64 bit time_t in 32 bit ABI?

[Jayachandran C.]

On Sat, Apr 30, 2011 at 12:54 AM, Warner Losh <imp at bsdimp.com> wrote:
> I thought we made all new architectures adopt a 32-bit time_t and that's why it is this way.

Although it is not required by standard, having time_t as long is (in
my opinion) would be the least surprising implementation. Having it
'long long' on 32-bit will avoid the 2038 issue, but it also makes us
incompatible with other platforms and might trip up some programs.

That said, if this is a decision that has been taken, we will abide by that.

JC.

> Warner
>
> On Apr 29, 2011, at 11:36 AM, Jayachandran C. wrote:
>
>> In sys/mips/include/_types.h, I see
>>
>> | 119 typedef __int64_t       __time_t;               /* time()... */
>>
>> Which as far as I can see, is not right for the 32-bit ABIs. But this
>> definition has been there for a long time, so I would like to know if
>> there is any reason for this?
>>
>> Since this is a user-visible type, changing it would break binary compatibility.
>>
>> Thanks,
>> JC.