Question about 'tzparse()'

[Olson, Arthur David (NIH/NCI) [E]]

The bad news is that there is indeed a code problem; the good news is
that it was only introduced back in February of this year (with the
first 64-bit-version release).

The fix should be to move the zeroing of sp->timecnt; the suggested
patch below incorporates both that change and a fix for last week's
report on bad goahead/goback logic. (I've chosen the path of minimal
change in both cases).

				--ado


------- localtime.c -------
*** /tmp/geta21479	Mon Jul 31 13:39:56 2006
--- /tmp/getb21479	Mon Jul 31 13:39:56 2006
***************
*** 5,11 ****
  
  #ifndef lint
  #ifndef NOID
! static char	elsieid[] = "@(#)localtime.c	8.3";
  #endif /* !defined NOID */
  #endif /* !defined lint */
  
--- 5,11 ----
  
  #ifndef lint
  #ifndef NOID
! static char	elsieid[] = "@(#)localtime.c	8.5";
  #endif /* !defined NOID */
  #endif /* !defined lint */
  
***************
*** 556,564 ****
  	}
  	i = 2 * YEARSPERREPEAT;
  	sp->goback = sp->goahead = sp->timecnt > i;
! 	sp->goback &= sp->types[i] == sp->types[0] &&
  		differ_by_repeat(sp->ats[i], sp->ats[0]);
! 	sp->goahead &=
  		sp->types[sp->timecnt - 1] == sp->types[sp->timecnt - 1
- i] &&
  		differ_by_repeat(sp->ats[sp->timecnt - 1],
  			 sp->ats[sp->timecnt - 1 - i]);
--- 556,564 ----
  	}
  	i = 2 * YEARSPERREPEAT;
  	sp->goback = sp->goahead = sp->timecnt > i;
! 	sp->goback = sp->goback && sp->types[i] == sp->types[0] &&
  		differ_by_repeat(sp->ats[i], sp->ats[0]);
! 	sp->goahead = sp->goahead &&
  		sp->types[sp->timecnt - 1] == sp->types[sp->timecnt - 1
- i] &&
  		differ_by_repeat(sp->ats[sp->timecnt - 1],
  			 sp->ats[sp->timecnt - 1 - i]);
***************
*** 916,922 ****
  	load_result = tzload(TZDEFRULES, sp, FALSE);
  	if (load_result != 0)
  		sp->leapcnt = 0;		/* so, we're off a
little */
- 	sp->timecnt = 0;
  	if (*name != '\0') {
  		if (*name == '<') {
  			dstname = ++name;
--- 916,921 ----
***************
*** 967,972 ****
--- 966,972 ----
  			atp = sp->ats;
  			typep = sp->types;
  			janfirst = 0;
+ 			sp->timecnt = 0;
  			for (year = EPOCH_YEAR;
  			    sp->timecnt + 2 <= TZ_MAX_TIMES;
  			    ++year) {

-----Original Message-----
From: David Lawless [mailto:lawless at spamcop.net]
Sent: Friday, July 28, 2006 1:50 PM
To: Olson, Arthur David (NIH/NCI) [E]; tz at lecserver.nci.nih.gov
Subject: Question about 'tzparse()'

Mr. Olson,

I'm wondering about line 920 in 'tzparse()' in 'localtime.c':

917    load_result = tzload(TZDEFRULES, sp, FALSE);
918    if (load_result != 0)
919       sp->leapcnt = 0;     /* so, we're off a little */
920    sp->timecnt = 0;
921    if (*name != '\0') {
922       if (*name == '<') {
923          dstname = ++name;

It appears to have the effect of inhibiting the application of DST per
'/usr/local/etc/timezone/posixrules' when a TZ string such as
TZ="EST+5EDT" is used.  If the line is removed DST is applied.

The effective default if 'posixrules' is not readable, per
TZDEFRULESTRING, is TZ="EST+5EDT,M4.1.0,M10.5.0" which does lead to the
application of DST.

Also RHEL 4.3 (CentOS 4.3) appears to apply DST when TZ="EST+5EDT".

Does 'tzparse()' behave in this manner by design?

Thank you for your efforts and contribution of the excellent. 
zone package.

Sincerely,

David Lawless