[PATCH] Fix mktime

[PATCH] Fix mktime

[Jakub Jelinek]

Hi!

The PR libc/2738 fix was not entirely correct, since even tm with tm_year 69
is representable in certain timezones. This caused e.g. perl-Date-Calc tests
to fail. Below is a fix. Years before 69 surely cannot be represented, for
69 it computes the year and checks for overflow afterwards.

2002-04-13  Jakub Jelinek  <jakub@redhat.com>

	* time/mktime.c (__mktime_internal): If year is 69, don't bail out
	early, but check whether it overflowed afterwards.
	* time/tst-mktime.c (main): Add new tests.

--- libc/time/mktime.c.jj	Sun Apr  7 17:52:27 2002
+++ libc/time/mktime.c	Sat Apr 13 11:12:05 2002
@@ -259,8 +259,10 @@ __mktime_internal (struct tm *tp,
 
   int sec_requested = sec;
 
-  /* Only years after 1970 are defined.  */
-  if (year < 70)
+  /* Only years after 1970 are defined.
+     If year is 69, it might still be representable due to
+     timezone differnces.  */
+  if (year < 69)
     return -1;
 
 #if LEAP_SECONDS_POSSIBLE
@@ -370,6 +372,14 @@ __mktime_internal (struct tm *tp,
 	return -1;
     }
 
+  if (year == 69)
+    {
+      /* If year was 69, need to check whether the time was representable
+	 or not.  */
+      if (t < 0 || t > 2 * 24 * 60 * 60)
+	return -1;
+    }
+
   *tp = tm;
   return t;
 }
--- libc/time/tst-mktime.c.jj	Sat Apr 13 11:45:06 2002
+++ libc/time/tst-mktime.c	Sat Apr 13 11:58:22 2002
@@ -5,7 +5,8 @@
 int
 main (void)
 {
-  struct tm time_str;
+  struct tm time_str, *tm;
+  time_t t;
   char daybuf[20];
   int result;
 
@@ -29,5 +30,38 @@ main (void)
       result = strcmp (daybuf, "Wednesday") != 0;
     }
 
+  setenv ("TZ", "EST", 1);
+#define EVENING69 1 * 60 * 60 + 2 * 60 + 29
+  t = EVENING69;
+  tm = localtime (&t);
+  if (tm == NULL)
+    {
+      (void) puts ("localtime returned NULL");
+      result = 1;
+    }
+  else
+    {
+      time_str = *tm;
+      t = mktime (&time_str);
+      if (t != EVENING69)
+        {
+          printf ("mktime returned %ld, expected %ld\n",
+		  (long) t, EVENING69);
+	  result = 1;
+        }
+      else
+        (void) puts ("Dec 31 1969 EST test passed");
+
+      setenv ("TZ", "CET", 1);
+      t = mktime (&time_str);
+      if (t != (time_t) -1)
+        {
+	  printf ("mktime returned %ld, expected -1\n", (long) t);
+	  result = 1;
+        }
+      else
+        (void) puts ("Dec 31 1969 CET test passed");
+    }
+
   return result;
 }

	Jakub

Re: [PATCH] Fix mktime

[Ulrich Drepper]

[-- Attachment #1: Type: text/plain, Size: 541 bytes --]

On Sat, 2002-04-13 at 02:59, Jakub Jelinek wrote:

> 2002-04-13  Jakub Jelinek  <jakub@redhat.com>
> 
> 	* time/mktime.c (__mktime_internal): If year is 69, don't bail out
> 	early, but check whether it overflowed afterwards.
> 	* time/tst-mktime.c (main): Add new tests.

I've applied the patch.  Thanks,

-- 
---------------.                          ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Red Hat          `--' drepper at redhat.com   `------------------------

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 232 bytes --]

Re: [PATCH] Fix mktime

[Thorsten Kukuk]

On Sat, Apr 13, Jakub Jelinek wrote:

> Hi!
> 
> The PR libc/2738 fix was not entirely correct, since even tm with tm_year 69
> is representable in certain timezones. This caused e.g. perl-Date-Calc tests
> to fail. Below is a fix. Years before 69 surely cannot be represented, for
> 69 it computes the year and checks for overflow afterwards.
> 
> 2002-04-13  Jakub Jelinek  <jakub@redhat.com>
> 
> 	* time/mktime.c (__mktime_internal): If year is 69, don't bail out
> 	early, but check whether it overflowed afterwards.
> 	* time/tst-mktime.c (main): Add new tests.

make check does not longer pass for me with this. Output of
tst-mktime.out is:

