* bug#37046: mod-test-add-nanosecond/valid fails
2019-08-16 5:40 bug#37046: mod-test-add-nanosecond/valid fails Glenn Morris
@ 2019-08-16 23:54 ` Paul Eggert
0 siblings, 0 replies; 2+ messages in thread
From: Paul Eggert @ 2019-08-16 23:54 UTC (permalink / raw)
To: Glenn Morris; +Cc: 37046-done
[-- Attachment #1: Type: text/plain, Size: 681 bytes --]
Glenn Morris wrote:
> Package: emacs
> Severity: minor
> Version: 27.0.50
>
> Somewhere between 3548fd8 and 96ef76e (inclusive),
> mod-test-add-nanosecond/valid began failing.
>
> Ref eg
> https://hydra.nixos.org/build/98764057
> https://hydra.nixos.org/build/98750439
Thanks for reporting that. It should be fixed by commit
2019-08-16T23:25:02Z!eggert@cs.ucla.edu (attached) so I'm boldly closing the bug
report.
On a related topic, perhaps tests shouldn't depend on time-of-day? That kind of
phase-of-the-moon stuff can make maintenance difficult - though, as it happened,
it wasn't a problem this time as I discovered and fixed the bug before reading
your bug report.
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-Fix-time-add-rounding-bug.patch --]
[-- Type: text/x-patch; name="0001-Fix-time-add-rounding-bug.patch", Size: 6673 bytes --]
From f9fd12a30b3d94eb48f7b309907d136d7b2682ac Mon Sep 17 00:00:00 2001
From: Paul Eggert <eggert@cs.ucla.edu>
Date: Fri, 16 Aug 2019 16:25:02 -0700
Subject: [PATCH] Fix time-add rounding bug
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Without this fix, time arithmetic yielded results that were not
mathematically accurate, even though the exact results were
representable; for example, (time-add 0 1e-13) yielded a timestamp
equal to 0 instead of to 1e-13.
* lisp/timezone.el (timezone-time-from-absolute):
Let time-add do its thing rather than using floating point
internally, which has rounding errors. We now have bignums and so
don’t need floating point to avoid overflow issues.
* src/timefns.c (timeform_sub_ps_p): New function.
(time_arith): If either argument is a float, represent the
result exactly instead of discarding sub-ps info.
* test/lisp/timezone-tests.el (timezone-tests-time-from-absolute):
Don’t assume (HI LO US PS) timestamp format.
* test/src/emacs-module-tests.el (mod-test-add-nanosecond/valid):
Don’t assume that time-add discards sub-ns info.
* test/src/timefns-tests.el (time-rounding-tests):
Add regression test to detect time-add rounding bug.
---
lisp/timezone.el | 4 ++--
src/timefns.c | 14 +++++++++++---
test/lisp/timezone-tests.el | 7 ++++---
test/src/emacs-module-tests.el | 7 +++++--
test/src/timefns-tests.el | 3 +++
5 files changed, 25 insertions(+), 10 deletions(-)
diff --git a/lisp/timezone.el b/lisp/timezone.el
index ff0b266245..ce881a8c95 100644
--- a/lisp/timezone.el
+++ b/lisp/timezone.el
@@ -284,14 +284,14 @@ timezone-zone-to-minute
(defun timezone-time-from-absolute (date seconds)
"Compute the UTC time equivalent to DATE at time SECONDS after midnight.
-Return a list suitable as an argument to `current-time-zone',
+Return a Lisp timestamp suitable as an argument to `current-time-zone',
or nil if the date cannot be thus represented.
DATE is the number of days elapsed since the (imaginary)
Gregorian date Sunday, December 31, 1 BC."
