Fix 5.42 regression with strftime() and daylight savings

ext/POSIX/POSIX.xs

@@ -3589,7 +3589,7 @@ difftime(time1, time2)
 #     sv_setpv(TARG, ...) could be used rather than
 #     ST(0) = sv_2mortal(newSVpv(...))
 void
-strftime(fmt, sec, min, hour, mday, mon, year, wday = -1, yday = -1, isdst = 0)
+strftime(fmt, sec, min, hour, mday, mon, year, wday = -1, yday = -1, isdst = -1)
 	SV *		fmt
 	int		sec
 	int		min
@@ -3605,10 +3605,8 @@ strftime(fmt, sec, min, hour, mday, mon, year, wday = -1, yday = -1, isdst = 0)
             PERL_UNUSED_ARG(wday);
             PERL_UNUSED_ARG(yday);
 
-            /* -isdst triggers backwards compatibility mode for non-zero
-             * 'isdst' */
             SV *sv = sv_strftime_ints(fmt, sec, min, hour, mday, mon, year,
-                                           -abs(isdst));
+                                      isdst);
 	    if (sv) {
                 sv = sv_2mortal(sv);
             }

ext/POSIX/lib/POSIX.pm

@@ -4,7 +4,7 @@ use warnings;
 
 our ($AUTOLOAD, %SIGRT);
 
-our $VERSION = '2.24';
+our $VERSION = '2.25';
 
 require XSLoader;
 

ext/POSIX/lib/POSIX.pod

@@ -1866,49 +1866,17 @@ Identical to the string form of C<$!>, see L<perlvar/$ERRNO>.
 =item C<strftime>
 
 Convert date and time information to string based on the current
-underlying locale of the program (except for any daylight savings time).
-Returns the string.
+underlying locale of the program.
+Returns the string in a mortalized SV; set to an empty string on error.
 
-Synopsis:
-
-	strftime(fmt, sec, min, hour, mday, mon, year,
-		 wday = -1, yday = -1, isdst = 0)
-
-The month (C<mon>) begins at zero,
-I<e.g.>, January is 0, not 1.  The
-year (C<year>) is given in years since 1900, I<e.g.>, the year 1995 is 95; the
-year 2001 is 101.  Consult your system's C<strftime()> manpage for details
-about these and the other arguments.
+ my $sv = strftime(fmt, sec, min, hour, mday, mon, year,
+                   wday = -1, yday = -1, isdst = -1)
 
 The C<wday> and C<yday> parameters are both ignored.  Their values are
 always determinable from the other parameters.
 
-C<isdst> should be C<1> or C<0>, depending on whether or not daylight
-savings time is in effect for the given time or not.
-
-If you want your code to be portable, your format (C<fmt>) argument
-should use only the conversion specifiers defined by the ANSI C
-standard (C99, to play safe).  These are C<aAbBcdHIjmMpSUwWxXyYZ%>.
-But even then, the B<results> of some of the conversion specifiers are
-non-portable.  For example, the specifiers C<aAbBcpZ> change according
-to the locale settings of the user, and both how to set locales (the
-locale names) and what output to expect are non-standard.
-The specifier C<c> changes according to the timezone settings of the
-user and the timezone computation rules of the operating system.
-The C<Z> specifier is notoriously unportable since the names of
-timezones are non-standard. Sticking to the numeric specifiers is the
-safest route.
-
-The arguments, except for C<isdst>, are made consistent as though by
-calling C<mktime()> before calling your system's C<strftime()> function.
-To get correct results, you must set C<isdst> to be the proper value.
-When omitted, the function assumes daylight savings is not in effect.
-
-The string for Tuesday, December 12, 1995 in the C<C> locale.
-
-	$str = POSIX::strftime( "%A, %B %d, %Y",
-				 0, 0, 0, 12, 11, 95, 2 );
-	print "$str\n";
+More details on the behavior and the specification of the other
+parameters are described in L<perlapi/sv_strftime_ints>.
 
