Re: POSIX Threads Compliance in LSB Test Suite
The motivation for The Open Group to make the VSTH-lite tests available
was so we can measure the difference between the current Linux
implementation and the standard. At the moment the LSB test team
is not proposing that the VSTH-lite tests be used as part of the
LSB certification tests, although they are available in the repository.
As with all the standards test suites that The Open Group has made
available as part of the LSB test program, the Artistic license requires
that any modifications leave the original test modes (for POSIX)
unchanged , and so either a Linux specific test be introduced and
a corresponding Linux test mode, or a better way is to keep a list of
expected and understood test failures. Its important to note that zero
failures in a test suite does not equate to conformance, there will always
be issues in test suites, standards, and implementations that require
a resolution process.
The Open Group
(LSB test leader)
On Jul 2, 11:54am in "POSIX Threads Compli", Howell, David P wrote:
> On Mon, 2 Jul 2001, Howell, David P wrote:
> > Would it make sense to just post this to lsb-discuss?
> Yes, I think it would. People need to see that it is important that the
> differences with POSIX continue to be eliminated. I suspect that there are
> that are satisfied with the current state of threads as fixing them is
> > > kernel clone() implementation. Knowing that there are problems, is LSB
> > > taking the approach that
> > > they are a) testing for POSIX threads compliance and then allowing
> > > implementations to get waivers
> > > (not sure that this is the process, it used to be back in the MP-RAS
> > > test days), or b) taking
> > > the tests and creating compliant versions that match the LinuxThreads
> > > implementation.
> > >
> > > We are hoping for the former, as this approach will keep the issues
> > > the community until
> > > there are better solutions found for true conformance with other Unixes
> > > that do POSIX threads right.
> > > For the industry that we are going after POSIX compliance is important,
> > > and we would want to see
> > > full compliance still be the goal for the Linux standard base.