Current subsurface breakage

[Dirk Hohndel]

On Fri, Jan 02, 2015 at 09:35:39PM -0800, Linus Torvalds wrote:
> And that, in turn, is because it turns out that the git loading and
> the XML loading are subtly different.

That's not good.

> so what happened was that since I use the git save format, any missing
> temperatures in my save files would get populated as "same as the old
> sample". Then we cleared duplicates, so when we saved, we saved the
> same thing.

Bleh. And I didn't test against git format but against XML format.

> And then commit 6cf3787a0ed1 removed the clearing, so it would save
> that changed state with all the duplicate temperatures.
> 
> But if you used the XML save format, any missing temperatures would
> never have been initialized in the first place, so that commit made no
> difference.
> 
> Anyway, I'd suggest applying the attached patch that makes the git
> save format not save redundant temperatures. That actually makes your
> partial revert be a non-issue.

The patch looked good and I took it.

> I *also* wonder whether we should perhaps try to make the git and xml
> saving logic be closer to each other. I think the git logic is perhaps
> better in that it's simpler and doesn't special-case a lot of
> different sample cases.

Yes, I would love for the two to be more similar. Whenever I add / modify
features I curse the subtle differences...

/D