Guard against duplicate definitions of __EXTENSIONS__

[Bill-Sommerfeld]

For whatever reason, g++ on illumos predefines __EXTENSIONS__.

gnucash defines it in four places; it's guarded with an #ifdef in config.h, but three other files define it unconditionally in a way that clashes with the predefinition, resulting in compile errors like:

/z/ws/gnucash-upstream/libgnucash/engine/gnc-date.cpp:28:9: error: "__EXTENSIONS__" redefined [-Werror]
   28 | #define __EXTENSIONS__
      |         ^~~~~~~~~~~~~~
<built-in>: note: this is the location of the previous definition

The fix is to change the other three files to match config.h.

[jralls]

Meh. __EXTENSIONS__ seems to be a SunOS/Solaris thing. Let's just get rid of them.

[jralls]

I just saw your wiki account application explaining that openindiana is an open source solaris. The reason for __EXTENSIONS__ is explained in standards(5) under "feature test macros".

[Bill-Sommerfeld]

I understand why __EXTENSIONS__ exists in the system headers (looks like it was also used on hp-ux). What I'm not sure of yet is why g++ (but not gcc) is always setting it.

Dropping the #define __EXTENSIONS__ from these three files and from config.h makes sense if it can land in CFLAGS instead. I'll revise this pull request accordingly..

[jralls]

It will also need to come out of CMakeLists.txt.

Test thoroughly, it's not going to have anything to do with modern Unix that I can test.

[jralls]

I'll revise this pull request accordingly..

Ping?

[Bill-Sommerfeld]

Will be getting back to this shortly (had it mostly ready when something else captured my attention..)

[jralls]

Thanks!