Time-Piece-1.13 test failures on HP-UX

This is the conversation I had with Matt about this issue and the  
patch I provided for fixing it.

--Gisle

Begin forwarded message:

> From: Gisle Aas <gisle@activestate.com>
> Date: May 5, 2008 11:30:12 GMT+02:00
> To: matt@sergeant.org
> Cc: Jan Dubois <jand@activestate.com>
> Subject: Time-Piece test failures on HP-UX
>
> Hi Matt,
>
> I just wanted to let you know that we experience test failures with  
> Time-Piece-1.13's t/07arith.t test on HP-UX.  The reason is that  
> the system strptime() function on HP-UX returns NULL for dates like  
> "2001-02-29".  For now I've fixed our copy by simply using the  
> bundled strptime even on HP-UX with the patch below.
>
> A better fix would probably be to rework add_months() to not rely  
> on strptime being able to parse illegal dates.
>
> Regards,
> Gisle
>
>
>
>
> Index: main/Camel/src/cpan/T/Time/Time-Piece/main/Piece.xs
> --- main/Camel/src/cpan/T/Time/Time-Piece/main/Piece.xs.~1~     Mon  
> May  5 11:24:20 2008
> +++ main/Camel/src/cpan/T/Time/Time-Piece/main/Piece.xs Mon May  5  
> 11:24:20 2008
> @@ -184,8 +184,10 @@
>  }
>
>  /* No strptime on Win32 or QNX4 */
> -#if defined(WIN32) || (defined(__QNX__) && defined(__WATCOMC__))
> +#if defined(WIN32) || (defined(__QNX__) && defined(__WATCOMC__))  
> || defined(__hpux)
> +#ifndef __hpux
>  #define strncasecmp(x,y,n) strnicmp(x,y,n)
> +#endif
>
>  #if defined(WIN32)
>  #define alloca _alloca
> End of Patch.
Begin forwarded message:

> From: Matt Sergeant <matt@sergeant.org>
> Date: May 5, 2008 15:06:15 GMT+02:00
> To: Gisle Aas <gisle@activestate.com>
> Cc: Jan Dubois <jand@activestate.com>
> Subject: Re: Time-Piece test failures on HP-UX
>
> On 5-May-08, at 5:30 AM, Gisle Aas wrote:
>
>> I just wanted to let you know that we experience test failures  
>> with Time-Piece-1.13's t/07arith.t test on HP-UX.  The reason is  
>> that the system strptime() function on HP-UX returns NULL for  
>> dates like "2001-02-29".  For now I've fixed our copy by simply  
>> using the bundled strptime even on HP-UX with the patch below.
>>
>> A better fix would probably be to rework add_months() to not rely  
>> on strptime being able to parse illegal dates.
>
> Hmm, any idea what posix says about this?
>
> Matt.
Begin forwarded message:

> From: Gisle Aas <gisle@activestate.com>
> Date: May 5, 2008 17:04:40 GMT+02:00
> To: Matt Sergeant <matt@sergeant.org>
> Cc: Jan Dubois <jand@activestate.com>
> Subject: Re: Time-Piece test failures on HP-UX
>
> On May 5, 2008, at 15:06, Matt Sergeant wrote:
>
>> On 5-May-08, at 5:30 AM, Gisle Aas wrote:
>>
>>> I just wanted to let you know that we experience test failures  
>>> with Time-Piece-1.13's t/07arith.t test on HP-UX.  The reason is  
>>> that the system strptime() function on HP-UX returns NULL for  
>>> dates like "2001-02-29".  For now I've fixed our copy by simply  
>>> using the bundled strptime even on HP-UX with the patch below.
>>>
>>> A better fix would probably be to rework add_months() to not rely  
>>> on strptime being able to parse illegal dates.
>>
>> Hmm, any idea what posix says about this?
>
> It doesn't appear that it says anything about this kind of behaviour.
>
>    http://www.opengroup.org/onlinepubs/009695399/functions/ 
> strptime.html
>
> --Gisle
>
Begin forwarded message:

