[PATCH] org-test: Fix zone-dependent miscalculation of days of week

[PATCH] org-test: Fix zone-dependent miscalculation of days of week

[Kyle Meyer]

Hello,

Three clock tests are failing on my end:

   FAILED  test-org-clock/clock-drawer-dwim
   (... :explanation
    (array-elt 35 (different-atoms (87 "#x57" "?W") (84 "#x54" "?T"))))

   FAILED  test-org-clock/org-clock-timestamps-change
   (... :explanation
    (array-elt 20 (different-atoms (97 "#x61" "?a") (117 "#x75" "?u"))))

   FAILED  test-org-clok/org-clock-update-time-maybe
   (... :explanation
    (array-elt 19 (different-atoms (70 "#x46" "?F") (83 "#x53" "?S"))))

Those stem from org-test-day-of-weeks-{abbrev,full} not having the
expected value.  Those variables are supposed to list Sunday through
Saturday in the machine's locale.  Here's what I see on my end:

  org-test-day-of-weeks-full’s value is
  ["Saturday" "Monday" "Monday" "Tuesday" "Wednesday" "Thursday"
   "Friday"]

The patch below fixes the issue on my end.  In addition to my usual
locale, I tested it with another one (de_BE.utf8), and all the tests
passed.

-- >8 --
Subject: [PATCH] org-test: Fix zone-dependent miscalculation of days of week

* testing/org-test.el (org-test-day-of-weeks-seconds): Specify seconds
for formatting with UTC as time zone.
(org-test-day-of-weeks-abbrev):
(org-test-day-of-weeks-full): Use UTC as time zone when formatting
input.

Avoid calling format-time-string with the local time zone because that
gives the wrong result in some cases.  For example, 2222222 is
supposed to produce the locale's name for "Tuesday" but, when the
local time zone is +0000, (format-time-string "%A" 2222222) returns
the locale's name for "Monday".
---
 testing/org-test.el | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/testing/org-test.el b/testing/org-test.el
index d9fe33284..643c5c766 100644
--- a/testing/org-test.el
+++ b/testing/org-test.el
@@ -549,26 +549,26 @@ (defmacro org-test-capture-warnings (&rest body)
      (nreverse messages)))
 
 (defconst org-test-day-of-weeks-seconds
-  [121223891                            ; Sun
-   30000000                             ; Mon
-   2222222                              ; Tue
-   500000                               ; Wed
-   1000                                 ; Thu
-   89173                                ; Fri
-   666666666]                           ; Sat
+  [302400                               ; Sun
+   388800                               ; Mon
+   475200                               ; Tue
+   561600                               ; Wed
+   648000                               ; Thu
+   734400                               ; Fri
+   820800]                              ; Sat
   "Epoch seconds for generating days of week strings.
 Starts at Sunday, ends at Saturday.")
 
 (defconst org-test-day-of-weeks-abbrev
   (apply #'vector
-         (seq-map (apply-partially #'format-time-string "%a")
+         (seq-map (lambda (s) (format-time-string "%a" s t))
                   org-test-day-of-weeks-seconds))
   "Vector of abbreviated names of days of week.
 See `org-test-day-of-weeks-seconds'.")
 
 (defconst org-test-day-of-weeks-full
   (apply #'vector
-         (seq-map (apply-partially #'format-time-string "%A")
+         (seq-map (lambda (s) (format-time-string "%A" s t))
                   org-test-day-of-weeks-seconds))
   "Vector of full names for days of week.
 See `org-test-day-of-weeks-seconds'.")

base-commit: 671ca44df04801514fd77faf06e7e0b3216188a6
-- 
2.41.0

Re: [PATCH] org-test: Fix zone-dependent miscalculation of days of week

[Ihor Radchenko]

Kyle Meyer <kyle@kyleam.com> writes:

