Avoid some excess precision in time arithmetic

* doc/misc/emacs-mime.texi (time-date):
Adjust example to match new behavior.
* etc/NEWS: Mention this.
* lisp/calendar/time-date.el (decoded-time-add)
(decoded-time--alter-second):
Don’t lose underestimate precision of seconds component.
* src/bignum.c (mpz): Grow by 1.
* src/timefns.c (trillion_factor): New function.
(timeform_sub_ps_p): Remove.
(time_arith): Avoid unnecessarily-large hz, by reducing the hz
to a value no worse than the worse hz of the two arguments.
The result is always exact unless an error is signaled.
* test/src/timefns-tests.el (timefns-tests--decode-time):
New function.
(format-time-string-with-zone): Test (decode-time LOOK ZONE t)
resolution as well as its numeric value.

1 files changed, 3 insertions, 1 deletions

diff --git a/etc/NEWS b/etc/NEWS
index 9f25cf4af51..3fdc185af4f 100644
--- a/etc/NEWS
+++ b/etc/NEWS
@@ -2166,7 +2166,9 @@ end and duration).
 +++
 *** 'time-add', 'time-subtract', and 'time-less-p' now accept
 infinities and NaNs too, and propagate them or return nil like
-floating-point operators do.
+floating-point operators do.  If both arguments are finite, these
+functions now return exact results instead of rounding in some cases,
+and they also avoid excess precision when that is easy.
 
 +++
 *** New function 'time-equal-p' compares time values for equality.