Java assertion related test case failures
Hi,
I ran into some test case failures with the trunk code in both Eclipse and
Maven (reported by my continumm build) related to the usage of java
assertions.
For example, in test case
org.apache.tuscany.spi.extension.ReferenceTestCase, we have the following
test:
public void testPrepare() throws Exception {
TestReference ref = new TestReference(null, null, null);
try {
ref.prepare(); //[rfeng] We assume the assert will catch null before
it moves on to NPE
fail();
} catch (AssertionError e) {
//expected // [rfeng] NPE is thrown if assertion is not enabled
}
}
By default, assertions are disabled by JVM unless you explicitly turned it
on using -ea option on the VM (In Eclipse, you need to set the VM
arguments either at the JRE or test case.profile level. For Maven, you may
need to set MAVEN_OPTS to include -ea). As a result, the test case fails
because it throws NullPointerException instead of AssertionError.
Should we improve these test cases to be more robust?
Thanks,
Raymond
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Re: Java assertion related test case failures
For those who are interested, there's a nice article about java assertions @
http://www.javaworld.com/javaworld/jw-11-2001/jw-1109-assert.html. And the
following is quoted from it:
Expressions within an assert statement should not produce side effects,
since doing so exposes program execution to potentially different behavior
with and without assertions enabled. You should use assertions to produce
more reliable programs, not less reliable ones.
Finally, caution must guide the development of the expressions used in
assert statements. In addition to not producing side effects, assertions
should not alter normal program execution.
I think I buy what the author says.
Thanks,
Raymond
- Original Message -
From: Yang ZHONG [EMAIL PROTECTED]
To: tuscany-dev@ws.apache.org
Sent: Thursday, July 20, 2006 3:11 PM
Subject: Re: Java assertion related test case failures
I hope our code are not designated to run in some specifically configured
JVM.
Since JVMs may turn assertion on or off, I'm not sure AssertionError
should
be an expected behavior in general.
Dedicated exceptions and errors are much better protocol, e.g.
IndexOutOfBoundsException and OutOfMemoryError give much more
specific/useful info than plain AssertionError, not to mention assertion
isn't really born for error reporting.
On 7/20/06, Jeremy Boynes [EMAIL PROTECTED] wrote:
Assertions should be enabled when running our test cases.
I have added an AssertionTestCase to the spi module that will cause
the build to fail if assertions are not enabled.
--
Jeremy
On Jul 20, 2006, at 9:36 AM, Raymond Feng wrote:
Hi,
I ran into some test case failures with the trunk code in both
Eclipse and Maven (reported by my continumm build) related to the
usage of java assertions.
For example, in test case
org.apache.tuscany.spi.extension.ReferenceTestCase, we have the
following test:
public void testPrepare() throws Exception {
TestReference ref = new TestReference(null, null, null);
try {
ref.prepare(); //[rfeng] We assume the assert will catch
null before it moves on to NPE
fail();
} catch (AssertionError e) {
//expected // [rfeng] NPE is thrown if assertion is not enabled
}
}
By default, assertions are disabled by JVM unless you explicitly
turned it on using -ea option on the VM (In Eclipse, you need to
set the VM arguments either at the JRE or test case.profile level.
For Maven, you may need to set MAVEN_OPTS to include -ea). As a
result, the test case fails because it throws NullPointerException
instead of AssertionError.
Should we improve these test cases to be more robust?
Thanks,
Raymond
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Yang ZHONG
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Re: Java assertion related test case failures
We covered assertions a little in the exception handling document
which should be on the website but which I got from here:
http://svn.apache.org/viewvc/incubator/tuscany/site/src/site/xdoc/
exception_handling.html?view=co
To me, assertions are things that the developer declares should never
happen at runtime. As in, I thought about it and decided this could
never happen but I decided to check anyway as I could be wrong.
Enabling them when running tests is necessary to verify those
assumptions.
This is different from things that should not happen but which might
through user error - like checking for a user passing in a null value
to an API call and throwing a IllegalArgumentException rather than
letting a NPE happen when you use the value.
And different from things that might quite realistically go wrong and
which your user should be informed about - such as an IOException
when reading from a file.