> Three clock tests are failing on my end:
>
>    FAILED  test-org-clock/clock-drawer-dwim
>    FAILED  test-org-clock/org-clock-timestamps-change
>    FAILED  test-org-clok/org-clock-update-time-maybe
>
> Those stem from org-test-day-of-weeks-{abbrev,full} not having the
> expected value.  Those variables are supposed to list Sunday through
> Saturday in the machine's locale.  Here's what I see on my end:
>
>   org-test-day-of-weeks-full’s value is
>   ["Saturday" "Monday" "Monday" "Tuesday" "Wednesday" "Thursday"
>    "Friday"]
>
> The patch below fixes the issue on my end.  In addition to my usual
> locale, I tested it with another one (de_BE.utf8), and all the tests
> passed.
> ...
>  (defconst org-test-day-of-weeks-seconds
> -  [121223891                            ; Sun
> -   30000000                             ; Mon
> -   2222222                              ; Tue
> -   500000                               ; Wed
> -   1000                                 ; Thu
> -   89173                                ; Fri
> -   666666666]                           ; Sat
> +  [302400                               ; Sun
> +   388800                               ; Mon
> +   475200                               ; Tue
> +   561600                               ; Wed
> +   648000                               ; Thu
> +   734400                               ; Fri
> +   820800]                              ; Sat

I suspect that the failures are because of your timezone.
If my guess is right, there will always be some timezone where a given
number seconds from epoch is a different day...
I am not sure how to address this problem.

-- 
Ihor Radchenko // yantar92,
Org mode contributor,
Learn more about Org mode at <https://orgmode.org/>.
Support Org development at <https://liberapay.com/org-mode>,
or support my work at <https://liberapay.com/yantar92>

Re: [PATCH] org-test: Fix zone-dependent miscalculation of days of week

[Kyle Meyer]

Ihor Radchenko writes:

> I suspect that the failures are because of your timezone.

Yes, that matches my conclusion too and was what I was trying to convey
in the commit message.

> If my guess is right, there will always be some timezone where a given
> number seconds from epoch is a different day...
> I am not sure how to address this problem.

How about my patch?  Notice that format-time-string is invoked with ZONE
set to t so that TIME is always taken as UTC.

Re: [PATCH] org-test: Fix zone-dependent miscalculation of days of week

[Kyle Meyer]

[ Sorry, I somehow managed to drop the References and In-Reply-to
  headers from my previous reply.  I'll include that message again here
  along with some examples. ]

Ihor Radchenko writes:

> I suspect that the failures are because of your timezone.

Yes, that matches my conclusion too and was what I was trying to convey
in the commit message.

> If my guess is right, there will always be some timezone where a given
> number seconds from epoch is a different day...
> I am not sure how to address this problem.

How about my patch?  Notice that format-time-string is invoked with ZONE
set to t so that TIME is always taken as UTC.

Examples
--------

;;; locale en_US.UTF-8
;; local time zone (Japan in this demo)
(format-time-string "%A %T %z" 259200) => "Sunday 09:00:00 +0900"
;; force UTC
(format-time-string "%A %T %z" 259200 t) => "Sunday 00:00:00 +0000"

;; locale de_BE.UTF-8
;; local time zone
(format-time-string "%A %T %z" 259200) => "Sonntag 09:00:00 +0900"
;; force UTC
(format-time-string "%A %T %z" 259200 t) => "Sonntag 00:00:00 +0000"

Re: [PATCH] org-test: Fix zone-dependent miscalculation of days of week

[Ihor Radchenko]

Kyle Meyer <kyle@kyleam.com> writes:

>> If my guess is right, there will always be some timezone where a given
>> number seconds from epoch is a different day...
>> I am not sure how to address this problem.
>
> How about my patch?  Notice that format-time-string is invoked with ZONE
> set to t so that TIME is always taken as UTC.
> ...
> ;;; locale en_US.UTF-8
> ;; local time zone (Japan in this demo)
> (format-time-string "%A %T %z" 259200) => "Sunday 09:00:00 +0900"
> ;; force UTC
> (format-time-string "%A %T %z" 259200 t) => "Sunday 00:00:00 +0000"

Oops. I missed that you added `t' argument.
It indeed makes sense.
Applied, onto bugfix.
https://git.savannah.gnu.org/cgit/emacs/org-mode.git/commit/?id=de0df5b92

-- 
Ihor Radchenko // yantar92,
Org mode contributor,
Learn more about Org mode at <https://orgmode.org/>.
Support Org development at <https://liberapay.com/org-mode>,
or support my work at <https://liberapay.com/yantar92>