careful with the latest master and your 'real' dives

[Dirk Hohndel]

Linus Torvalds <torvalds at linux-foundation.org> writes:

> On Thu, Sep 20, 2012 at 3:01 PM, Dirk Hohndel <dirk at hohndel.org> wrote:
>>
>> Something in the new time handling code wasn't as well tested as I hoped
>
> Hmm. The time conversion functions themselves were actually tested
> with millions (literally) of random dates between 1970 and 2099.
>
> But the testing I did was just verifying that the new utc_mkdate()
> followed by the (old, and thus tested) utc_mktime() got back the same
> 64-bit timestamp.
>
> My test program didn't actually test any of the *users* of utc_mkdate().
>
> I can reproduce it by just loading a non-trip dive list, asking for
> auto-trip (which makes it all *show* correctly), and then saving the
> result.
>
> The trip data save is bogus, even if the dates showed correctly in the
> divelist before the saving.

Yes - see my later email. Mea culpa. I pointed at the wrong code change.
It wasn't the new time handling code (you'll admit, it was a likely
suspect), it was the change in data structures...

Fixes to this (and to the bug that Lubomir found an hour ago) have been
pushed out to master.

/D