--
Jeremy
On Jul 20, 2006, at 3:33 PM, Raymond Feng wrote:
For those who are interested, there's a nice article about java
assertions @ http://www.javaworld.com/javaworld/jw-11-2001/jw-1109-
assert.html. And the following is quoted from it:
Expressions within an assert statement should not produce side
effects, since doing so exposes program execution to potentially
different behavior with and without assertions enabled. You should
use assertions to produce more reliable programs, not less reliable
ones.
Finally, caution must guide the development of the expressions used
in assert statements. In addition to not producing side effects,
assertions should not alter normal program execution.
I think I buy what the author says.
Thanks,
Raymond
- Original Message - From: Yang ZHONG
[EMAIL PROTECTED]
To: tuscany-dev@ws.apache.org
Sent: Thursday, July 20, 2006 3:11 PM
Subject: Re: Java assertion related test case failures
I hope our code are not designated to run in some specifically
configured
JVM.
Since JVMs may turn assertion on or off, I'm not sure
AssertionError should
be an expected behavior in general.
Dedicated exceptions and errors are much better protocol, e.g.
IndexOutOfBoundsException and OutOfMemoryError give much more
specific/useful info than plain AssertionError, not to mention
assertion
isn't really born for error reporting.
On 7/20/06, Jeremy Boynes [EMAIL PROTECTED] wrote:
Assertions should be enabled when running our test cases.
I have added an AssertionTestCase to the spi module that will cause
the build to fail if assertions are not enabled.
--
Jeremy
On Jul 20, 2006, at 9:36 AM, Raymond Feng wrote:
Hi,
I ran into some test case failures with the trunk code in both
Eclipse and Maven (reported by my continumm build) related to the
usage of java assertions.
For example, in test case
org.apache.tuscany.spi.extension.ReferenceTestCase, we have the
following test:
public void testPrepare() throws Exception {
TestReference ref = new TestReference(null, null, null);
try {
ref.prepare(); //[rfeng] We assume the assert will catch
null before it moves on to NPE
fail();
} catch (AssertionError e) {
//expected // [rfeng] NPE is thrown if assertion is not
enabled
}
}
By default, assertions are disabled by JVM unless you explicitly
turned it on using -ea option on the VM (In Eclipse, you need to
set the VM arguments either at the JRE or test case.profile level.
For Maven, you may need to set MAVEN_OPTS to include -ea). As a
result, the test case fails because it throws NullPointerException
instead of AssertionError.
Should we improve these test cases to be more robust?
Thanks,
Raymond
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Yang ZHONG
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Re: Java assertion related test case failures
If the purpose of the test case is to verify the assert, I suggest that we
use ClassLoader.setClassAssertionStatus(targetClassName, true) to make
sure assert is on before the target class is initialized (for example, in
a static block or TestCase.setUp()) instead of requesting the whole JVM to
be launched with -ea. Making sense?
Thanks,
Raymond
- Original Message -
From: Jeremy Boynes [EMAIL PROTECTED]
To: tuscany-dev@ws.apache.org
Sent: Thursday, July 20, 2006 4:00 PM
Subject: Re: Java assertion related test case failures
We covered assertions a little in the exception handling document which
should be on the website but which I got from here:
http://svn.apache.org/viewvc/incubator/tuscany/site/src/site/xdoc/
exception_handling.html?view=co
To me, assertions are things that the developer declares should never
happen at runtime. As in, I thought about it and decided this could
never happen but I decided to check anyway as I could be wrong. Enabling
them when running tests is necessary to verify those assumptions.
This is different from things that should not happen but which might
through user error - like checking for a user passing in a null value to
an API call and throwing a IllegalArgumentException rather than letting a
NPE happen when you use the value.
And different from things that might quite realistically go wrong and
which your user should be informed about - such as an IOException when
reading from a file.
--
Jeremy
On Jul 20, 2006, at 3:33 PM, Raymond Feng wrote:
For those who are interested, there's a nice article about java
assertions @ http://www.javaworld.com/javaworld/jw-11-2001/jw-1109-
assert.html. And the following is quoted from it:
Expressions within an assert statement should not produce side effects,
since doing so exposes program execution to potentially different
behavior with and without assertions enabled. You should use assertions
to produce more reliable programs, not less reliable ones.
