RE: [PATCH 1/1] poll: use GetTickCount64() to avoid wrap-around issues

From: Randall S. Becker <rsbecker@nexbridge.com>
Date: Nov 4, 2018, 23:44 UTC
Message-ID: <001101d47498$512b4bf0$f381e3d0$@nexbridge.com>
In-Reply-To: <xmqqefc0mnlh.fsf@gitster-ct.c.googlers.com>

On November 4, 2018 6:26 PM, Junio C Hamano, wrote:

Show 16 quoted lines

> Johannes Sixt <j6t@kdbg.org> writes:
> 
> > Am 03.11.18 um 09:14 schrieb Carlo Arenas:
> >> On Fri, Nov 2, 2018 at 9:44 AM Johannes Sixt <j6t@kdbg.org> wrote:
> >>>
> >>> +      timeout = elapsed >= orig_timeout ? 0 : (int)(orig_timeout -
> >>> + elapsed);
> >>
> >> nitpick: cast to DWORD instead of int
> >
> > No; timeout is of type int; after an explicit type cast we don't want
> > to have another implicit conversion.
> >
> > -- Hannes
> 
> OK, thanks.  It seems that the relative silence after this message is a

sign that

> the resulting patch after squashing is what everybody is happey with?

On my platform (HPE NonStop), DWORD is being defined as unsigned int (32-bit) rather than unsigned long long (64 bit). The definition comes through the odbc/windows.h include, not the compiler or any core definition. It's only a nano-quibble (if even that), because GetTickCount64 is not defined on the platform anyway, so this is probably not a big deal.

Cheers, Randall