Test on macOS-13 (x86) and macOS-14 (arm) #206
Conversation
|
@kali I see that you retried the CI run. I authored this PR hoping that the issues I was having in my CI's setup were specific to my project. I authored a repo and CI hoping it was the runner itself to file against the runner-images repo. I'm not sure where to go from here. |
|
It would make more sense to me to target aarch64-apple-ios when the host is aarch64-apple-darwin, and x86_64-apple-ios when the host is x86_64-apple-darwin. |
|
Maybe try again now? I was able to |
Turns out you're right. Maybe rosetta-2 isn't in the m1 runners or has some quirks. |
| @@ -67,12 +67,12 @@ then | |||
| title "boot a simulator" | |||
| rustup target add x86_64-apple-ios; | |||
| RUNTIME_ID=$(xcrun simctl list runtimes | grep iOS | cut -d ' ' -f 7 | tail -1) | |||
| export SIM_ID=$(xcrun simctl create My-iphone7 com.apple.CoreSimulator.SimDeviceType.iPhone-8 $RUNTIME_ID) | |||
| export SIM_ID=$(xcrun simctl create My-iphone-se com.apple.CoreSimulator.SimDeviceType.iPhone-SE-3rd-generation $RUNTIME_ID) | |||
There was a problem hiding this comment.
Please add a comment on why we are using this specific choice of iPhone to simulate. If iPhone 8 works when running on M1 then maybe this is better done in a separate PR?
.travis.sh
Outdated
| @@ -60,19 +60,52 @@ tests_sequence() { | |||
| ) | |||
| } | |||
|
|
|||
| tests_sequence_ios_sim() { | |||
There was a problem hiding this comment.
It seems like this should have aarch64 or arm64 in the name, and test_sequence should have x86_64 in the name.
There was a problem hiding this comment.
Yeah, I don't love it. The test_sequence function is used for android, qemu and the iOS targets. I think tests_sequence_aarch64_ios_sim might be okay but it's pretty un-DRY.
There was a problem hiding this comment.
Ah, I see. I agree that's a better name.
| && cargo clean \ | ||
| && $CARGO_DINGHY -d $1 -p auto-ios-aarch64-sim test pass \ | ||
| && ! $CARGO_DINGHY -d $1 -p auto-ios-aarch64-sim test fails \ | ||
| && ! $CARGO_DINGHY -d $1 -p auto-ios-aarch64-sim test \ |
There was a problem hiding this comment.
I am new to this tool so I don't know what -p does or why it seems to be needed here. I think maybe we're working around a bug where cargo-dinghy isn't automatically choosing -sim when the host is aarch64, even though it does automatically choose the right thing for x86_64?
if this is indeed a bug, perhaps we should fix that bug so that we can DRY these two functions?
There was a problem hiding this comment.
I was just wondering about that. It's a bug. To verify locally, I ran rustup target remove x86_64-appl-ios (on my m1), then cargo dinghy -d <sim id> test and it failed because dinghy was compiling for x86_64-apple-ios rather than aarch64-apple-ios-sim
There was a problem hiding this comment.
cargo dinghy -d test and it failed because dinghy was compiling for x86_64-apple-ios rather than aarch64-apple-ios-sim
Maybe the best thing to do by default is query the target device and choose the architecture that matches the host, if available, and then choose the default target triple based on what's compatible for that device? If the target device is a real device then it would be aarch64-apple-ios and if it is a simulator then either aarch64-apple-ios-sim or x86_64-apple-ios? And do analogously if the target device is a tvos (simulator) or watchos (simulator).
There was a problem hiding this comment.
Also, I think the kind of improvement I'm suggesting in this thread could be postponed to a future PR. The duplication here is unfortunate but this is good progress that would unblock work on several issues.
The symptom is an indefinite hang, right? Maybe macOS is showing the GUI prompt about installing Rosetta, and that's what's causing the hang? |
* Use newer iphone simulator for CI * Add aarch64-apple-ios-sim for testing
simlay
force-pushed
the
test-on-macos-13-and-macos-14
branch
from
3f866e2 to
90d6a97
Compare
Thanks for the review! I'm glad to hear getrandom (and perhaps ring) is using dinghy! @kali could you review/merge this when you get a chance? |
That can't be the problem because I've run the prebuild x86_64 binary on a macos-14 runner. |
|
Thanks! The -p/--platform is needed because, |
Github now has m1 runners. Using macOS-14 will exclusively run on M1s while macOS-13 will still be on intel.