Re: [PATCH] tests: move -D_FORTIFY_SOURCE to CPPFLAGS

[Roland McGrath <roland at hack dot frob dot com>]

I think the change is fine.  Using CPPFLAGS rather than CFLAGS for the
cpp flags is just perfectly sensible pedanticism in its own right.
Indeed, it is undesireably subtle that this is the way to make sure
that you override flags set in a less-specific CPPFLAGS variable.
What that really means is that we have an implicit presumption that
choice of CPPFLAGS vs CFLAGS for a given flag will be made with equal
pedanticism in all the different places flags can be added (configure
time, sysdeps Makefiles, per-file variables).  Arguably it would be
better if we used an ordering of:
 $(CPPFLAGS-global) $(CFLAGS-global) $(CPPFLAGS-specific) $(CFLAGS-specific)
for all the various kinds of global vs specific.  But I think the
relative simplicity of all the kinds of *CPPFLAGS* variables being put
into CPPFLAGS (resp. CFLAGS, CXXFLAGS) trumps that idea.

The notion of adding another variable TESTFLAGS seems to me like a
false hope for simplicity, that actually is just adding yet more
complexity.  It's yet another variable whose ordering relationships
with the others is important and has to be remembered.


Thanks,
Roland