 =item C<strlen>
 

ext/POSIX/t/posix.t

@@ -273,7 +273,12 @@ print POSIX::strftime("ok $test # %H:%M, on %m/%d/%y\n", localtime());
 # input fields to strftime().
 sub try_strftime {
     my $expect = shift;
-    my $got = POSIX::strftime("%a %b %d %H:%M:%S %Y %j", @_);
+    my @input = @_;
+
+    # Add zeros to missing parameters.  The final 0 is for isdst, and the zero
+    # forces use of mini_mktime (unless the code changes).
+    push @input, 0 while @input < 9;
+    my $got = POSIX::strftime("%a %b %d %H:%M:%S %Y %j", @input);
     is($got, $expect, "validating mini_mktime() and strftime(): $expect");
 }
 

ext/POSIX/t/time.t

@@ -9,7 +9,7 @@ use strict;
 
 use Config;
 use POSIX;
-use Test::More tests => 31;
+use Test::More tests => 49;
 
 # For the first go to UTC to avoid DST issues around the world when testing.  SUS3 says that
 # null should get you UTC, but some environments want the explicit names.
@@ -226,3 +226,67 @@ SKIP: {
     is(strftime(undef, CORE::localtime), '', "strftime() works if format is undef");
     like($warnings, qr/^Use of uninitialized value in subroutine entry /, "strftime(undef, ...) produces expected warning");
 }
+
+SKIP: {   # GH #23878; test that dst fall back works properly
+    my $skip_count = 9;
+    skip "No mktime()", $skip_count if $Config{d_mktime} ne 'define';
+    my $locale = "Europe/Paris";
+    $ENV{TZ} = $locale;
+    my $t = 1761436800;     # an hour before time should have changed
+
+    # The time in the first test case is such that UTC gives a different day.
+    # If the locale above is unknown, libc is supposed to use UTC; so that's
+    # how we check if the system knows about the rules for Paris; which on
+    # some systems differ from plain CET-1CEST.
+    skip "'$locale' not understood", $skip_count if
+                POSIX::strftime("%F %T%z", localtime($t - 1)) !~ /2025-10-26/;
+
+    # On Windows, the documentation says that it won't understand what our
+    # $locale is set to, but instead of falling back to UTC, it uses the
+    # system timezone, or U.S.  Pacific time.  The trick above therefore
+    # doesn't work, so just skip this batch of tests.
+    skip "'$locale' not understood", $skip_count if $^O eq "MSWin32";
+
+    my @fall = (
+                 [   -1, "2025-10-26 01:59:59+0200", "Chg -1 hr, 1 sec" ],
+                 [    0, "2025-10-26 02:00:00+0200", "Chg -1 hr, 0 sec" ],
+                 [    1, "2025-10-26 02:00:01+0200", "Chg -59 min, 59 sec" ],
+                 [ 3599, "2025-10-26 02:59:59+0200", "Chg -1 sec" ],
+                 [ 3600, "2025-10-26 02:00:00+0100", "At Paris DST fallback" ],
+                 [ 3601, "2025-10-26 02:00:01+0100", "Chg +1 sec" ],
+                 [ 7199, "2025-10-26 02:59:59+0100", "Chg +1 hr, 59m, 59s" ],
+                 [ 7200, "2025-10-26 03:00:00+0100", "Chg +1 hr" ],
+                 [ 7201, "2025-10-26 03:00:01+0100", "Chg +1 hr, 1 sec" ],
+               );
+    for (my $i = 0; $i < @fall; $i++) {
+        is(POSIX::strftime("%F %T%z", localtime $t + $fall[$i][0]),
+           $fall[$i][1], $fall[$i][2]);
+    }
+}
+
+SKIP: {   # GH #23878: test that dst spring forward works properly; use a
+          # locale that MS narcissism should be able to handle
+    my $skip_count = 9;
+    skip "No mktime()", $skip_count if $Config{d_mktime} ne 'define';
+    my $locale = "PST8PDT";
+    $ENV{TZ} = $locale;
+    POSIX::tzset();
+    my $t = 1741510800;     # an hour before time should have changed
+
+    my @spring = (
+                  [   -1, "2025-03-09 00:59:59-0800", "Chg -1 hr, 1 sec" ],
+                  [    0, "2025-03-09 01:00:00-0800", "Chg -1 hr, 0 sec" ],
+                  [    1, "2025-03-09 01:00:01-0800", "Chg -59 min, 59 sec" ],
+                  [ 3599, "2025-03-09 01:59:59-0800", "Chg -1 sec" ],
+                  [ 3600, "2025-03-09 03:00:00-0700",
+                                            "At Redmond DST spring forward" ],
+                  [ 3601, "2025-03-09 03:00:01-0700", "Chg +1 sec" ],
+                  [ 7199, "2025-03-09 03:59:59-0700", "Chg +1 hr, 59m, 59s" ],
+                  [ 7200, "2025-03-09 04:00:00-0700", "Chg +1 hr" ],
+                  [ 7201, "2025-03-09 04:00:01-0700", "Chg +1 hr, 1 sec" ],
+               );
+    for (my $i = 0; $i < @spring; $i++) {
+        is(POSIX::strftime("%F %T%z", localtime $t + $spring[$i][0]),
+           $spring[$i][1], $spring[$i][2]);
+    }
+}