For constant field initializers, compile the initializer expression
to Kernel, perform local type inference, and perform necessary
rewrites during outline construction. This will support constant
evaluation during separate compilation.
Change-Id: I65fe601595c04c45d586d0bac97c2ade6ab15a90
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104564
Reviewed-by: Dmitry Stefantsov <dmitryas@google.com>
Commit-Queue: Kevin Millikin <kmillikin@google.com>
An internal benchmark via kernel worker of lots of outline
calculations in worker mode with reuse and the incremental compiler
(and lots of dill loaded dependencies) goes from ~160 seconds to ~150 seconds.
Change-Id: I80afa10caacef5e14e569928a1d421bd3e8ab342
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104003
Reviewed-by: Kevin Millikin <kmillikin@google.com>
Commit-Queue: Jens Johansen <jensj@google.com>
So that we can start dogfooding it.
I noticed that we logged too much data, so I reworked what we
log in LibraryContext.
Currently performance will be subpar, due to using content hash as
file's API signature. So, any change to a file, even in a method
body will cause the whole library cycle (and everything that depends
on it) invalidation. I will implement computing API signatures for
parsed unit (hopefully reusing something already existing) in a
following CL.
R=brianwilkerson@google.com, paulberry@google.com
Change-Id: Ifb77e29188484b6784edbaa6a6d5daca6800ef2a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104603
Reviewed-by: Paul Berry <paulberry@google.com>
Commit-Queue: Konstantin Shcheglov <scheglov@google.com>
This is required because otherwise
Object() == null
requires the type signature of Object.operator==(Object o) to be
changed to Object.operator==(Object? o). Which I don't think is the
behavior we want.
Confirmation that this CL is correct has been sent to the language
team. I recommend we land, which will unblock my subtyping CL, and
roll back/readdress later if need be.
Change-Id: I498f9870e7128b2cac3012fff0cb1ab50fcc8df7
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104344
Commit-Queue: Mike Fairhurst <mfairhurst@google.com>
Reviewed-by: Brian Wilkerson <brianwilkerson@google.com>
Reviewed-by: Konstantin Shcheglov <scheglov@google.com>
Reviewed-by: Paul Berry <paulberry@google.com>
Mostly this is adding an abstraction so we can produce the jsonl files that will
allow us to track the status of each test individually.
On a separate CL I'll be moving the tests out of the current unit tests into
their own step in the test matrix.
Change-Id: Id4537db1458949370195124164b120b49a7a526c
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104384
Commit-Queue: Sigmund Cherem <sigmund@google.com>
Reviewed-by: Johnni Winther <johnniwinther@google.com>
So, that we don't have to allocate these lists for each instance.
This makes it cheaper to have serialize more objects.
In a synthetic performance test, where we turn IDL builders into
bytes 100 times, this improves time to bytes from 45 to 34 seconds.
R=paulberry@google.com
Change-Id: Ie8c9a8ed01b4c0d4abeba96ed5e99bd430f9309d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104600
Reviewed-by: Paul Berry <paulberry@google.com>
Commit-Queue: Konstantin Shcheglov <scheglov@google.com>
Exception: also create hard edges for the relation between a type
argument and the corresponding type parameter bound, because that
can't be fixed by a null check.
Change-Id: I71b21ddb11c1bb1060cb6f13efde13855770e2b5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104485
Reviewed-by: Brian Wilkerson <brianwilkerson@google.com>
Reviewed-by: Konstantin Shcheglov <scheglov@google.com>
A number of places in the code will free `Dart_KernelCompilationResult.kernel`
if non-nullptr even if `Dart_KernelCompilationResult.error` is non-nullptr.
Normally the two are mutually exclusive (either `error` is set or `kernel` is set) but
some code frees both if they are non-null.
This issue has caused free() of uninitialized memory which in return
caused some of our benchmarks to break.
Change-Id: Ie65a0c51a32a3616361f3852c13c859e560285b5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104563
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
HttpClientResponse.certificate is documented to be null on a insecure
connection:
> HttpClientResponse's X509Certificate get certificate:
>
> Returns the certificate of the HTTPS server providing the response.
