Tim Quinn wrote on 9/17/09 11:23 AM:
> Bill Shannon wrote:
>>> //Util.java
>>> InputStream iStream =
>>> Util.class.getClassLoader().getResourceAsStream("config"+File.separator
>>> + fileName);
>>>
>>> while (iStream.available() > 0) { <-- line 149
>>> oStream.write(iStream.read());
>>> }
>>>
>>> iStream is null.... but the files are there in
>>> v3\security\core\target\classes\config.
>>> Can someone from security team take a look at this? In the meantime can
>>> we disable the ejb.EmbeddedTest since it's breaking Windows build?
>>>
>>
>> No doubt someone is catching Exception and hiding this error? Grrr...
>>
>> I believe resources are always named using URL syntax (forward slashes)
>> instead of file syntax. The code above is buggy, which is why it fails
>> on Windows.
>>
>
> ...which is one reason I've lobbied in the past for a build + QL Hudson
> build to run on a Windows box, to try to catch these. Of course no set
> of tests will catch every possible such problem, and I know Terena and
> others have worked toward this in the past. Let's see if we can at
> least have a Hudson build + QL run on a Windows box triggered after an
> svn repository change in place soon.
Is there some reason that's hard to do? Why don't we just do it?
Do our tests depend on running on Unix?