(let* ((current-time-origin 719163)
;; (timezone-absolute-from-gregorian 1 1 1970)
(days (- date current-time-origin))
- (seconds-per-day (float 86400))
+ (seconds-per-day 86400)
(day-seconds (* days seconds-per-day)))
(condition-case nil (time-add day-seconds seconds)
(range-error))))
diff --git a/src/timefns.c b/src/timefns.c
index e9d1a9bf64..a4c1c4cb28 100644
--- a/src/timefns.c
+++ b/src/timefns.c
@@ -661,10 +661,18 @@ enum timeform
TIMEFORM_HI_LO_US, /* seconds plus microseconds (HI LO US) */
TIMEFORM_NIL, /* current time in nanoseconds */
TIMEFORM_HI_LO_US_PS, /* seconds plus micro and picoseconds (HI LO US PS) */
+ /* These two should be last; see timeform_sub_ps_p. */
TIMEFORM_FLOAT, /* time as a float */
TIMEFORM_TICKS_HZ /* fractional time: HI is ticks, LO is ticks per second */
};
+/* True if Lisp times of form FORM can express sub-picosecond timestamps. */
+static bool
+timeform_sub_ps_p (enum timeform form)
+{
+ return TIMEFORM_FLOAT <= form;
+}
+
/* From the valid form FORM and the time components HIGH, LOW, USEC
and PSEC, generate the corresponding time value. If LOW is
floating point, the other components should be zero and FORM should
@@ -1016,8 +1024,8 @@ lispint_arith (Lisp_Object a, Lisp_Object b, bool subtract)
/* Given Lisp operands A and B, add their values, and return the
result as a Lisp timestamp that is in (TICKS . HZ) form if either A
- or B are in that form, (HI LO US PS) form otherwise. Subtract
- instead of adding if SUBTRACT. */
+ or B are in that form or are floats, (HI LO US PS) form otherwise.
+ Subtract instead of adding if SUBTRACT. */
static Lisp_Object
time_arith (Lisp_Object a, Lisp_Object b, bool subtract)
{
@@ -1077,7 +1085,7 @@ time_arith (Lisp_Object a, Lisp_Object b, bool subtract)
otherwise the (HI LO US PS) form for backward compatibility. */
return (EQ (hz, make_fixnum (1))
? ticks
- : aform == TIMEFORM_TICKS_HZ || bform == TIMEFORM_TICKS_HZ
+ : timeform_sub_ps_p (aform) || timeform_sub_ps_p (bform)
? Fcons (ticks, hz)
: ticks_hz_list4 (ticks, hz));
}
diff --git a/test/lisp/timezone-tests.el b/test/lisp/timezone-tests.el
index 4b5f5617ec..c374042fa5 100644
--- a/test/lisp/timezone-tests.el
+++ b/test/lisp/timezone-tests.el
@@ -135,9 +135,10 @@
(should (equal (timezone-zone-to-minute "*invalid*") 0)))
(ert-deftest timezone-tests-time-from-absolute ()
- (should (equal (timezone-time-from-absolute (* 2020 365) ; Jan 1 2020
- (* 12 60 60)) ; 12:00
- '(23911 48704 0 0))))
+ (should (time-equal-p
+ (timezone-time-from-absolute (* 2020 365) ; Jan 1 2020
+ (* 12 60 60)) ; 12:00
+ '(23911 48704 0 0))))
;; TODO: Write tests for timezone-tests-time-zone-from-absolute, which is a
;; bit tricky since the results depend on `current-time-zone'.
diff --git a/test/src/emacs-module-tests.el b/test/src/emacs-module-tests.el
index c44c386d30..c510784731 100644
--- a/test/src/emacs-module-tests.el
+++ b/test/src/emacs-module-tests.el
@@ -335,12 +335,15 @@ module--test-assertion
;; New (TICKS . HZ) format.
'(123456789 . 1000000000)))
(ert-info ((format "input: %s" input))
- (let ((result (mod-test-add-nanosecond input)))
+ (let ((result (mod-test-add-nanosecond input))
+ (desired-result
+ (let ((hz 1000000000))
+ (time-add (time-convert input hz) (cons 1 hz)))))
(should (consp result))
(should (integerp (car result)))
(should (integerp (cdr result)))
(should (cl-plusp (cdr result)))
- (should (time-equal-p result (time-add input '(0 0 0 1000))))))))
+ (should (time-equal-p result desired-result))))))
(ert-deftest mod-test-add-nanosecond/nil ()
(should (<= (float-time (mod-test-add-nanosecond nil))
diff --git a/test/src/timefns-tests.el b/test/src/timefns-tests.el
index 1b1032deaa..362e7655a9 100644
--- a/test/src/timefns-tests.el
+++ b/test/src/timefns-tests.el
@@ -145,6 +145,9 @@ timefns-tests--have-leap-seconds
(< 0.99 (/ (- (float-time a)) (float-time b))
1.01))))))))
+(ert-deftest time-rounding-tests ()
+ (should (time-equal-p 1e-13 (time-add 0 1e-13))))
+
(ert-deftest encode-time-dst-numeric-zone ()
"Check for Bug#35502."
(should (time-equal-p
--
2.17.1
^ permalink raw reply related [flat|nested] 2+ messages in thread