mu_scan_datetime: Fix error recovery. %Z handles both time-zone forms.

The value stored in the memory location pointed to by the endp argument
has changed.  On the MU_ERR_PARSE error, it is (as earlier) the position
in input where conversion stopped.  On the MU_ERR_FORMAT, however, it
is the character in fmt where the format error was detected.

* libmailutils/datetime/scantime.c (peek_state): Return MU_ERR_FORMAT
if mu_list_tail fails (the list is empty).  This reflects unbalanced
closing parenthesis or bracket.
(scan_recovery): fix algorithm.
(mu_scan_datetime): %Z handles both abbreviated time-zones and
time-zones as hour offset from GMT.
Improve error handling.
On MU_ERR_PARSE error, point *endp to the character in input where
conversion stopped.
On MU_ERR_FORMAT error, point it to the character in fmt, where the
format error was detected.
* libmailutils/tests/scantime.c: Improve error handling.
* libmailutils/tests/scantime.at: More testcases.

2 files changed, 97 insertions, 1 deletions

diff --git a/libmailutils/tests/scantime.at b/libmailutils/tests/scantime.at
index 3b5a1b1a1..7969158f4 100644
--- a/libmailutils/tests/scantime.at
+++ b/libmailutils/tests/scantime.at
@@ -153,5 +153,89 @@ sec=26,min=25,hour=13,mday=3,mon=4,year=111,wday=2,yday=122,tz=7200
 scantime: 5: parse failed near ; 03 May 2011 13:25:26 +0200
 ])
 
+# This expanded scan specification takes into account all possible
+# variants of the date/time specification that ever existed in the
+# From_ line of UNIX mbox mailboxes.
+SCANTIME([MBOX From_ time],[from_],
+[[%a %b %e %H:%M%[:%S%] %(%(%z%|%Z%) %Y%|%Y %(%z%|%Z%)%|%Y%)]],
+[Wed Dec  2 05:53 1992
+Wed Dec  2 05:53:22 1992
+Wed Dec  2 05:53 PST 1992
+Wed Dec  2 05:53:22 PST 1992
+Wed Dec  2 05:53 -0700 1992
+Wed Dec  2 05:53:22 -0700 1992
+Wed Dec  2 05:53 1992 PST
+Wed Dec  2 05:53:22 1992 PST
+Wed Dec  2 05:53 1992 -0700
+Wed Dec  2 05:53:22 1992 -0700
+],
+[sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=0
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=0
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+])
+
+# This one is a simplification of the above, taking into account the
+# fact that %Z handles correctly numeric time zones as well (after
+# 2020-11-27).  Notice ordering of the alternative forms.
+SCANTIME([MBOX From_ time (alternative)],[from_],
+[[%a %b %e %H:%M%[:%S%] %(%(%Z %Y%|%Y %Z%)%|%Y%)]],
+[Wed Dec  2 05:53 1992
+Wed Dec  2 05:53:22 1992
+Wed Dec  2 05:53 PST 1992
+Wed Dec  2 05:53:22 PST 1992
+Wed Dec  2 05:53 -0700 1992
+Wed Dec  2 05:53:22 -0700 1992
+Wed Dec  2 05:53 1992 PST
+Wed Dec  2 05:53:22 1992 PST
+Wed Dec  2 05:53 1992 -0700
+Wed Dec  2 05:53:22 1992 -0700
+],
+[sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=0
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=0
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+])
+
+# This is a final version of the From_ format line.  It eliminates
+# one level of alternative specs by joining "%Y %Z" and "%Y" into
+# one spec.
+SCANTIME([MBOX From_ time (simplified)],[from_],
+[[%a %b %e %H:%M%[:%S%] %(%Z %Y%|%Y%[ %Z%]%)]],
+[Wed Dec  2 05:53 1992
+Wed Dec  2 05:53:22 1992
+Wed Dec  2 05:53 PST 1992
+Wed Dec  2 05:53:22 PST 1992
+Wed Dec  2 05:53 -0700 1992
+Wed Dec  2 05:53:22 -0700 1992
+Wed Dec  2 05:53 1992 PST
+Wed Dec  2 05:53:22 1992 PST
+Wed Dec  2 05:53 1992 -0700
+Wed Dec  2 05:53:22 1992 -0700
+],
+[sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=0
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=0
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-28800
+sec=0,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+sec=22,min=53,hour=5,mday=2,mon=11,year=92,wday=3,yday=336,tz=-25200
+])
+
 m4_popdef([SCANTIME])
 dnl ---------------------------------------------------------------------
diff --git a/libmailutils/tests/scantime.c b/libmailutils/tests/scantime.c
index a9a01af71..5a0d51141 100644
--- a/libmailutils/tests/scantime.c
+++ b/libmailutils/tests/scantime.c
@@ -65,13 +65,25 @@ main (int argc, char **argv)
       if (!*buf)
 	continue;
       rc = mu_scan_datetime (buf, format, &tm, &tz, &endp);
-      if (rc)
+      switch (rc)
 	{
+	case 0:
+	  break;
+	  
+	case MU_ERR_PARSE:
 	  if (*endp)
 	    mu_error ("%d: parse failed near %s", line, endp);
 	  else
 	    mu_error ("%d: parse failed at end of input", line);
 	  continue;
+
+	case MU_ERR_FORMAT:
+	  mu_error ("%d: error in format string near %s", line, endp);
+	  continue;
+
+	default:
+	  mu_error ("%d: %s", line, mu_strerror (rc));
+	  exit (1);
 	}
       if (*endp)
 	mu_printf ("# %d: stopped at %s\n", line, endp);