Re: ✘gpsinfo

[Fred Wright]

On Mon, 15 Mar 2021, Gary E. Miller wrote:

> On Mon, 15 Mar 2021 18:05:33 -0700 (PDT)
> Fred Wright <fw@fwright.net> wrote:
[...]
> > The only Python that matters is the one referenced in the shebang
> > lines. Aside from error checking, that means something like this:
> >
> >         $(head -1 $(which gpscat) | cut -b 3-) -V
>
> gpsdebuginfo alreay does this:
>
>    head -n 1 `which xgps`
>
> Your command assumes that the shebang is present and working.  And hides
> valuable debug info.  gpsdebuginfo is for when things are broken.

Which is why I said "aside from error checking".

> The current test also prints gpsd module version, which may differ
> from the python client version.  Which your does not do.

My test was intended solely for the Python version test, and nothing else.

> And, just today on IRC, a user wanted both python and python3, so best to
> keep the script inclusive of options.

When invoked normally, the programs will only use the one python version 
they're shebanged for.  Reporting the version of a Python they're *not* 
using is rather pointless.

If you're really trying to report on all possible Pythons, regardless of 
whether gpsd's programs actually use them, you won't necessarily get that 
even if you use 'python', 'python2', and 'python3'.  I have a program 
which attempts to report some information about all Pythons installed in 
the system, but it would be gross overkill for this use case.  E.g.:

        MacPro:Python fw$ ./python_paths.py | tail -1
        Found 18 unique Python installations out of 25 total

> > Also note that running "gpscat -V" is sufficient to check that
> > imports work, including the packetizer.
>
> Assuming gpscat is not broken.  People run the script when things are
> broken.

If gpscat is present and able to do -V, then the imports work.  If it's 
present and not able to do -V, that almost certainly means that the 
imports don't work.  Hence, it's a straightforward import test.

> > Anything related to Python needs to be optional, in case the install
> > used "python=no".
>
> No way to know if python=no.  Worse, this will usually be run when
> gpsd installed from packages.  It fails nicely, so I'm not worried about
> running when not needed.  The problem is not being run when it needs to.

It can't directly know "python=no", but if everything else works and 
Python-based programs are missing, then that's a pretty good bet, and 
probably worth an appropriate message.

Fred Wright