https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98055
Martin Sebor <msebor at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
Resolution|--- |WONTFIX
CC| |msebor at gcc dot gnu.org
--- Comment #3 from Martin Sebor <msebor at gcc dot gnu.org> ---
The -Wunused-result warning applies to calls to the GCC alloca built-in
function that allocate space on the stack. To use replacement implementations
of the function the built-in must be disabled by -fno-builtin-alloca. That
also suppresses the warning. This behavior applies in general to other
built-in functions as well, including malloc (and including calls to malloc(0)
which need not allocate any memory and returns null on some implementations).
Diagnosing unportable code, either to other implementations or even to
different configurations of GCC, could be a useful feature, but it's not the
purpose of existing warnings. Their goal is to detect potential problems given
the current target, under the options in effect. As an example, take
-Wpointer-to-int-cast which only diagnoses slicing conversion under the
selected data model (e.g., pointer to int in LP64 but not in ILP32).
Broadening the scope of existing warnings to portability would need to be done
under the control of some sort of a -Wportability option that, to be useful and
not overly noisy, would also need to specify the target(s) to which the code is
intended to be portable. Otherwise, to continue with the example of
-Wpointer-to-int-cast, the warning would have to diagnose every conversion of a
pointer to an integer except intptr_t (which might be useful to some but not
others).
I don't think it would be appropriate to make these implementation choices on a
case-by-case basis and have GCC behave differently for one set of constructs or
functions than another. Doing so would only result in inconsistencies and make
the warnings less useful, not more.
