bug#26632: TAI<->UTC conversion botches 1961 to 1971

[Zefram <zefram@fysh.org>]

Mark H Weaver wrote:
>patch adds the TAI-UTC tables for 1961-1971 and uses them to implement
>TAI<->UTC conversions over that time range with nanosecond accuracy.

On a quick inspection of the code, that looks good.

>        I'm vaguely concerned about violating widely-held assumptions,
>e.g. that UTC runs at the same rate as TAI

If an application assumes that for pre-1972 times, then the application
is broken.  Note that any application currently using the srfi-19 library
for pre-1972 TAI<->UTC conversions already has a bigger problem, in that
it's getting false answers from the library.  It's hard to see how fixing
the library could make any previously-working program stop working.

>             which might cause some code on top of Guile to misbehave if
>the system clock is set pre-1972,

If the system clock is incorrect by decades, there will be many other
problems to deal with.

>I'm curious to hear opinions on this.

My view is that this change should definitely be applied.  But it's also
worth thinking about what the alternative is, if the correct conversions
are somehow too shocking for innocent programs to be exposed to them.
Making no change isn't a realistic option: the library is producing
false answers, which are no use to anyone.  It's clearly a bug in the
library, and needs to be addressed somehow.  The only other defensible
option would be to declare pre-1972 UTC out of scope for the library,
having attempted conversions signal an error.  That would have to be
documented, and it seems like it would still amount to a deviation from
the requirements of SRFI-19.

-zefram