Java7CallGraphTest.testOcamlHelloHash() failing

[wala]

Java7CallGraphTest.testOcamlHelloHash() failing

[msridhar]

@cos if you have time to see if you get this failure, that would be great.

[juliandolby]

I have 1.7.0_71, and it seems to work. I'll try to get _72 and check...

-- Julian

From: Manu Sridharan [email protected]

To: wala/WALA [email protected]

Cc: Julian Dolby/Watson/IBM@IBMUS

Date: 11/11/2014 02:11 PM

Subject: [WALA] Java7CallGraphTest.testOcamlHelloHash() failing (#60)

This regression test fails for me when running on JDK 1.7.0_72. I wonder if
the JDK update caused the problem? @juliandolby can you reproduce?

—
Reply to this email directly or view it on GitHub.

[cos]

I have Java 8 on my system so it is a bit harder to test locally. Still, the #59 pull request has passed on JDK 1.7.0_72:
https://travis-ci.org/cos/wala/builds/40612103
so the problem doesn't seem to be strictly caused by _72.

The exception is: java.lang.VerifyError: Bad type on operand stack. This heavily upvoted stackoverflow answer says it is usually caused by a runtime library which is different from the compile-time one.

The exception is thrown here:

WALA/com.ibm.wala.core.tests/src/com/ibm/wala/core/tests/shrike/DynamicCallGraphTestBase.java

Line 99 in f4f6f06

Method testMain = testClass.getDeclaredMethod("main", String[].class);

It means the resources from the URLClassLoader are somehow corrupted. Is there any guarantee that all jars are recreated cleanly between runs? Maybe it worked for my Travis instance because it was fresh, with nothing created in previous runs. I have another branch (with some refactoring which should pass; a fixing #58), which now also fails on Travis with the same error in the same spot:
https://travis-ci.org/cos/wala/builds/40615655

There's also an open class loader there. I'll try closing it and see how it goes.

[msridhar]

Ugh. I would think that each run on Travis boots a fresh instance? I
think this must have something to do with a recent Travis change, as I
tried re-running build 190, corresponding to Julian's latest commit, and
that failed too:

https://travis-ci.org/wala/WALA/builds/40020629 (note the output is from
the successful run for some reason)

Anyway, I was unable to reproduce locally, and I really don't understand
how the build at https://travis-ci.org/cos/wala/builds/40612103 managed to
pass. Let me know if there's some way I can help narrow things down.

On Tue, Nov 11, 2014 at 4:38 PM, Cosmin Radoi [email protected]
wrote:

I have Java 8 on my system so it is a bit harder to test locally. Still,
the #59 #59 pull request has passed on
JDK 1.7.0_72:
https://travis-ci.org/cos/wala/builds/40612103
so the problem doesn't seem to be strictly caused by _72.

The exception is: java.lang.VerifyError: Bad type on operand stack. This
heavily upvoted stackoverflow answer
http://stackoverflow.com/questions/100107/reasons-of-getting-a-java-lang-verifyerror/2518002#2518002
says it is usually caused by a runtime library which is different from the
compile-time one.

The exception is thrown here:

WALA/com.ibm.wala.core.tests/src/com/ibm/wala/core/tests/shrike/DynamicCallGraphTestBase.java

Line 99 in f4f6f06

Method testMain = testClass.getDeclaredMethod("main", String[].class);

It means the resources from the URLClassLoader are somehow corrupted. Is
there any guarantee that all jars are recreated cleanly between runs? Maybe
it worked for my Travis instance because it was fresh, with nothing created
in previous runs. I have another branch (with some refactoring which should
pass; a fixing #58 #58), which now
also fails on Travis with the same error in the same spot:
https://travis-ci.org/cos/wala/builds/40615655

There's also an open class loader there. I'll try closing it and see how
it goes.

—
Reply to this email directly or view it on GitHub
#60 (comment).

[msridhar]

@juliandolby, did you fix this one? It looks like commit 71f16d4 is related

[juliandolby]

I don't know... hard to tell because this bug only happens sometimes
anyway.

From: Manu Sridharan [email protected]

To: wala/WALA [email protected]

Cc: Julian Dolby/Watson/IBM@IBMUS

Date: 11/11/2014 09:33 PM

Subject: Re: [WALA] Java7CallGraphTest.testOcamlHelloHash() failing (#60)

Julian, did you fix this one? It looks like commit 71f16d4 is related

—
Reply to this email directly or view it on GitHub.

[msridhar]

Closing as this seems to be working on Travis now.