I wrote:
> Andrew Dunstan <andrew@dunslane.net> writes:
>> On 04/23/2016 05:30 PM, Christian Ullrich wrote:
>>> In this case, I would prefer this:
>>>
>>> #ifdef WIN32_ONLY_COMPILER
>>> -typedef int pid_t;
>>> +typedef intptr_t pid_t;
>>> #endif
>> That's a change that will have a pretty wide effect. Everything up to
>> now has been pretty low risk, but this worries me rather more. Maybe
>> it's safe, but I'd like to hear others' comments.
> Yeah, it makes me a bit nervous too.
One other thought: even if this is safe for HEAD, I think we could
*not* back-patch it into 9.5, because it would amount to an ABI
break on Windows anywhere that pid_t is used in globally visible
structs or function signatures. (Maybe there are no such places,
but I doubt it.) So we'd need to go with the messy-cast solution
for 9.5.
regards, tom lane