Switch the Windows ARM64 builds to use MSVC. Clang disagrees with itself about handling of small structs in variadic functions, allowing splitting between the last argument register and the stack as the callee but not as the caller.
TEST=ci
Cq-Include-Trybots: luci.dart.try:vm-ffi-android-debug-arm-try,vm-ffi-android-debug-arm64c-try,vm-ffi-android-release-arm-try,vm-ffi-android-release-arm64c-try,vm-ffi-qemu-linux-release-arm-try,vm-linux-release-arm64-try,vm-mac-debug-arm64-try,vm-mac-release-arm64-try,vm-win-debug-arm64-try,vm-win-release-arm64-try,vm-ffi-qemu-linux-release-riscv64-try,vm-linux-debug-ia32-try,vm-linux-release-ia32-try,vm-win-release-ia32-try,vm-linux-debug-x64-try,vm-linux-release-x64-try,vm-mac-debug-x64-try,vm-mac-release-x64-try,vm-win-debug-x64-try,vm-win-release-x64-try
Bug: https://github.com/dart-lang/sdk/issues/52644
Bug: https://github.com/dart-lang/sdk/issues/53829
Change-Id: I2fd6c40620a885479f11bb8528ca1e9df3948a2f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/331209
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
stringref target won't be used any time soon (probably ever). To help
with build times and avoid keeping it updated remove it for now.
Change-Id: I0df33b7ab2e19bae5090e8ea32ea6a3dc3751652
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/334881
Reviewed-by: Aske Simon Christensen <askesc@google.com>
Commit-Queue: Ömer Ağacan <omersa@google.com>
The Dart2JS team has agreed that this mode of compilation is no longer worth investing in at this time and its existence adds some overhead to other feature work so it is worth fully removing. In the future we may revisit this mode of compilation. Below is some more context for any future exploration in this space.
What didn't work with modular analysis:
- current modular analysis was computing impacts, which were dense (50% of the size of kernel)
- using it moved work to a modular phase, and cut Phase1 in half however end-to-end time was not better
- data overhead was very high
- it made it much harder to maintain invariants throughout the pipeline: the data is tightly coupled with the kernel AST, making it hard to make late modifications to the AST.
How to potentially make it better:
- make the data much more sparse
- make the data more independent from the kernel AST so that transformations are not breaking
- reduce the critical path in a more substantial way.
Note: We retain and ignore the commandline flags used for modular analysis in order to avoid breaking build pipelines that were passing them. We may remove these at a later date.
Change-Id: If574ce2358280ab5fedd89c62665328601e72e22
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/333360
Reviewed-by: Mayank Patke <fishythefish@google.com>
Reviewed-by: Sigmund Cherem <sigmund@google.com>
Commit-Queue: Nate Biggs <natebiggs@google.com>
Build one package with three elf_test_runner components: one for each of the binaries that the test harness invokes on the target. (Though currently only tests that use "dart" are setup.)
Create includable CML shards for the capabilites required by the AOT and JIT VMs.
Fold test_runner's fuchsia_cfv2.dart back into fuchsia.dart.
TEST=ci
Bug: https://github.com/dart-lang/sdk/issues/38752
Change-Id: I7bd8d43d184cbcb11903c7aed77ce31fb30cb894
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/332451
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Zijie He <zijiehe@google.com>
Reviewed-by: Alexander Thomas <athom@google.com>
Also renames the test configurations for x64 to include the architecture.
Cq-Include-Trybots: luci.dart.try:pkg-linux-debug-try,pkg-linux-release-try,pkg-mac-release-arm64-try,pkg-mac-release-try,pkg-win-release-try
Change-Id: I76e259a059e9fe16f4799d65f2bb5380a9c6fb94
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/330028
Reviewed-by: William Hesse <whesse@google.com>
Commit-Queue: Alexander Thomas <athom@google.com>
Many tests under `pkg`, like `pkg/js_ast` and `pkg/analyzer_cli` are
meant to run as unit tests in the Dart VM. These were also being run in
dart2js and causing flakiness because they took a long time to build on
our CQ.
Locally some of these tests compile in under 5s, but we see the
compile-time vary among the bots a lot. I've observed cases close to 30s
and some that reach the 1-minute timeout cutoff. Flakiness data show
that this is happening between 5-7% of the runs, especially on linux
bots.
This CL stops running these tests on the dart2js configuration by
changing which subset of the `pkg` suite is provided to dart2js. It also
adjusts the .status file to skip a few suites for convenience on local
testing only. This allows developers locally to run `test.py -n... pkg`
without worrying about filtering out skipped suites.
