On Mon, Apr 15, 2019 at 11:34:01AM +0200, Branko Čibej wrote:
> On 12.04.2019 11:28, Stefan Sperling wrote:
> > On Wed, Apr 10, 2019 at 12:07:23AM +0200, Branko Čibej wrote:
> >> This is nonsense. We should replace the single use of
> >> SVN_APR_INT64_T_PYCFMT in swigutil_py.c with "L" and cast to
> >> PY_LONG_LONG and be done with it. Such checks will always have failing
> >> edge cases, so let's just get rid of them.
> > Agreed. See https://svn.apache.org/r1857391 Does this look good?
>
> Better. Except that, IIRC, PY_LONG_LONG might not be defined on some
> platform/compiler combinations.
Comments in python's own header files suggest that PY_LONG_LONG will work
with C compilers which support a 'long long' type. This isn't C89 but I
would say this a safe assumption for all compilers we intend to support
going forward, isn't it? In any case, our code base already relies on
'long long' types in other places, e.g. swig bindings and lz4.
Received on 2019-04-15 11:44:20 CEST