[qmtest] os.system0 use in resource SetUp methods

[Mark Mitchell]

On Mon, 2003-11-24 at 07:56, Steve Byan wrote:
> On Nov 21, 2003, at 8:39 PM, Mark Mitchell wrote:
> 
> > On Fri, 2003-11-21 at 13:54, Steve Byan wrote:
> >> I've written a couple of test extension classes modeled on
> >> ShellCommandTest and ShellScriptTest which use a remote shell to
> >> execute the tests on a remote host specified by a "node" resource
> >> extension class, which selects one host from those provided in the
> >> context by a target extension class.
> >>
> >> I'd like to check the state of the selected host in the resource SetUp
> >> method. I've tried using code like this:
> >>
> >>          # test that pServer is up and running
> >>          if os.system("ping -c 3 " + remote_host + " >/dev/null") != 
> >> 0:
> >>              cause = "Can't ping " + blade + " at hostname " +
> >> remote_host
> >>              result.Fail(cause=cause)
> >>              return
> >
> > Does this problem happen with the CVS version of QMTest or only with
> > QMTest 2.0.3?  In 2.0.3, there were some bugs involving test/resource
> > dependencies that might have caused problems here.
> 
> I've only tried 2.0.3; I haven't set up access to the QMTest CVS 
> repository. I'll retest when 2.1 is out.

You could also try the 2.1 prerelease that was announced last night.

> > I can't think of any reason that the GUI should present particular
> > problems here; we'll have to have more information to figure out what's
> > going wrong.
> 
> I suspect the problem is somewhere in the platform thread support. The 
> problem turns out to be platform-specific, or at least python-version 
> specific. The GUI hangs on MacOS X, 10.3.1  which comes with python 
> 2.3, but not on RedHat 9.0, which comes with python 2.2.2. In previous 
> versions of Mac OS X, the POSIX thread support was reputed to be weak. 
> (Of course, so is Linux's :-)

Yes, that's a plausible hypothesis.

-- 
Mark Mitchell <mark at codesourcery.com>
CodeSourcery, LLC