* src/timefns.c (decode_lisp_time) [FASTER_TIMEFNS]:
Speed up when SPECIFIED_TIME is an integer.
(time_cmp) [FASTER_TIMEFNS]: Speed up when comparing integers.
This is a followup to commit b70369c557 of 2022-08-05
"time-convert): Deprecate calls without an explicit FORM arg".
* doc/lispref/os.texi (Time Conversion):
* src/timefns.c (Ftime_convert): Describe FORM argument as required
as per the advertised calling convention.
* src/timefns.c (decode_float_time): Assume T is finite.
All callers changed.
(decode_time_components): Assume FORM is not TIMEFORM_FLOAT.
All callers changed.
(decode_lisp_time): If the specified time is a float,
signal an error if it is not finite.
(Ffloat_time): If the specified time is a float,
simply return it.
Since time comparison says X == X, have time arithmetic behave
similarly for X - X. This should also be a bit faster due to not
having to test for floats and NaNs.
* src/timefns.c (time_arith, time_cmp):
Simplify by not worrying about NaNs, which are not time values.
(time_arith): Simplify by not worrying about subtracting nil from
nil; the caller now handles this.
(Ftime_subtract): Handle subtracting X from X specially.
* doc/lispref/os.texi (Time Parsing): Clarify which functions
take/return timestamps and which ones take decoded time structures.
* src/timefns.c (Fdecode_time): Clarify TIME argument (bug#46505).
* src/lisp.h (lisp_h_BASE2_EQ, BASE2_EQ): New macros and functions.
* src/timefns.c (tzlookup, Fdecode_time): Use them.
(Ftime_convert): Convert to symbol once, instead of many times.
* src/timefns.c (tzlookup, lisp_time_hz_ticks)
(decode_time_components, lisp_to_timespec, lispint_arith)
(time_arith, time_cmp, Fdecode_time, Fencode_time)
(Ftime_convert): Prefer BASE_EQ to EQ where either will do.
Change current-time and related functions back to using the
traditional list form. Also, add a new boolean variable
current-time-list that lets people try out (TICKS . HZ) form,
with the goal of smoothing the transition.
* src/timefns.c (CURRENT_TIME_LIST): Change default back to true.
(current-time-list): New boolean Lisp variable, which defaults to
CURRENT_TIME_LIST. All uses of CURRENT_TIME_LIST changed to
use current_time_list, and all documentation changed.
* src/timefns.c (Fencode_time): Add support for a 6-elt list arg.
Requested by Max Nikulin for Org (bug#54764).
* test/src/timefns-tests.el (encode-time-alternate-apis): New test.
* doc/lispref/os.texi (Time of Day, Time Conversion):
Move the warnings about DST being -1 to closer to where DST is
discussed, and reword and improve the discussions and warnings.
Be more precise about years before 1969 (possible west of UTC) vs the
Epoch. Mention some problems due to leap seconds, leap years,
daylight saving transitions, and time zone changes. Modernize
discussion of OS timestamp range. Prefer secular ‘BCE’ to religious
‘BC’. Omit discussion of decoded-time-add and make-decoded-time, as
they are in a library and are not always available; instead, mention
the library. Warn about common mistakes when doing simple date
arithmetic.
* src/timefns.c (Fencode_time): In doc string, mention date
arithmetic and tighten up the wording a bit.
* doc/lispref/os.texi (Time Conversion): Add a warning that blind
changing of code calling `encode-time' to use single list instead of
multiple values may cause deferred bugs since it is common to use nil
for ignored arguments such as DST in the old calling convention.
* src/timefns.c (encode-time): Mention the warning added to the elisp
reference in the docstring.
Refactoring related to `encode-time' caused (bug#54731), so it is better
to make apparent the difference between the recommended and the
obsolescent ways to call the function. More details concerning the
purpose and limitations of the DST field are added after discussion with
Paul Eggert in (bug#54764).
* src/sysdep.c [GNU_LINUX]:
(time_from_jiffies): Simplify by using time-convert. Change args
and result type. All uses changed.
(ltime_from_jiffies): Remove; call time_from_jiffies instead.
(put_jiffies): New function.
(get_up_time): Return Lisp_Object not struct timespec.
All uses changed. Simplify by using time-add.
(system_process_attributes): Simplify by using the above.
This fixes some minor problems where timestamps promised more
precision than was actually available. When info is not
available (e.g., sysconf fails) do not place it into the alist.
* src/timefns.c (float_time): Now extern.
Do not warn about timestamps like (1 . 1000). This warning was added
in Emacs 27 as a temporary transition aid, and has now served its
purpose. These timestamps, which Emacs 26 and earlier treated as (HI
. LO) instead of as (TICKS . HZ), were never generated by Emacs
primitives, and in practice the warning seems to have been triggered
only by test cases designed to generate it.
* src/timefns.c (WARN_OBSOLETE_TIMESTAMPS): Remove.
All uses changed to assume it’s false.
(decode_lisp_time): Simplify by taking a bool instead of an
integer bitmask. All uses changed.
* src/pdumper.c (hash_table_thaw): Pacify -Wconversion so
we can use make_nil_vector again.
* src/timefns.c (syms_of_timefns): Prefer make_nil_vector
to make_vector with Qnil.
* src/timefns.c (frac_to_double): Omit optimization that is
invalid on machines with excess precision (e.g., gcc x86 with 387
FPU), because it double-rounds. Found via ‘gcc -m32’ on x86-64.
* src/timefns.c (frac_to_double) [FASTER_TIMEFNS]: Prefer intmax_t
division or double division to mpz division if they also yield the
correctly rounded result.
* src/timefns.c (frac_to_double): Pass FLT_RADIX to mpz_sizeinbase
instead of doing the radix calculation ourselves, not always
correctly. Fix off-by-one error in scale, which caused
double-rounding.
(decode_time_components): Use frac_to_double (via decode_ticks_hz)
to fix double-rounding error that can occur even though
intermediate results are long double.
* test/src/timefns-tests.el (float-time-precision):
Test the above fixes.
