Hi Alex, > On 14 Jan 2021, at 22:13, Alexandre Oliva <ol...@adacore.com> wrote: > > Hello, Olivier, > > On Dec 18, 2020, Olivier Hainque <hain...@adacore.com> wrote: > >> Ping for https://gcc.gnu.org/pipermail/gcc-patches/2020-October/557548.html >> (copied below for convenience), please ? > > I think defining alloca as a macro in analyzer-alloca.h might conflict > with system headers included before or after analyzer-alloca.h > > Say, if a system header #defines alloca(n), GCC might warn when > encountering a different #define in analyzer-alloca.h, included later. > > OTOH, if analyzer-alloca.h is included first, then a later system header > that attempts to declare alloca as a function could be macro-substituted > into declaring __builtin_alloca, which probably wouldn't end well. > > While the alloca macro would render other changes from alloca to > __builtin_alloca unnecessary, there's such an explicit change in the > patch. > > If it were up to me, I'd rather use __builtin_alloca explicitly all > over. I think we already do that in several other tests, for similar > reasons (that some target systems don't have alloca.h or alloca) > > Would you mind if I submitted an alternate patch to do so?
Not at all, thanks for your feedback and for proposing an alternative! Olivier
