Re: [gpsd-dev] Major progress!

[Eric S. Raymond]

Hal Murray <address@hidden>:
> I'm assuming that both gpsd and the test setup think they own a
> chunk of SHM.  Yet they are both using the same Key so they are both
> using/sharing the same block of memory.

Not any more.  I fixed that about an hour ago.

>  I don't know my way around the test package.  I'd expect there
> would be a test flag that would get passed in when running tests.
> It would use separate Keys for SHM and listen on separate TCP/UDP
> ports.  Or something like that.

Yes.  SConstruct now sets an environment variable that overrides the
default SHM key.  The separate TCP port is allocated using the port-zero
trick you dug up.

> Does it make sense to run several instances of the test setup in
> parallel?  If so, the flag needs to be more complicated that just a
> boolean.

Yes, I thought about that.  In theory the environment variable could
be set to a base value plus the SConstruct run's PID. But running multiple
tests concurrently doesn't make a lot of sense, so I haven't actually
implemented this yet.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>