Re: [gpsd-dev] Major progress!

[Eric S. Raymond]

Gary E. Miller <address@hidden>:
> Which should not be a problem.  Having the production gpsd use the
> same shmid as the test gpsd causes many problems.  Not the least of
> which is test adta leaking into the production data.
> > 
> > > As long as the test shmid is cleaned up after the tests are done.
> > 
> > Hm.  Should I move the segment removal to the end of the test
> > sequence? Right now it's done first.
> 
> Maybe both?  And double check you have not the same shmid as the prodcution
> gpsd.

What I've done is implemented and documented a GPSD_SHM_KEY
environment variable that overrides the default GPSD_KEY value
everywhere it's used - three places, the daemon's shm export code, the
library's code for the other end, and gpsctl's -R option to remove the
segment.

SConstruct sets this variable to a non-default value as part of the
execution environment for the tests.  That should solve the problem.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>