harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gregory Shimansky <gshiman...@gmail.com>
Subject Re: [general][M3] Classlibrary test suite crashes (Was: Re: [general] M3 - code frozen)
Date Thu, 27 Sep 2007 15:10:33 GMT
Alexei Zakharov wrote:
> Stepan, Tim,
> 
> I'm currently working on
> org.apache.harmony.security.tests.java.security.cert.serialization.CertificateTest
> crash on Linux. During last few hours I've successfully removed all
> security stuff from it and created simpler reproducer:
> 
> Test113.java
> ---
> import java.io.*;
> 
> public class Test113 {
> 
>     static class A implements Serializable {
>         private final byte[] encoding = null;
> 
>         protected A() {}
> 
>         protected Object writeReplace() {
>             return new String(encoding.clone());
>         }
>     }
> 
>     public static void main(String argv[]) throws Exception {
>         ByteArrayOutputStream out = new ByteArrayOutputStream();
>         ObjectOutputStream oos = new ObjectOutputStream(out);
> 	A a = new A();
> 
>         for (int i = 0; i < 2; i++) {
>            try {
> 	       System.out.println("ITERATION " + i);
> 	       oos.writeObject(a);
> 	   } catch (NullPointerException e) {
> 	       System.out.println("NPE catched");
> 	   }
> 	}
>         System.out.println("PASS");
>     }
> }
> ---
> 
> It passes on J9 and RI but leads to segmentation fault on DRLVM at my
> Debian stable Linux IA32 machine. Tomorrow I'm going to unwrap it
> further till I see obvious bug in DRLVM or probably in LUNI natives.
> However, if someone else like to join - I open to any kind of help.

I tried this test and I am quite sure that the bug is in DRLVM. It 
doesn't exit signal context when it transfers control to the NPE 
exception handler, and there it receives a second SIGSEGV which isn't 
handled by signal handler since it is still in signal handler context. 
Please file a bug on DRLVM with [exception] category.

BTW the same bug should happen on Linux x86_64.

> I can't reproduce
> org.apache.harmony.security.tests.asn1.der.DerUTCTimeEDTest failure on
> Windows.
> 
> Besides crash of CertificateTest I also see
> tests.api.java.net.MulticastSocketTest failure on Linux. It can be a
> bogus test indeed. I haven't looked at it yet.
> 
> With Best Regards,
> Alexei
> 
> 2007/9/27, Tim Ellison <t.p.ellison@gmail.com>:
>> Stepan Mishura wrote:
>>> On 9/25/07, Alexei Zakharov <alexei.zakharov@gmail.com> wrote:
>>>>> As I remember Sean took pack200 tests. And Alexei Zakharov took
>>>>> security test crash.
>>>> I've just said I'm able to reproduce failures :) However, I'll take a
>>>> look. May be it can be fixed easily.
>>> Alexei, any progress in resolving classlib test crashes? Do we have
>>> serious issue here?
>> There are no serious issues with the pack200 tests which are now passing
>> since the code was cleaned and rebuilt.
>>
>> There is a reported VM crash running the classlib tests on r579330, but
>> no details available so maybe they timed out.
>>
>> Windows:
>> org.apache.harmony.security.tests.asn1.der.DerUTCTimeEDTest
>>
>> Linux:
>> org.apache.harmony.security.tests.java.security.cert.serialization.CertificateTest
>>
>> The other failure on Linux java.net code looks to me like a potentially
>> bogus test?
>>
>> Regards,
>> Tim
> 


-- 
Gregory


Mime
View raw message