Re: [gpsd-dev] Fw: [gpsd-commit-watch] [SCM] GPSD branch, master, updated. release-3.17-356-ga7dadab

[Gary E. Miller]

Yo Fred!

On Sat, 29 Sep 2018 15:41:48 -0700 (PDT)
Fred Wright <address@hidden> wrote:

> > Odd that I do not have that problem.  I tested on over a dozen hosts
> > of wildly different configurations.  If you do, lets find a real fix
> > that is not a regression.  
> 
> It definitely behaves that way on the Mac (though perhaps not if you
> have no X11 support installed).

Ouch.  The xgps and xgpsspeed don't run any of the Gtk code for -V and
-h on Linux.  They just do the import.  So unclear why your X11 was
triggered.  Mine is not.  On systems with X11 running, X11 installed
but not running, and X11 not installed.

Probably best not to refer to this as an X11 issue.  Gtk does not
require X11 to run.  My son just did an embedded RasPi project
that did not run X11.  The Gtk directly managed the frame buffer.

I also don't think Gtk on Windows needs X11, but I'm unsure about that.

> > Maybe a simple check for $DISPLAY and exit if unset?  That woulda
> > saved you code.  
> 
> Actually it's sort of the opposite.  My recollection was incorrect
> about the imports failing when DISPLAY is unset, and they don't
> bother to do the X11 launch in that case.  So the simple fix is to
> clear DISPLAY in test-only cases (which aren't going to try to do any
> X11 display, anyway).

Cool.  I see you pushed a fix.  I fixed the pycodestyle issues you
created in SConstruct.

> It might be nice if the case where xgps/xgpsspeed tries to actually
> use X11 without a proper DISPLAY setting had a friendlier error
> message than a segfault, but that isn't a new bug. :-)

I also found that in testing yesterday.  Seems to be a bug deep in the
Python GTK code.  Nothing we can do about that.

> >> I'm sure Bernd's Travis
> >> checks would have failed as soon as you introduced that if they
> >> hadn't stopped updating two months ago.  
> >
> > Dunno anything about that.  Do you know who can fix that?  
> 
> Bernd, presumably.  I think his gpsd-mirror repo stopped mirroring.

Ouch.

> >>> This new check found a ton of bugs.  
> >>
> >> Quite likely so, but none that my better-behaved check wouldn't
> >> have found.  
> >
> > Easy to say since you did not see my results.  But not true.  
> 
> Did you have any cases where cairo and gi imported successfully but
> there were still missing prerequisites?

Nope.  But other issues, like failed gps.__version__ checks.  -V not
implemented, etc.

The easiest way to see if a Python program is syntactically correct
is to run the -V check.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        address@hidden  Tel:+1 541 382 8588

            Veritas liberabit vos. -- Quid est veritas?
    "If you can’t measure it, you can’t improve it." - Lord Kelvin

pgpi2pF2NsIPE.pgp
Description: OpenPGP digital signature