ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bruce Atherton <>
Subject Can't submit bugfix because of JDK 1.1 issues
Date Fri, 28 Jun 2002 19:33:54 GMT
I fixed a bug in selectors some time ago (see for details) 
but haven't been able to submit it because I have been struggling with 
getting the tests to work in all environments. In particular, because 
Magesh had troubles with tests on a 1.1 JVM, I've downloaded and am testing 
against Sun's JDK 1.1.8 for Windows.

My first problem was that Ant would not build with JDK 1.1.8 because there 
were classes that referred to the Collection classes. Here is a patch to 
get it working again (also included as an attachment):

Index: build.xml
RCS file: /home/cvspublic/jakarta-ant/build.xml,v
retrieving revision 1.304.2.19
diff -u -r1.304.2.19 build.xml
--- build.xml	25 Jun 2002 11:16:03 -0000	1.304.2.19
+++ build.xml	28 Jun 2002 18:05:30 -0000
@@ -135,6 +135,10 @@
               unless="jdk1.2+" />
      <exclude name="${optional.package}/ejb/IPlanet*.java"
               unless="jdk1.2+" />
+    <exclude name="${optional.package}/ejb/Borland*.java"
+             unless="jdk1.2+" />
+    <exclude name="${optional.package}/j2ee/*.java"
+             unless="jdk1.2+" />
      <!-- uses the context classloader -->
      <exclude name="${optional.package}/junit/"
               unless="jdk1.2+" />

Getting it compiling only takes me so far, though, at least under Windows 
2000 whether I am running cygwin, 4NT, or CMD. It appears that any tests 
which use BuildFileTest are failing because the base directory used when 
executing the buildfile tasks is the top level jakarta-ant directory rather 
than the directory where the buildfile lives.

I've done some investigation into this but I can't really see where it is 
going wrong. I suspect it has something to do with paths that use mixed 
separators (ie. c:\jakarta-ant\src/etc/testcases/taskdefs) but I'm not 
sure. Perhaps someone who knows this code better could figure it out.

There are other errors, too, although they may all be related to the 
problem above.

   ImmutableTest: expected <override> but was <original>
   IncludeTest: FileNotFoundException "\"
   IntrospectionHelperTest: expected <:\2> but was <2>
   AvailableTest: expected <true> but was <null>
   ConditionTest: expected <null> but was <true>
   CopyTest: Cannot perform operation between directory and file

My conclusion from all this is that we aren't all that worried about 
whether tests run everywhere. Is that correct? Should I submit my bugfix 
with the proviso that it "works for me"?

View raw message