bug#43677: 27.1; Calc holidays off-by-one error, incorrect business day calculation

[Mattias Engdegård]

30 sep. 2020 kl. 15.40 skrev Lars Ingebrigtsen <larsi@gnus.org>:

> That's quite a stretch to be bisecting, though...

Bisecting is logarithmic so large stretches are seldom a problem. You can also 
bisect restricted to one path which reduces the number of versions 
dramatically. (Building old versions of Emacs, on the other hand -- that can be 
messy!)

But you don't have to! I found it. The error was introduced in the calendar 
changes between c9f618db441 and fd1f1603a4f, during which the internal date 
numbering changed: date Jan 1, 1 AD went from date 0 to 1. However, the 
business days code remained unchanged and thus started getting the weekdays 
wrong.

Of course there were no tests but that is excusable since this was 2012 and 
tests weren't invented back then.

I'll put together a patch a little later today or tomorrow.