> From: Gisle Aas <gisle@activestate.com>
> Date: May 6, 2008 22:06:11 GMT+02:00
> To: Matt Sergeant <matt@sergeant.org>
> Cc: Jan Dubois <jand@activestate.com>
> Subject: Re: Time-Piece test failures on HP-UX
>
> On May 5, 2008, at 15:06, Matt Sergeant wrote:
>
>> On 5-May-08, at 5:30 AM, Gisle Aas wrote:
>>
>>> I just wanted to let you know that we experience test failures  
>>> with Time-Piece-1.13's t/07arith.t test on HP-UX.  The reason is  
>>> that the system strptime() function on HP-UX returns NULL for  
>>> dates like "2001-02-29".  For now I've fixed our copy by simply  
>>> using the bundled strptime even on HP-UX with the patch below.
>>>
>>> A better fix would probably be to rework add_months() to not rely  
>>> on strptime being able to parse illegal dates.
>>
>
> This is such a patch:
>
> Index: Piece.pm
> --- Piece.pm.~1~        Tue May  6 22:00:41 2008
> +++ Piece.pm    Tue May  6 22:00:41 2008
> @@ -607,12 +607,8 @@
>          $final_month = $final_month % 12;
>      }
>
> -    my $string = ($time->year + $num_years) . "-" .
> -                 ($final_month + 1) . "-" .
> -                 ($time->mday) . " " . $time->hms;
> -    my $format = "%Y-%m-%d %H:%M:%S";
> -    #warn("Parsing string: $string\n");
> -    my @vals = _strptime($string, $format);
> +    my @vals = _mini_mktime($time->sec, $time->min, $time->hour,
> +                            $time->mday, $final_month, $time->year  
> - 1900 + $num_years);
>  #    warn(sprintf("got vals: %d-%d-%d %d:%d:%d\n", reverse(@vals)));
>      return scalar $time->_mktime(\@vals, $time->[c_islocal]);
>  }
> ==== //depot/Piece.xs#5 - /Users/gisle/camel/src/cpan/T/Time/Time- 
> Piece/main/Piece.xs ====
> Index: Piece.xs
> --- Piece.xs.~1~        Tue May  6 22:00:41 2008
> +++ Piece.xs    Tue May  6 22:00:41 2008
> @@ -915,3 +915,37 @@
>         PUSHs(sv_2mortal(newSViv(0)));
>         /* islocal */
>         PUSHs(sv_2mortal(newSViv(0)));
> +
> +void
> +_mini_mktime(int sec, int min, int hour, int mday, int mon, int year)
> +  PREINIT:
> +       struct tm mytm;
> +       time_t t;
> +  PPCODE:
> +       t = 0;
> +       mytm = *gmtime(&t);
> +
> +       mytm.tm_sec = sec;
> +       mytm.tm_min = min;
> +       mytm.tm_hour = hour;
> +       mytm.tm_mday = mday;
> +       mytm.tm_mon = mon;
> +       mytm.tm_year = year;
> +
> +       my_mini_mktime(&mytm);
> +
> +       EXTEND(SP, 11);
> +       PUSHs(sv_2mortal(newSViv(mytm.tm_sec)));
> +       PUSHs(sv_2mortal(newSViv(mytm.tm_min)));
> +       PUSHs(sv_2mortal(newSViv(mytm.tm_hour)));
> +       PUSHs(sv_2mortal(newSViv(mytm.tm_mday)));
> +       PUSHs(sv_2mortal(newSViv(mytm.tm_mon)));
> +       PUSHs(sv_2mortal(newSViv(mytm.tm_year)));
> +       PUSHs(sv_2mortal(newSViv(mytm.tm_wday)));
> +       PUSHs(sv_2mortal(newSViv(mytm.tm_yday)));
> +       /* isdst */
> +       PUSHs(sv_2mortal(newSViv(0)));
> +       /* epoch */
> +       PUSHs(sv_2mortal(newSViv(0)));
> +       /* islocal */
> +       PUSHs(sv_2mortal(newSViv(0)));
> End of Patch.
>

• Time-Piece-1.13 test failures on HP-UX by Gisle Aas