Remove some timestamp format assumptions

Don’t assume that current-time and plain encode-time return
timestamps in (HI LO US PS) format.
* lisp/gnus/gnus-art.el (article-make-date-line)
(article-lapsed-string):
* lisp/gnus/gnus-demon.el (gnus-demon-time-to-step):
* lisp/gnus/gnus-diary.el (gnus-user-format-function-d):
* lisp/gnus/nnmaildir.el (nnmaildir-request-expire-articles):
* lisp/net/pop3.el (pop3-uidl-dele):
* lisp/org/ox-publish.el (org-publish-sitemap):
* lisp/vc/vc-hg.el (vc-hg-state-fast):
Simplify and remove assumptions about timestamp format.
* lisp/gnus/gnus-art.el (article-lapsed-string):
* lisp/gnus/gnus-diary.el (gnus-user-format-function-d):
Do not worry about time-subtract returning nil; that's not possible.
* lisp/gnus/gnus-diary.el (gnus-user-format-function-d):
Avoid race due to duplicate current-time calls.
* lisp/vc/vc-hg.el (vc-hg--time-to-integer): Remove; no longer used.

1 files changed, 21 insertions, 24 deletions

diff --git a/lisp/gnus/gnus-diary.el b/lisp/gnus/gnus-diary.el
index 51e39958798..ceb0d4a30da 100644
--- a/lisp/gnus/gnus-diary.el
+++ b/lisp/gnus/gnus-diary.el
@@ -159,32 +159,29 @@ There are currently two built-in format functions:
   ;; Code partly stolen from article-make-date-line
   (let* ((extras (mail-header-extra header))
 	 (sched (gnus-diary-header-schedule extras))
-	 (occur (nndiary-next-occurrence sched (current-time)))
 	 (now (current-time))
+	 (occur (nndiary-next-occurrence sched now))
 	 (real-time (time-subtract occur now)))
-    (if (null real-time)
-	"?????"
-      (let* ((sec (+ (* (float (car real-time)) 65536) (cadr real-time)))
-	     (past (< sec 0))
-	     delay)
-	(and past (setq sec (- sec)))
-	(unless (zerop sec)
-	  ;; This is a bit convoluted, but basically we go through the time
-	  ;; units for years, weeks, etc, and divide things to see whether
-	  ;; that results in positive answers.
-	  (let ((units `((year . ,(* 365.25 24 3600))
-			 (month . ,(* 31 24 3600))
-			 (week . ,(* 7 24 3600))
-			 (day . ,(* 24 3600))
-			 (hour . 3600)
-			 (minute . 60)))
-		unit num)
-	    (while (setq unit (pop units))
-	      (unless (zerop (setq num (ffloor (/ sec (cdr unit)))))
-		(setq delay (append delay `((,(floor num) . ,(car unit))))))
-	      (setq sec (- sec (* num (cdr unit)))))))
-	(funcall gnus-diary-delay-format-function past delay)))
-    ))
+    (let* ((sec (encode-time real-time 'integer))
+	   (past (< sec 0))
+	   delay)
+      (and past (setq sec (- sec)))
+      (unless (zerop sec)
+	;; This is a bit convoluted, but basically we go through the time
+	;; units for years, weeks, etc, and divide things to see whether
+	;; that results in positive answers.
+	(let ((units `((year . ,(round (* 365.25 24 3600)))
+		       (month . ,(* 31 24 3600))
+		       (week . ,(* 7 24 3600))
+		       (day . ,(* 24 3600))
+		       (hour . 3600)
+		       (minute . 60)))
+	      unit num)
+	  (while (setq unit (pop units))
+	    (unless (zerop (setq num (floor sec (cdr unit))))
+	      (setq delay (append delay `((,num . ,(car unit))))))
+	    (setq sec (mod sec (cdr unit))))))
+      (funcall gnus-diary-delay-format-function past delay))))
 
 ;; #### NOTE: Gnus sometimes gives me a HEADER not corresponding to any
 ;; message, with all fields set to nil here. I don't know what it is for, and