Ito Kazumitsu wrote:
From: Ito Kazumitsu <[EMAIL PROTECTED]>
Date: Thu, 22 Dec 2005 11:08:15 +0900 (JST)
The runtime access control is much simpler than the compile-time
access control. And we may be able to omit the checking of nested
relations of classes at runtime.
Simply omitting the checking of nested relations of classes
causes IllegalAccessError, which was reported in November, 2003.
I am afraid Sun's VM spec:
A class or interface C is accessible to a class or interface D if and only if
either of the following conditions are true:
(Runtime):
- C is public.
- C and D are members of the same runtime package.
is incomplete and something should be said about the case
where C is protected.
Attached below is my proposed patch. I have deleted the
checking of outer classes for determining slot_acc,
but kept such checking for determining class_acc.
With this patch applied,
- All the regression tests pass.
- Mauve test gnu.testlet.java.lang.Class.newInstance shows the
same results as Sun's JDK.
- The testcase attached to
http://www.kaffe.org/pipermail/kaffe/2003-November/096187.html
passes.
Hi Ito,
Thanks for spending some more time in access.c. My previous patch was
minimalist. But if this one passes everything it's great and please
apply it to CVS ! I think we should add a regression test for this
feature also. I will do it before release if you do not.
Regards,
Guilhem.
_______________________________________________
kaffe mailing list
[email protected]
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe
