messing with the gas / tank handling

[Dirk Hohndel]

On Tue, Oct 28, 2014 at 02:58:50PM -0700, Linus Torvalds wrote:
> On Tue, Oct 28, 2014 at 2:13 PM, Dirk Hohndel <dirk at hohndel.org> wrote:
> >
> > Bug #742 shows that we got things pretty badly wrong for the Cobalt if the
> > primary gas used on a dive (or in the case of his sample data, the only
> > gas used on a dive) was not tank 0. We showed that ugly gas change, we
> > showed that two gases were used during the dive, we associated the
> > pressure from the samples with the wrong tank, etc.
> >
> > I just pushed out a commit that pretends to fix this. And it makes sense
> > to me. But I'd really appreciate if you (or anyone else for that matter)
> > would take a look and double check that this is ok.
> 
> It looks fine, but the whole "<= 30s" and "first sample" thing looks a
> bit confused. Do you really need *both*?

>From reading the Cobalt code we don't need the <= 30s code anymore - it's
ALWAYS on the first sample.

I wondered if there was another dive computer that did something similar
where we might still need it. I can't think of one, though. Maybe I should
just remove it and see if something breaks :-)

> Also, that whole
> 
>     /* if we have an explicit first cylinder */
>     if (sample->sensor == 0 && first_cylinder != 0)
>         sample->sensor = first_cylinder;
> 
> looks a bit iffy. I guess we're stuck with it, but my gut feel is that
> it's actually an Atomic Cobalt back-end bug, and that the Cobalt
> should just have reported the right sensor, rather than reporting
> sensor 0.

I was wondering about that as well. The backend clearly has code there to
try to report this data - but without access to a Cobalt this is hard to
figure out... the Cobalt is another one of the dive computers where we
can't just do a memory dump.

Jef - I think you were working on a tool to dump the raw data for a dive?
That might be helpful here...

Or I could just send the user a hacked version of Subsurface that dumps
the raw data to a file. He is responsive on trac...

> The reason it's iffy is that another dive computer that does this
> *right* might be impacted.

And that's exactly why I wanted to show this to you...

> But right now, I guess only the Cobalt has
> this whole "first_cylinder" issue, so maybe worry about that
> possibility later.. And quite frankly, we'll need to eventually fix
> the fact that we can only have one cylinder pressure per sample, so
> this code will end up needing changes in more fundamental ways
> eventually. So the whole "this might break in the future" is not a
> very strong objection. The code is pretty much *guaranteed* to break
> in the future for other reasons anyway.

Lovely. See my comment above :-/

Thanks for taking a look. I'll remove the <= 30s checks and leave it as is
otherwise.

/D