[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Jikesrvm-regression] classpath CVS head regression FAILED 4 tests;
From: |
Casey Marshall |
Subject: |
Re: [Jikesrvm-regression] classpath CVS head regression FAILED 4 tests; 0 mauve failures |
Date: |
Mon, 08 Nov 2004 21:44:44 -0800 |
User-agent: |
Gnus/5.1002 (Gnus v5.10.2) Emacs/21.3 (gnu/linux) |
>>>>> "Mark" == Mark Wielaard <address@hidden> writes:
Mark> Hi,
Mark> On Mon, 2004-11-08 at 23:46, Steven Augart wrote:
>> > development: Failed in SPECjbb2000
>> Another new problem, this time during class loading. I'm CC'ing
>> this to the classpath list, since it probably points to a problem
>> in Classpath CVS head.
>>
>> at java.util.jar.JarFile.verify(JarFile.java:669)
>> at java.util.jar.JarFile.readSignatures(JarFile.java:561)
>> at java.util.jar.JarFile.getEntry(JarFile.java:406)
>> at java.util.jar.JarFile.getJarEntry(JarFile.java:475)
Mark> I was just hunting this one down.
Mark> Attached is my first attempt.
Unfortunately, this is a different bug :(
The code checks if the signature file identifier is DSA or RSA, but
unfortunately leaves the Signature object 'null' if it is
neither. Which is odd; Steven, is this Jar file signed? Any chance I
could get some info about it, if it is? Like what algorithm signed it?
I'll check in the attached patch that fixes this.
Mark> It makes sure to not check the verified Map when the jar file is
Mark> explicitly opened with verify set to false. It also just
Mark> compares the value in the Map to Boolean.TRUE or Boolean.FALSE.
Mark> I am still thinking about synchronization issues. It looks like
Mark> we need to be more careful about synchronization when
Mark> manipulating/reading the verified Map since a JarFile can
Mark> probably be accessed from multiple threads at once.
Eek, yeah, I didn't consider synchronization.
--
Casey Marshall || address@hidden
jarfile.patch
Description: Text Data