[PATCH 1 of 2 v2] Allow commit date before Unix epoch

[Florent Gallaire]

> We would be happy to get the ability to use negative timestamps back into
> Mercurial, thanks for looking into it. (I double checked with Matt Mackall,
> there is no sea monster hiding under the bed that would awaken if we get
> that back.)

Happy to have confirmation that this patch could be useful to some people.

> 1) We'll want your second patch, the one with an alternative fix for
> issue2513 to be the first one so that we don't regress in the middle of the
> series. This is useful for bisecting and also give us the ability to
> partially accept series if some parts needs reword.
> (note that I've not reviewed this patch 2 in details yet and it might need
> changes, from a quick glance we'll want a larger description of the
> changeset to explain what is happening and why it is correct)

OK for no regression, I just posted a one patch version (I think it's
compatible with the "a one change per patch" concept).

> 3) We'll want some tests. If you are to allow a new range of value, we want
> to make sure it is properly recorded and displayed. Ideally, you would run
> theses new tests both on unix* and windows, but if you don't have access to
> a windows machine, this is fine.

I set up a Windows Seven on VirtualBox just to make the tests. I know
time gestion is a sensible part of a DVCS, but one more time, this was
allowed from 2005 to 2010, there's nothing new to test !

Florent

-- 
FLOSS Engineer & Lawyer