Finally, caution must guide the development of the expressions used in
assert statements. In addition to not producing side effects, assertions
should not alter normal program execution.
I think I buy what the author says.
Thanks,
Raymond
- Original Message - From: Yang ZHONG
[EMAIL PROTECTED]
To: tuscany-dev@ws.apache.org
Sent: Thursday, July 20, 2006 3:11 PM
Subject: Re: Java assertion related test case failures
I hope our code are not designated to run in some specifically
configured
JVM.
Since JVMs may turn assertion on or off, I'm not sure AssertionError
should
be an expected behavior in general.
Dedicated exceptions and errors are much better protocol, e.g.
IndexOutOfBoundsException and OutOfMemoryError give much more
specific/useful info than plain AssertionError, not to mention
assertion
isn't really born for error reporting.
On 7/20/06, Jeremy Boynes [EMAIL PROTECTED] wrote:
Assertions should be enabled when running our test cases.
I have added an AssertionTestCase to the spi module that will cause
the build to fail if assertions are not enabled.
--
Jeremy
On Jul 20, 2006, at 9:36 AM, Raymond Feng wrote:
Hi,
I ran into some test case failures with the trunk code in both
Eclipse and Maven (reported by my continumm build) related to the
usage of java assertions.
For example, in test case
org.apache.tuscany.spi.extension.ReferenceTestCase, we have the
following test:
public void testPrepare() throws Exception {
TestReference ref = new TestReference(null, null, null);
try {
ref.prepare(); //[rfeng] We assume the assert will catch
null before it moves on to NPE
fail();
} catch (AssertionError e) {
//expected // [rfeng] NPE is thrown if assertion is not
enabled
}
}
By default, assertions are disabled by JVM unless you explicitly
turned it on using -ea option on the VM (In Eclipse, you need to
set the VM arguments either at the JRE or test case.profile level.
For Maven, you may need to set MAVEN_OPTS to include -ea). As a
result, the test case fails because it throws NullPointerException
instead of AssertionError.
Should we improve these test cases to be more robust?
Thanks,
Raymond
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Yang ZHONG
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED
Re: Java assertion related test case failures
Agree with Jeremy on assertion usage, especially *assertion is not anything
user should be informed about*
which is exactly my point: assertion (failure) should *not* be user/TestCase
expectation.
On 7/20/06, Jeremy Boynes [EMAIL PROTECTED] wrote:
We covered assertions a little in the exception handling document
which should be on the website but which I got from here:
http://svn.apache.org/viewvc/incubator/tuscany/site/src/site/xdoc/
exception_handling.html?view=co
To me, assertions are things that the developer declares should never
happen at runtime. As in, I thought about it and decided this could
never happen but I decided to check anyway as I could be wrong.
Enabling them when running tests is necessary to verify those
assumptions.
This is different from things that should not happen but which might
through user error - like checking for a user passing in a null value
to an API call and throwing a IllegalArgumentException rather than
letting a NPE happen when you use the value.
And different from things that might quite realistically go wrong and
which your user should be informed about - such as an IOException
when reading from a file.
--
Jeremy
On Jul 20, 2006, at 3:33 PM, Raymond Feng wrote:
For those who are interested, there's a nice article about java
assertions @ http://www.javaworld.com/javaworld/jw-11-2001/jw-1109-
assert.html. And the following is quoted from it:
Expressions within an assert statement should not produce side
effects, since doing so exposes program execution to potentially
different behavior with and without assertions enabled. You should
use assertions to produce more reliable programs, not less reliable
ones.
Finally, caution must guide the development of the expressions used
in assert statements. In addition to not producing side effects,
assertions should not alter normal program execution.
I think I buy what the author says.
Thanks,
Raymond
- Original Message - From: Yang ZHONG
[EMAIL PROTECTED]
To: tuscany-dev@ws.apache.org
Sent: Thursday, July 20, 2006 3:11 PM
Subject: Re: Java assertion related test case failures
I hope our code are not designated to run in some specifically
configured
JVM.
Since JVMs may turn assertion on or off, I'm not sure
AssertionError should
be an expected behavior in general.
Dedicated exceptions and errors are much better protocol, e.g.