Addendum: after discussions on the CL we decided not to cmpletely drop
coverage of js_ast tests on JS to ensure it can be used in browser
environments in the future. To support this we made made the tests
cheaper by removing the dependency on package:test. They will now only
be run in a single configuration, which is all we need for this purpose.
We can reevaluate and remove this if we continue to see timeouts.
Change-Id: Idf0dbdd37e412ef71ba117ec979cb1e52585c431
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/330704
Reviewed-by: Stephen Adams <sra@google.com>
Commit-Queue: Sigmund Cherem <sigmund@google.com>
- Update dart2wasm script to accept a `-O` argument. When passed it also
passed `--inlining-limit 10` to dart2wasm (unless it's already passed
by the caller) and runs wasm-opt on the output, with the same
arguments as `compile_benchmark`.
- Add new test configurations to run tests with optimized binaries.
Example configuration: `dart2wasm-linux-optimized-d8`.
Change-Id: Ia7bd8d9c99d3a4baecb72b92b2ba38c0e92807c6
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/326620
Commit-Queue: Ömer Ağacan <omersa@google.com>
Reviewed-by: Aske Simon Christensen <askesc@google.com>
This change created a CFv2 build target and provided two
implementations of the FuchsiaEmulator, so that I can continually work
on both CFv2 build and test script integration independently.
The v2 target can be built with
```
tools/build.py --os fuchsia --mode release fuchsia_ffi_test_package_cfv2
```
And v2 FuchsiaEmulator implementation can be triggered with
FUCHSIA_CFV2 environment variable.
Bug: #38752
Change-Id: I31936a2ca967fbfeb2bc5628e2f005aef6762687
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/322583
Reviewed-by: Alexander Thomas <athom@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Zijie He <zijiehe@google.com>
- Updates test infra to use the assets built by the new "ddc" build
targets.
- The old "dartdevc" targets are still present to avoid breaking
golem when this change lands.
- Old targets are still used when packaging assets for the SDK.
- Golem and the SDK packaging will be updated in following changes.
Change-Id: I1926e0c86833c812f5ed8355d7cd4df8740d79ee
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/315224
Commit-Queue: Nicholas Shahan <nshahan@google.com>
Reviewed-by: Sigmund Cherem <sigmund@google.com>
- Introduces a new output subdirectory `gen/utils/ddc` where the
new build targets output files. This is intended to make the
migration to the new assets easier since landing this change
shouldn't immediately break any dependencies.
- Enables building the canary and stable assets at the same time.
- Changes more names in our workflow from "dartdevc" to "ddc".
- Updates the ddc-canary-linux-chrome and
ddc-canary-linux-chrome-unsound configurations to use the
new assets.
The new outputs appear in directories that are more consistently
named, (no more sound vs kernel confusion). They are named in
preparation for the eventual deletion of the unsound targets
without any lingering cruft to be cleaned up in the sound
directories.
The new structure is:
```
gen/utils/ddc/
|- canary
| |- pkg
| |- sdk
|- canary_unsound
| |- pkg
| |- sdk
|- stable
| |- pkg
| |- sdk
|- stable-unsound
|- pkg
|- sdk
```
The 'pkg' and 'sdk' directories all contain outputs with the same
files names, each compiled in the respective build modes
(sound/unsound and stable/canary).
Change-Id: I66822ebf03bba487b6d359a8e0aa818b9e7b6bef
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/313081
Reviewed-by: Sigmund Cherem <sigmund@google.com>
Commit-Queue: Nicholas Shahan <nshahan@google.com>
This reverts commit f827eb3a78.
Reason for revert: The tests being run, and the results being compared to, are changing dramatically, causing all sorts of test results going from [status] -> Skipped and from New Test -> [status].
Something is really going wrong.
Original change's description:
> Balance tests equally across shards.
>
> The sharded test runner invocations are now passed the previous results
> which contains the test timing, which are used to simulate how long each
> shard would take to run. The shards are now balanced as evenly as
> possible on a test level, taking multiple cores into account.
>
> Sharded tests are now run starting with the slowest test first, such
> that extremely long running tests finish as early as possible. This
> behavior ensures the cores are saturated and can be padded with fast
> tests near the end, rather than waiting for a few slow tests to complete
> while the rest of the system is idle.
