RE: epicsTime_localtime() currently reports daylight saving time incorrectly on Windows

[<freddie.akeroyd@stfc.ac.uk>]

Hi Michael,

All tests now pass OK
Thank you

Regards,

Freddie

-----Original Message-----
From: Michael Davidsaver [mailto:mdavidsaver@gmail.com] 
Sent: 19 October 2015 17:43
To: Akeroyd, Freddie (STFC,RAL,ISIS); tech-talk@aps.anl.gov
Subject: Re: epicsTime_localtime() currently reports daylight saving time incorrectly on Windows

> ... should be honoured by the built-in localtime() function

I've make a quick attempt to do this which passes w/ mingw+wine.  Please pull and retest.


On 10/19/2015 12:17 PM, freddie.akeroyd@stfc.ac.uk wrote:
> Hi Michael,
> 
> test results enclosed for running on windows-x64, _tzset() is indeed 
> ignored by epicsTime_localtime() (or rather by 
> GetTimeZoneInformation()) but should be honoured by the built-in 
> localtime() function
> 
> Regards,
> 
> Freddie
> 
>> -----Original Message-----
>> From: Michael Davidsaver [mailto:mdavidsaver@gmail.com]
>> Sent: 19 October 2015 14:59
>> To: Akeroyd, Freddie (STFC,RAL,ISIS); tech-talk@aps.anl.gov
>> Subject: Re: epicsTime_localtime() currently reports daylight saving 
>> time incorrectly on Windows
>>
>> I've tried to create a test case for this.  Please run 
>> epicsTimeZoneTest in src/libCom/test and report the results.
>>
>> git clone --branch tztest 
>> https://github.com/mdavidsaver/epics-base.git
>> cd epics-base
>> make -C src libCom/test
>> ./src/libCom/test/O.*/epicsTimeZoneTest.exe
>>
>> I've already found that _tzset() with WINE 1.6.2 doesn't seem to have 
>> any effect, so the tests fail.
>>
>> On 10/19/2015 07:10 AM, freddie.akeroyd@stfc.ac.uk wrote:
>>> Hi,
>>>
>>> I've noticed that the "tm_isdst" member of the "struct tm" returned 
>>> by the
>> Win32 implementation of epicsTime_localtime() is not currently 
>> reporting DST status correctly. The problem is that though the 
>> StandardDate / DaylightDate members of TIME_ZONE_INFORMATION are 
>> SYSTEMTIME structures, they have slightly different interpretations of their contents than
>> usual in particular   wDay == 5  means "last occurrence of wDayOfWeek in
>> the month" rather than "5th day of month". One solution would be to 
>> use the status returned by GetTimeZoneInformation() which now seems 
>> to include the DST status, but I think the epicsTime_localtime() 
>> implementation could actually be replace by something simpler like 
>> (give or take a NULL pointer
>> check):
>>>
>>> 	    memcpy(pTM, localtime(pAnsiTime), sizeof(struct tm));
>>>
>>> as on WIN32 the localtime() function uses thread local storage 
>>> rather than
>> global static storage (though the same TLS is shared with gmtime() )
>>>
>>> Regards,
>>>
>>> Freddie
>>>
>>>