Avoid macro redefined warning

[catap]

We redefines some macros on FreeBSD and OpenBSD, and it trigers some warning which may lead to failure when -Werror is used.

[WojciechMazur]

I guess it would be better to keep original macro

Suggested change

	#ifdef _XOPEN_VERSION
	#undef _XOPEN_VERSION
	#endif
	#define _XOPEN_VERSION 0
	#ifndef _XOPEN_VERSION
	#define _XOPEN_VERSION 0
	#endif

[catap]

Not sure, because only part of constant are missed and all of them is switched of.

Let me ping @LeeTibbert as well.

[ekrich]

I think there was a discussion about this and we concluded that it was getting ugly so this second approach on top was followed. Perhaps a better approach to using #undef and all this stuff on top is to create a macro to replace the following where needed:

#ifdef _PC_ALLOC_SIZE_MIN
    return _PC_ALLOC_SIZE_MIN;
#else
    return 0;
#endif

You know, when macros get bad make another :-)

[catap]

Ok, I'll take care of this in different way.

But I'd like to finish CI for BSD platform before I move.

[LeeTibbert]

@catap

re: Let me ping @LeeTibbert as well.

I have been offline for awhile and am about to go offline for another longish (weeks) while.
I'll take a look if I can before I go offline again.

My rapid "wet" reading is that section probably long predates me and has to do with
what is and is_not POSIX and what is intended to be allowed in SN posixlib.

I believe, but would have to re-ground myself, that XOPEN is a superset of POSIX and
that the original intent was that only POSIX and possibly XSI were allowed in posixlib.

Compiling with _XOPEN_VERSION macro defined enabled the compilation of "intended-to-be-excluded"
XOPEN functions, so it was #undef'd. I think the #undef was because we do not have absolute control
over the compiler flags passed in; a user could -D_XOPEN_VERSION.

TL;DR - in case events need action before I can do all my evidence collection & tracing.

I think the ugly but cleanest solution is to put explicit OS specific guards around the _XOPEN_VERSION
handling. Again, ugly but gives future devos a sporting chance of understanding the logic chain.

That is, have the existing section for linux, MacOS (Windows?) and a new section
which does the alternate define of this PR for FreeBSD, NetBSD, and known others as they
may apply.

I would end with a "Unsupported OS" #error pragma.

Each OS gets what they are happy with and what has been tested & shown to work.

[catap]

On Fri, 19 Apr 2024 17:28:14 +0200, LeeTibbert wrote: I would end with a "Unsupported OS" #error pragma. Each OS gets what they are happy with and what has been tested & shown to work.

Feel free to push into this branch nessesary changes, and we may discuss it. Right now no rush with it, we have plenty of time to discus the right solution.

…

-- wbr, Kirill