Subsurface deco calculations

[Linus Torvalds]

On Wed, Feb 8, 2017 at 7:24 AM, Miika Turkia <miika.turkia at gmail.com> wrote:
>
> So the index used is correctly 1 with o2 32%. Only used cylinder was
> that 32% as I find it easier to choose one of the common mixes than
> adjusting the mix for first cylinder. However, with Vyper the used
> cylinder index was 0 as that was the 32% mix in that computer. And
> that is the only cylinder stored in the log in my original DL.
>
> Should we do the merge more intelligently mapping the cylinder indexes
> to match the gas mixes, or should I start using same cylinder indexes
> on different DCs?

We actually do do that, but:

 (a) it gets very little testing in practice

 (b)  it has some heuristics for figuring out the closest gases

 (c) it might easily have broken for various reasons (just one of
which is the change in gas change events: with the new gaschange, the
event itself has _only_ the index, no longer a gasmix, maybe that
confused something, but it could have been any other change too)

.. and for all I know, _your_ particular case never worked at all,
because this code really doesn't get very much coverage.

So I suspect I can easily fix this (famous last words), but I'd like
to see a test-case, and to do that I'd ask to get the XML for the two
computers for that dive (before merging), and also know your workflow.
When did you delete the cylinders, for example? It sounds like you
downloaded the Vyper data first (with just one cylinder), and after
that you downloaded the OSTC data (that had three cylinders and you
used index=1)?

                Linus