Bit more details on windows. We do have builds going internally on
windows , but not the latest trunk which we will be doing soon. 
Regarding auth-test yes thats an issue and storm security in general not
going to work in windows. We disable this test for windows. I think we
should put a check in in the code itself to run this test on *nix only.

On Thu, Nov 5, 2015, at 09:06 AM, Bobby Evans wrote:
> Chuck, sorry but windows really does end up being a bit of an after
> though here.  We do have groups that test the official releases on
> windows and go back to fix issues that they find, but it is not
> guaranteed that you will be able to run on windows all the time.  I do
> know that symlinks are supported on windows, but it requires some special
> permissions to make it work.  Hadoop uses symlinks in a very similar way
> to how storm does now, so it might just end up being a documentation
> issue.  Please file a JIRA indicating the issues you are seeing and
> hopefully someone who works with windows will be better able to help you
> out.
> 
> 
> The auth-test failing to impersonate a user might be related to a bug we
> have had in the tests where some of the auth tests failed to stop
> impersonating a user and this leaked into other tests.  Those should have
> been fixed on master, but if you are still seeing them on master please
> file a JIRA.
> 
>  - Bobby 
> 
> 
> 
> On Thursday, November 5, 2015 10:20 AM, Chuck Burgess
> <[email protected]> wrote:
> Hey everyone,
> Trying to build Storm on Windows, I consistently see clojure test
> failures regarding the supervisor-test (seems to fail trying to create
> symlinks) and the auth-test (seems to fail trying to "impersonate user").
>  At first glance, I could see these as not being possible in Windows, but
> that's just a hunch.
> 
> Should these pass on a Windows build?  Or is a Windows build not
> typically considered "in scope" for the project?  I need to understand if
> Windows testing is significant or not, so I'll know how to proceed with
> trying to write up some contributions.
> 
> Thanks.
> CRB 

Reply via email to