Add _pointerBitWidth platform condition #41534
Conversation
|
CC: @xwu |
|
@swift-ci test |
|
Fantastic to see this resurrected. As you deduced correctly, I never did have the time to return to this (most particularly, the proposal process), but it would be lovely to have and I'm glad to see you pick it up. |
lib/Basic/LangOptions.cpp
Outdated
| @@ -356,6 +364,25 @@ std::pair<bool, bool> LangOptions::setTarget(llvm::Triple triple) { | |||
| break; | |||
| } | |||
|
|
|||
| // Set the "_native_word_size" platform condition. | |||
There was a problem hiding this comment.
Nit:
| // Set the "_native_word_size" platform condition. | |
| // Set the "_pointerBitWidth" platform condition. |
But also could you document where these particular target arches come from so we can keep things in sync in the future?
There was a problem hiding this comment.
I'm not entirely sure what you mean by "where these particular target arches come from". It uses the same targets as the rest of the code in this file.
drexin
force-pushed
the
wip-word-size-condition
branch
from
87f9ad4 to
61f3518
Compare
drexin
changed the title
drexin
force-pushed
the
wip-word-size-condition
branch
from
61f3518 to
ddd5479
Compare
|
@swift-ci smoke test |
| @@ -7,7 +7,7 @@ | |||
| let i: Int = "Hello" | |||
| #endif | |||
|
|
|||
| #if arch(arm) && os(Android) && _runtime(_Native) && _endian(little) | |||
| #if arch(arm) && os(Android) && _runtime(_Native) && _endian(little) && _pointerBitWidth(_32) | |||
There was a problem hiding this comment.
This seems pointless if it's not possible to use other pointer sizes on platforms like Android armv7.
There was a problem hiding this comment.
I'm not sure I understand. In what cases would you want other pointer sizes on armv7?
There was a problem hiding this comment.
My point is that isn't arch(arm) here redundant with _pointerBitWidth(_32), ie the former always implies the latter?
There was a problem hiding this comment.
Oh, but the whole point of this test is to check that all of these are properly set. If any of them wasn't, the test would fail.
There was a problem hiding this comment.
OK, so the goal is to just lump together all compile-time properties of this platform, even though some of them are redundant so nobody would ever use some combos of those checks together, got it.
There was a problem hiding this comment.
Not quite. It is to ensure that if we are compiling for armv7-unknown-linux-androideabi, all of the expected conditions hold. We have these tests for all targets.
drexin
force-pushed
the
wip-word-size-condition
branch
from
ddd5479 to
6881b41
Compare
|
@swift-ci test |
drexin
requested review from
ahoppen,
bnbarham,
DougGregor,
rintaro,
hborla,
slavapestov and
xedin
as code owners
|
Wooooo thanks @drexin for making this a reality! |
* Add _pointerBitWidth platform condition * Fixes after rebasing --------- Co-authored-by: Xiaodi Wu <[email protected]>
The original PR was #14413, but was seemingly abandoned. Proposal is forthcoming.
rdar://107070345