Re: [gpsd-dev] regression tests blow up on os x 10.9 with shm issues

[Eric S. Raymond]

Hal Murray <address@hidden>:
> 
> address@hidden said:
> > I think something went wrong with my attempt to set a non-default shm
> > segment for testing.  Possibly I happened to pick a key that is in use by
> > a service on your system.
> 
> The gpsctl -R line doesn't have the GPSD_SHM_KEY set.

Well, hell.

What a difference one letter can make.  Turns out I poked that into 'envs',
the scons build variable dictionary, when I wanted 'env', which means
what you think it does.

Fix pushed.  That ought to improve matters a lot.

>  My testing got more 
> interesting after I changed that line to:
> 
>             'echo "Foo $GPSD_SHM_KEY"; GPSD_SHM_KEY=0x47505345 $SRCDIR/gpsctl 
> -R 2>/dev/null; $SRCDIR/gpsfake -T; $SRCDIR/regress-driver $REGRESSOPTS 
> test/daemon/*.log')
> 
> >From scons check:
> 
> Matrix-algebra regression test succeeded
> echo "Foo "; GPSD_SHM_KEY=0x47505345 ./gpsctl -R 2>/dev/null; ./gpsfake -T; 
> ./regress-driver test/daemon/*.log
> Foo 
> sys linux2 platform Linux-3.18.5-201.fc21.i686-i686-with-fedora-21-Twenty_One:
>  WRITE_PAD = 0.00000

Note something about how you modified the command.  GPSD_HOME_KEY
will be in the environment of gpsctl -R but not the regress-driver call.
That might explain a lot.

Try cleaning up the dangling segments, then running scons check with my
latest patch in.  That will put the non-default GPSD_SHM_KEY in 
regress-driver's environment, too, which should help.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>