Inconsistent (system-type 'arch) on BC x86_64 FreeBSD?

[LiberalArtist]

This is Racket BC's definition for SCHEME_ARCH on FreeBSD:

racket/racket/src/bc/sconfig.h

Lines 346 to 359 in 89a8b08

	# define SCHEME_OS "freebsd"
	# if defined(__i386__)
	# define SCHEME_ARCH "i386"
	# define REGISTER_POOR_MACHINE
	# define MZ_USE_JIT_I386
	# define MZ_TRY_EXTFLONUMS
	# if defined(__FreeBSD_kernel__)
	# define ASM_DBLPREC_CONTROL_87
	# else
	# define FREEBSD_CONTROL_387
	# endif
	# elif defined(__amd64__)
	# define SCHEME_ARCH "amd64"

It appears to be the only circumstance when amd64 is used instead of x86_64 for (system-type 'arch) and (system-library-subpath). In particular, Racket CS uses x86_64 on ta6fb:

racket/racket/src/cs/rumble/system.ss

Lines 74 to 83 in 89a8b08

	(define arch-symbol
	(case (reflect-machine-type)
	[(a6osx ta6osx
	a6nt ta6nt
	a6le ta6le
	a6ob ta6ob
	a6nb ta6nb
	a6fb ta6fb
	a6s2 ta6s2)
	'x86_64]

I think the BC definition should probably changed, but I'm not a BSD user myself.

[LiberalArtist]

I just came across this in the docs, which, while it doesn't strictly promise not to return 'amd64, does give a stronger implication to expect 'x86_64 than I'd realized:

racket/pkgs/racket-doc/scribblings/reference/runtime.scrbl

Lines 34 to 36 in 89a8b08

	In @indexed-racket['arch] mode, the result is a symbol representing an
	architecture. Possible results include @racket['x86_64], @racket['i386],
	@racket['aarch64], @racket['arm] (32-bit), and @racket['ppc] (32-bit).

[mflatt]

I agree that it should be x86_64.