On Tue, Oct 29, 2019 at 02:32:15AM +0000, Ramsay Jones wrote:
> > + static const struct arch {
> > + const char *name;
> > + int mach;
> > + int bits;
> > + } archs[] = {
> > + { "aarch64", MACH_ARM64, 64 },
> > + { "arm64", MACH_ARM64, 64 },
> > + { "arm", MACH_ARM, 32 },
> > + { "i386", MACH_I386, 32 },
> > + { "m68k", MACH_M68K, 32 },
> > + { "mips", MACH_MIPS64 },
> > + { "powerpc", MACH_PPC64 },
> > + { "ppc", MACH_PPC64 },
> > + { "riscv", MACH_RISCV64 },
>
> I would rather these were explicitly set to 0.
Hmm ... To me, the difference pop out better like so.
The absence of a value is supposed to mean "there is not
a known size for this, it needs to be calculated/guessed".
I find that an explicit 0 conveys this les well.
> > + { "s390", MACH_S390X, 64 },
> > + { "s390x", MACH_S390X, 64 },
> > + { "sparc", MACH_MIPS64 },
>
> Er, I suppose this should be MACH_SPARC64, right? (also 0 init).
Ooops, yes, thank you. Bad copy-paste, bad.
> > + if (bits == 0) {
> > + // guess the size of the architecture
> > + if (!strcmp(suf, "")) {
> > + if (arch_m64 == ARCH_LP32)
> > + bits = 32;
> > + else
> > + bits = 64;
>
> So, this is a 50-50 bet. ;-)
No, not really.
The -m32/-m64 flags are still taken in account, before the
--arch and after it too. If no -m32/-m64 is given at all
then the used size is the one of the native arch (because
arch_m64 is initialized so). I think it's quite natural to
use but it lacks some error checking (e.g.: --arch=arm -m64
will purposely *not* select arm64, aka aarach64, but no
warnings will be issued).
-- Luc
