At 10:05 18.06.2001 -0500, Jim Moore wrote:
>"Contributors are expected to be familiar with the log4j test environment.
>They are also required to add a new test case with each new feature or
>component."
>
>What is the test environment? In the org.apache.log4j.test package, there
>are "ad-hoc" tests (like PatternTest) and JUnit tests (like
>UnitTestCategory). Do we add unit tests to the "unitTest" script? Wouldn't
>it make more sense to add them to a TestSuite to use JUnit's built-in
>support? Also, shouldn't the ant declaration have a "test.all" task so it's
>not platform dependant (ie, "#!/bin/sh") and more easily done as part of the
>regular build process? Am I missing something fundimental, or do we simply
>need a more defined "test environment"?
>
>-Jim Moore
Hi Jim,
The test environment is the code and scripts that are under org.apache.log4j.test
and org.apache.log4j.xml.test. These contain shell scripts as well as JUnit unit tests
plus a few perl scripts.
I started writing the bash scripts long before JUnit was fashionable. The log4j test
suite has been verified to run on both Windows NT/2000 and Linux. The log4j test
scripts require the bash shell and perl. On most Unix systems these are included by
default. On Windows systems you need to install the Cygwin tools from
http://sources.redhat.com/cygwin
The JUnit-based tests are useful for testing small components whereas the bash scripts
do regression tests at a higher level, i.e. they compare the output of the tested code
against witness output. In some cases, the test code took more time to write than the
code being tested. I would not any shed tears if we could get rid of the bash/perl
scripts. However, I am not sure JUnit is suitable for regression testing. Is it? Your
comments and suggestions are welcome. Ceki
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
