011260: fatal error: LineNumberTable attribute has wrong length in class file

[David Holmes]

Hi Harold,

On 9/04/2013 11:55 PM, harold seigel wrote:
> Hi,
>
> Please review the following bug fix:
>
> Summary:   The assertion was prevented by changing the code to verify
> classes loaded by the anonymous class loader, unless -noverify was

So if I understand this right:

- previously: loader == NULL -> BytecodeVerificationLocal
- now: loader == NULL && anonymous-loader ->
BytecodeVerificationRemote

in other words we were skipping verification for anonymous classes.

I guess I don't fully understand the implications here. Given we 
normally skip verification for classes loaded by the boot-loader, does 
this mean we now verify anonymous classes "linked" to the boot-loader? 
Are all anonymous classes considered to be non-boot-loader classes?

Makes me wonder if there are other places where we check a loader for 
NULL but also need to check it is not related to anonymous classes ?

> specified.  Also, the relevant assertion messages were improved to
> include additional information, including class name.

Thanks for that. I hadn't realized the classfile_parse_error routines 
created and threw exceptions.

David
-----

> Open webrev at http://cr.openjdk.java.net/~hseigel/bug_8011260/
> <http://cr.openjdk.java.net/%7Ehseigel/bug_8011260/>
>
> Bug link at http://bugs.sun.com/view_bug.do?bug_id=8011260
>
> The change was tested by hand to ensure that the assertion no longer
> occurs when the test is run.  Additional testing was done using JCK Lang
> and VM, JTREG tests, ute vm.quick.testlist and vm.mlvm.testlist tests,
> and JPRT tests.
>
> Thank you!
> Harold