Re: popen and pclose redefinitions causing many warning in Windows build

Heikki Linnakangas <hlinnakangas(at)vmware(dot)com> writes:
> On 05/08/2014 08:01 AM, Michael Paquier wrote:
>> Since commit a692ee5, code compilation on windows is full of warnings
>> caused by the re-definitions of popen and pclose:

> Hmm. Does the MinGW version of popen() and system() do the quoting for
> you? If we just #ifdef the defines, then we will not use the wrappers on
> MinGW, which would be wrong if the quoting is needed there. If it's not
> needed, then we shouldn't be compiling the wrapper functions in the
> first place.

Another problem, if we do need the wrappers on mingw, is that the
"#undef" commands in system.c will presumably result in the wrong
things happening in the wrapper functions, since the platform needs
us to use their macros there.

The simplest workaround I can think of is to change the stanza in
port.h to be like

#ifndef DONT_DEFINE_SYSTEM_POPEN
#undef system
#define system(a) pgwin32_system(a)
#undef popen
#define popen(a,b) pgwin32_popen(a,b)
#undef pclose
#define pclose(a) _pclose(a)
#endif

and then have system.c do #define DONT_DEFINE_SYSTEM_POPEN before
including postgres.h. This is pretty grotty, but on the other hand
it'd remove the need for the #undef's in system.c.

regards, tom lane