Wednesday
Dec 31 1969 EST test passed
mktime returned 3749, expected -1

  Thorsten

-- 
Thorsten Kukuk       http://www.suse.de/~kukuk/        kukuk@suse.de
SuSE Linux AG        Deutschherrenstr. 15-19       D-90429 Nuernberg
--------------------------------------------------------------------    
Key fingerprint = A368 676B 5E1B 3E46 CFCE  2D97 F8FD 4E23 56C6 FB4B

[PATCH] tst-mktime.c

[Jakub Jelinek]

On Wed, Apr 17, 2002 at 02:21:46PM +0200, Thorsten Kukuk wrote:
> On Sat, Apr 13, Jakub Jelinek wrote:
> 
> > Hi!
> > 
> > The PR libc/2738 fix was not entirely correct, since even tm with tm_year 69
> > is representable in certain timezones. This caused e.g. perl-Date-Calc tests
> > to fail. Below is a fix. Years before 69 surely cannot be represented, for
> > 69 it computes the year and checks for overflow afterwards.
> > 
> > 2002-04-13  Jakub Jelinek  <jakub@redhat.com>
> > 
> > 	* time/mktime.c (__mktime_internal): If year is 69, don't bail out
> > 	early, but check whether it overflowed afterwards.
> > 	* time/tst-mktime.c (main): Add new tests.
> 
> make check does not longer pass for me with this. Output of
> tst-mktime.out is:
> 
> Wednesday
> Dec 31 1969 EST test passed
> mktime returned 3749, expected -1

Weird, I get
Wednesday
Dec 31 1969 EST test passed
Dec 31 1969 CET test passed

on all arches I've bootstrapped it (i386,i686,ia64,alpha,alphaev6).
Looks like tst-mktime cannot find the CET zone file on your box.
Does the following work for you?
Passes for me with current glibc, fails with:
Wednesday
mktime returned -1, expected 3749
Dec 31 1969 CET test passed
with glibc before my mktime patch (expected).

2002-04-17  Jakub Jelinek  <jakub@redhat.com>

	* time/tst-mktime.c: Include <stdlib.h>.
	Use %d, not %ld format for EVENING69.
	Include offsets in TZ environment variable.

--- libc/time/tst-mktime.c.jj	Wed Apr 17 14:34:34 2002
+++ libc/time/tst-mktime.c	Wed Apr 17 14:48:32 2002
@@ -1,3 +1,4 @@
+#include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
 #include <time.h>
@@ -30,7 +31,7 @@ main (void)
       result = strcmp (daybuf, "Wednesday") != 0;
     }
 
-  setenv ("TZ", "EST", 1);
+  setenv ("TZ", "EST+5", 1);
 #define EVENING69 1 * 60 * 60 + 2 * 60 + 29
   t = EVENING69;
   tm = localtime (&t);
@@ -45,14 +46,14 @@ main (void)
       t = mktime (&time_str);
       if (t != EVENING69)
         {
-          printf ("mktime returned %ld, expected %ld\n",
+          printf ("mktime returned %ld, expected %d\n",
 		  (long) t, EVENING69);
 	  result = 1;
         }
       else
         (void) puts ("Dec 31 1969 EST test passed");
 
-      setenv ("TZ", "CET", 1);
+      setenv ("TZ", "CET-1", 1);
       t = mktime (&time_str);
       if (t != (time_t) -1)
         {

	Jakub

Re: [PATCH] tst-mktime.c

[Thorsten Kukuk]

On Wed, Apr 17, Jakub Jelinek wrote:

> > make check does not longer pass for me with this. Output of
> > tst-mktime.out is:
> > 
> > Wednesday
> > Dec 31 1969 EST test passed
> > mktime returned 3749, expected -1
> 
> Weird, I get
> Wednesday
> Dec 31 1969 EST test passed
> Dec 31 1969 CET test passed
> 
> on all arches I've bootstrapped it (i386,i686,ia64,alpha,alphaev6).
> Looks like tst-mktime cannot find the CET zone file on your box.

Arg, ok, this was it. On this machine was really no timezone data
installed :(

 Thanks for the hint,
  
   Thorsten

-- 
Thorsten Kukuk       http://www.suse.de/~kukuk/        kukuk@suse.de
SuSE Linux AG        Deutschherrenstr. 15-19       D-90429 Nuernberg
--------------------------------------------------------------------    
Key fingerprint = A368 676B 5E1B 3E46 CFCE  2D97 F8FD 4E23 56C6 FB4B