>
> The algorithm works very well whenever it's able to accurately predict
> the time to run shards. In a number of cases, the model doesn't quite
> reflect reality and the data, which makes it fairly imperfect but still
> reasonably good. I think a second order feedback loop might kick in once
> it reorders the tests across shards and the test timing data reflects
> the new test timings.
>
> Multitests are no longer always sent to the same shard, since the data
> isn't available at the moment, and the change as-is speeds up the test
> running considerably.
>
> The front end unit test suites currently ignore the feature as there are
> no benefits yet to improving those quick shards.
>
> Upgrade the language version to 3.0.0 so patterns can be used and fix
> a mixin not being a mixin.
>
> Fixes: b/291585137
> Change-Id: I3cc1b1d96038d5b46e836b091e299097717c226c
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/314081
> Reviewed-by: William Hesse <whesse@google.com>
> Commit-Queue: Jonas Termansen <sortie@google.com>
Change-Id: I233e4bfa6d6ecf0cea4f97c1e47f1635f7b9040c
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/315060
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Auto-Submit: William Hesse <whesse@google.com>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
The sharded test runner invocations are now passed the previous results
which contains the test timing, which are used to simulate how long each
shard would take to run. The shards are now balanced as evenly as
possible on a test level, taking multiple cores into account.
Sharded tests are now run starting with the slowest test first, such
that extremely long running tests finish as early as possible. This
behavior ensures the cores are saturated and can be padded with fast
tests near the end, rather than waiting for a few slow tests to complete
while the rest of the system is idle.
The algorithm works very well whenever it's able to accurately predict
the time to run shards. In a number of cases, the model doesn't quite
reflect reality and the data, which makes it fairly imperfect but still
reasonably good. I think a second order feedback loop might kick in once
it reorders the tests across shards and the test timing data reflects
the new test timings.
Multitests are no longer always sent to the same shard, since the data
isn't available at the moment, and the change as-is speeds up the test
running considerably.
The front end unit test suites currently ignore the feature as there are
no benefits yet to improving those quick shards.
Upgrade the language version to 3.0.0 so patterns can be used and fix
a mixin not being a mixin.
Fixes: b/291585137
Change-Id: I3cc1b1d96038d5b46e836b091e299097717c226c
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/314081
Reviewed-by: William Hesse <whesse@google.com>
Commit-Queue: Jonas Termansen <sortie@google.com>
The purpose of the wasm_js_compatibility target is to facilitate experiments with a JS compatibility mode for Dart2Wasm. Initially, we're just going to focus on typed data, but this will give us a place to experiment with moving List and String to JS as well.
In addition, someday down the road we hope to experiment with two additional compatibility changes:
1) Exclusively using double for all Dart numbers
2) Allowing undefined to flow as null.
The two major benefits of this approach are:
1) Much faster JS interop
2) To make it easier to bring up Dart2JS applications on Dart2Wasm
The only downside will be access overhead on the Wasm side, but the JS builtins proposal could potentially bring us close to parity with Wasm builtins someday.
Tested: Wasm specific trivial refactor.
Change-Id: I2c09426b6999507c1de6e584e9bc7072a088bda9
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/313240
Commit-Queue: Joshua Litt <joshualitt@google.com>
Reviewed-by: Ömer Ağacan <omersa@google.com>
Reviewed-by: William Hesse <whesse@google.com>
Add test.py --default-suites that includes all the default suites in
addition to the ones explicitly requested, so the test matrix can run
co19 together with the default suites in one sharded test step.
Bug: b/290617138
Change-Id: I5dd5d1aaf3b1ee38adf88c6e9ee6ec13d97fe1ce
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/313567
Reviewed-by: William Hesse <whesse@google.com>
The non-sharded tests takes 10 minutes and each of the sharded tests
takes 14 minutes. It's faster to shard immediately (costing one bot
more) and concurrently run the local tests.
Fix end to end dart2js test that times out when sharded and run outside
a directory called sdk.
Bug: b/290617138
Change-Id: If71f0d301edf565c9f15847098320106ca383635
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/312983
Reviewed-by: William Hesse <whesse@google.com>
The try builder is among the fastest commonly used builders at 21
minutes average, which is less than the current 25 minute best case
goal. Free up 2 of 10 shards to be available for other work during peak
load, which will slow it down by 2 or 3 minutes and finish at the same
time as the other try builders, resulting in an overall faster commit
queue experience.
Bug: b/290617138
Change-Id: Iea2ba71248fa0d3c21f1d80b540b55ce810d96af
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/313565
Reviewed-by: William Hesse <whesse@google.com>
Ben Konyi