Tweak some macro definition on Windows

[Girgias]

Does this seem reasonable @cmb69 ?
The change for HAVE_GRP_H is that because instead of doing an #ifdef we used #if making this feature disabled on Windows although it should have been enabled.

[KalleZ]

Reading over at the MSDN docs, it does sound to me like wspiapi.h is always available as of Windows 2000, which is way below our minimum required version of Windows, which I guess could mean we could simplify the IPV6 support check at the same time in config.w32 a little.

For HAVE_GRP_H we supply our own minimalistic header in win32/grp.h, so even if we mark HAVE_GRP_H as 1, any preprocessor checks needs to still check for PHP_WIN32 to conditionally include that file instead, so I think it might be better left as it is to not give a false illusion that it exists. When I implemented the nice() functionality, I opted to define HAVE_NICE as its a single function, not a header and I think we should follow the same principle here. On a similar note, I think we do the same for unistd.h too (provide our own, don't define HAVE_UNISTD_H).

Looking at the 3 files that uses HAVE_GRP_H all have a conditional check for Windows inside it, so yes it should enable some functionality, but I think the checks should be changed instead with an #elif defined(PHP_WIN32).

[Girgias]

Reading over at the MSDN docs, it does sound to me like wspiapi.h is always available as of Windows 2000, which is way below our minimum required version of Windows, which I guess could mean we could simplify the IPV6 support check at the same time in config.w32 a little.

For HAVE_GRP_H we supply our own minimalistic header in win32/grp.h, so even if we mark HAVE_GRP_H as 1, any preprocessor checks needs to still check for PHP_WIN32 to conditionally include that file instead, so I think it might be better left as it is to not give a false illusion that it exists. When I implemented the nice() functionality, I opted to define HAVE_NICE as its a single function, not a header and I think we should follow the same principle here. On a similar note, I think we do the same for unistd.h too (provide our own, don't define HAVE_UNISTD_H).

Looking at the 3 files that uses HAVE_GRP_H all have a conditional check for Windows inside it, so yes it should enable some functionality, but I think the checks should be changed instead with an #elif defined(PHP_WIN32).

Seems reasonable I'll amend the PR with a change to elif.

[cmb69]

It seems to me that setting HAVE_GRP_H on Windows is currently absolutely useless, since none of the respective functions are available, and I consider it somewhat unlikely that they ever will. Why do we even have win32/grp.h in the first place?

[nikic]

Based on @cmb69's comment, it looks like the correct action is the other way around: Keep HAVE_GRP_H=0 and remove the win32/grp.h header, which is currently unused (included only in dead paths).