llvm dev - May 2017 - Why not alias aarch64-apple-ios to arm64-apple-ios

Hi,

when calling clang with —target=arm64-apple-ios, the triple
results in arm64-apple-ios7.0, however when using
—target=aarch64-apple-ios, it ends up being unknown-apple-macosx10.4.0

What is the reasoning behind this? On the other hand the
lovely autotools (with a recent config.{guess,sub} recognize)
aarch64-apple-ios, but not arm64-apple-ios.  Thus resulting in
mismatches between the autoconf target/build/host and clangs
—target flag, and by that extension llvms -triple flags?

Cheers,
 Moritz

On 21 May 2017 at 03:47, Moritz Angermann via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> when calling clang with —target=arm64-apple-ios, the triple
> results in arm64-apple-ios7.0, however when using
> —target=aarch64-apple-ios, it ends up being unknown-apple-macosx10.4.0
>
> What is the reasoning behind this?
There's no real reasoning, it's a bug in the Clang driver. It's
persisted this long because it's mostly just a minor nuisance for
developers.

The official Apple tools use "-arch arm64" instead of specifying
-target directly; -arch only works when the default target is
Darwin-based to put Clang into "Apple mode" as it were. But it does
sort of explain why arm64 works better than aarch64, which is the
preferred name everywhere else.

Cheers.

Tim.