time_t

[Derek Atkins]

Graham Leggett <minfrin at sharp.fm> writes:

> Nathan Buchanan wrote:
>
>> I'm a bit out of my league here...but I believe a long (or long int)
>> is defined to be a minimum of 32 bits - so if you're still using a
>> 32 bit system(?) (or processor?) you may still get a 32 bit time_t.
>
> You're right - the 64 bit RHEL5 system showed sizeof(time_t) to be 8,
> while Leopard on my Powerbook G4 said 4, as did RHEL4/i386.
>
> Looking at what APR (the Apache portable runtime) does with its
> portable version of time_t, it explicitly defines it as a 64 bit
> signed type. I suspect gnucash may have to do the same.

GnuCash has a 64-bit time type, called "Timespec".  The problem
is that over time we've moved more and more code to using general
widgets that use time_t instead of Gnucash custom widgets that
use Timespec.

Eventually the OSes will upgrade time_t to 64 bits at which point
we're fine.   We store dates as ISO-8861 strings so #bits doesn't
matter to our XML storage.  In SQL we should be sure to use a 64-bit
time value (or a string).

> Regards,
> Graham

-derek

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord at MIT.EDU                        PGP key available