IndexOutOfBoundsException and OutOfMemoryError give much more
specific/useful info than plain AssertionError, not to mention
assertion
isn't really born for error reporting.
On 7/20/06, Jeremy Boynes [EMAIL PROTECTED] wrote:
Assertions should be enabled when running our test cases.
I have added an AssertionTestCase to the spi module that will cause
the build to fail if assertions are not enabled.
--
Jeremy
On Jul 20, 2006, at 9:36 AM, Raymond Feng wrote:
Hi,
I ran into some test case failures with the trunk code in both
Eclipse and Maven (reported by my continumm build) related to the
usage of java assertions.
For example, in test case
org.apache.tuscany.spi.extension.ReferenceTestCase, we have the
following test:
public void testPrepare() throws Exception {
TestReference ref = new TestReference(null, null, null);
try {
ref.prepare(); //[rfeng] We assume the assert will catch
null before it moves on to NPE
fail();
} catch (AssertionError e) {
//expected // [rfeng] NPE is thrown if assertion is not
enabled
}
}
By default, assertions are disabled by JVM unless you explicitly
turned it on using -ea option on the VM (In Eclipse, you need to
set the VM arguments either at the JRE or test case.profile level.
For Maven, you may need to set MAVEN_OPTS to include -ea). As a
result, the test case fails because it throws NullPointerException
instead of AssertionError.
Should we improve these test cases to be more robust?
Thanks,
Raymond
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Yang ZHONG
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Yang ZHONG
Re: Java assertion related test case failures
If you like my previous proposal, here's the patch I tried.
Thanks,
Raymond
- Original Message -
From: Raymond Feng [EMAIL PROTECTED]
To: tuscany-dev@ws.apache.org
Sent: Thursday, July 20, 2006 4:14 PM
Subject: Re: Java assertion related test case failures
If the purpose of the test case is to verify the assert, I suggest that
we use ClassLoader.setClassAssertionStatus(targetClassName, true) to
make sure assert is on before the target class is initialized (for
example, in a static block or TestCase.setUp()) instead of requesting the
whole JVM to be launched with -ea. Making sense?
Thanks,
Raymond
- Original Message -
From: Jeremy Boynes [EMAIL PROTECTED]
To: tuscany-dev@ws.apache.org
Sent: Thursday, July 20, 2006 4:00 PM
Subject: Re: Java assertion related test case failures
We covered assertions a little in the exception handling document which
should be on the website but which I got from here:
http://svn.apache.org/viewvc/incubator/tuscany/site/src/site/xdoc/
exception_handling.html?view=co
To me, assertions are things that the developer declares should never
happen at runtime. As in, I thought about it and decided this could
never happen but I decided to check anyway as I could be wrong.
Enabling them when running tests is necessary to verify those
assumptions.
This is different from things that should not happen but which might
through user error - like checking for a user passing in a null value to
an API call and throwing a IllegalArgumentException rather than letting
a NPE happen when you use the value.
And different from things that might quite realistically go wrong and
which your user should be informed about - such as an IOException when
reading from a file.
--
Jeremy
On Jul 20, 2006, at 3:33 PM, Raymond Feng wrote:
For those who are interested, there's a nice article about java
assertions @ http://www.javaworld.com/javaworld/jw-11-2001/jw-1109-
assert.html. And the following is quoted from it:
Expressions within an assert statement should not produce side
effects, since doing so exposes program execution to potentially
different behavior with and without assertions enabled. You should use
assertions to produce more reliable programs, not less reliable ones.
Finally, caution must guide the development of the expressions used in
assert statements. In addition to not producing side effects,
assertions should not alter normal program execution.
I think I buy what the author says.
Thanks,
Raymond
- Original Message - From: Yang ZHONG
[EMAIL PROTECTED]
To: tuscany-dev@ws.apache.org
Sent: Thursday, July 20, 2006 3:11 PM
Subject: Re: Java assertion related test case failures
I hope our code are not designated to run in some specifically
configured
JVM.
Since JVMs may turn assertion on or off, I'm not sure AssertionError
should
be an expected behavior in general.
Dedicated exceptions and errors are much better protocol, e.g.
IndexOutOfBoundsException and OutOfMemoryError give much more
specific/useful info than plain AssertionError, not to mention
assertion
isn't really born for error reporting.
On 7/20/06, Jeremy Boynes [EMAIL PROTECTED] wrote:
Assertions should be enabled when running our test cases.
I have added an AssertionTestCase to the spi module that will cause
the build to fail if assertions are not enabled.
--
Jeremy
On Jul 20, 2006, at 9:36 AM, Raymond Feng wrote:
Hi,
I ran into some test case failures with the trunk code in both
Eclipse and Maven (reported by my continumm build) related to the
usage of java assertions.
For example, in test case
org.apache.tuscany.spi.extension.ReferenceTestCase, we have the
following test:
public void testPrepare() throws Exception {
TestReference ref = new TestReference(null, null, null);
try {
ref.prepare(); //[rfeng] We assume the assert will catch
null before it moves on to NPE
fail();
} catch (AssertionError e) {
//expected // [rfeng] NPE is thrown if assertion is not
enabled
}
}
By default, assertions are disabled by JVM unless you explicitly
turned it on using -ea option on the VM (In Eclipse, you need to
set the VM arguments either at the JRE or test case.profile level.
For Maven, you may need to set MAVEN_OPTS to include -ea). As a
result, the test case fails because it throws NullPointerException
instead of AssertionError.
Should we improve these test cases to be more robust?
Thanks,
Raymond
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Yang ZHONG
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands
Re: Java assertion related test case failures
I'm not sure what happens to the continumm build which underneath runs maven. It seems that assertions are not enabled for the test cases by default and I have to set -ea in the mvn.bat so that the latest code can be built successfully. Thanks, Raymond - Original Message - From: Jeremy Boynes [EMAIL PROTECTED] To: tuscany-dev@ws.apache.org Sent: Thursday, July 20, 2006 4:37 PM Subject: Re: Java assertion related test case failures On Jul 20, 2006, at 4:14 PM, Raymond Feng wrote: If the purpose of the test case is to verify the assert, I suggest that we use ClassLoader.setClassAssertionStatus (targetClassName, true) to make sure assert is on before the target class is initialized (for example, in a static block or TestCase.setUp()) instead of requesting the whole JVM to be launched with -ea. Making sense? I was thinking that when running testcases we want do assertions on across the entire VM so that any problems in peripheral classes would cause the test to fail. As in, all tests should pass if assertions are enabled so let's test that assumption so that we can't mistakenly run the tests with assertions off. IMO it is not a class-by-class thing. In the maven build assertions are enabled in the surefire plugin so tests should always run with them on. -- Jeremy - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Java assertion related test case failures
These test cases were my bone-headed mistake...What they should be doing is testing that prepare() executes properly so I'll fix them once I get a couple of other checkins cleared out. That said, we really ought to run mvn was assertions on as we need to verify that assertion checking is done properly. For example, there have been times in the past when assertion checking was incorrectly done (i.e. threw errors for legitimate cases) and problems arose when someone ran with -ea. Jim On Jul 20, 2006, at 4:57 PM, Raymond Feng wrote: I'm not sure what happens to the continumm build which underneath runs maven. It seems that assertions are not enabled for the test cases by default and I have to set -ea in the mvn.bat so that the latest code can be built successfully. Thanks, Raymond - Original Message - From: Jeremy Boynes [EMAIL PROTECTED] To: tuscany-dev@ws.apache.org Sent: Thursday, July 20, 2006 4:37 PM Subject: Re: Java assertion related test case failures On Jul 20, 2006, at 4:14 PM, Raymond Feng wrote: If the purpose of the test case is to verify the assert, I suggest that we use ClassLoader.setClassAssertionStatus (targetClassName, true) to make sure assert is on before the target class is initialized (for example, in a static block or TestCase.setUp()) instead of requesting the whole JVM to be launched with -ea. Making sense? I was thinking that when running testcases we want do assertions on across the entire VM so that any problems in peripheral classes would cause the test to fail. As in, all tests should pass if assertions are enabled so let's test that assumption so that we can't mistakenly run the tests with assertions off. IMO it is not a class-by-class thing. In the maven build assertions are enabled in the surefire plugin so tests should always run with them on. -- Jeremy - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