> Returns null if the connection is not a secure TLS or SSL connection.
However, it instead throws an UnsupportedException as of Dart 2 due to an
oversight during refactoring. This change restores the original behavior.
Fixes https://github.com/dart-lang/sdk/issues/31931
Change-Id: I1ef22fe3d7165ec876d22309d8aa4316d94c5576
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/103843
Reviewed-by: Lasse R.H. Nielsen <lrn@google.com>
Commit-Queue: Jonas Termansen <sortie@google.com>
This was a lot of bookkeeping for very little added value. If we want
to experiment with changing the behavior of the migration tool in the
future, we should experimentally change it and see whether the quality
of the migration improves; it's not worth speculatively keeping
multiple behaviors around while the tool is being developed.
Change-Id: I9a59e2626be7e9e84a937e287421abcdc971aba5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104482
Commit-Queue: Paul Berry <paulberry@google.com>
Reviewed-by: Konstantin Shcheglov <scheglov@google.com>
Previously we were creating hard edges for the default values of
optional parameters. This was incorrect; we don't want a hard edge
because we don't want evidence of non-nullability to propagate
upstream from an optional parameter to the const expression that sets
its default value.
Change-Id: I44e12ae96987fd7a9defc538b0704cf76262473f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104500
Reviewed-by: Konstantin Shcheglov <scheglov@google.com>
Rather than use an edge from `always` to ensure that these nodes are
nullable (which is error prone since it could be overridden by an
assertion), we simply place these nodes in the ordinary nullable state
at the time they are created.
As a side effect we can reduce the amount of code we have to plumb the
nullability graph through.
Change-Id: I46cf5b2026aeb2917cb35a7ad0d36a0a0562ecc5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104481
Reviewed-by: Konstantin Shcheglov <scheglov@google.com>
Prior to this change, connection errors such as "Connection refused"
etc, are caught either by some error condition being detected by
the eventhandler thread and that error condition being propagated
to the Dart thread, or by the first read or write after a
"successful" connection failing with an error.
However, posix documentation for non-blocking sockets recommends
either checking getsockopt(SO_ERROR), or calling connect() a
second time following the first write event on a socket after
the first connect call returns EINPROGRESS.
On Fuchsia, this check is mandatory because errors like
"Connection refused" are neither signaled on the socket's
underlying OS handle, nor indicated by error returns from
read() and write().
This change adds the check, which is optional on other platforms,
but mandatory on Fuchsia, to the Dart socket connection loop
after the first write signal is asserted.
Change-Id: I8aeea4665913f79e8138bdd830676ae43066a959
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104263
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Zach Anderson <zra@google.com>
This reverts commit ab6aeaa106.
Revert "[vm/compiler] Speed up the compiler part which deals with kernel reading up in DEBUG mode"
This reverts commit b316210d94.
Reason for revert: regression of snapshot sizes (DNO-599).
Original change's description:
> [vm/kernel] Use GC-tracked ExternalTypedData/TypedDataView for kernel buffers
>
> Until now we often leaked kernel buffers (e.g. hot reload buffers) because various
> objects were referencing ExternalTypedData objects pointing into the middle of
> c-allocated memory. This made it impossible for the GC to determine when the last
> reference is gone.
>
> This CL ensures that the actual buffers are *always* made available via
> ExternalTypedData and any inner pointers into it are created via TypedDataViews.
>
> The embedder guarantees to the free kernel buffers it has provided to:
> - Dart_CreateIsolateFromKernel
> - Dart_LoadScriptFromKernel
> - Dart_LoadLibraryFromKernel
> - Dart_SetDartLibrarySourcesKernel
> on isolate shutdown.
>
> All other kernel buffers will get a finalizer attached, which ensures the
> kernel buffers get freed by the GC once they are no longer referenced:
> - Kernel blobs for expression evaluation
> - Kernel blobs for Hot-Reload
> - Kernel blobs for cc tests
>
> Fixes https://github.com/dart-lang/sdk/issues/33973
> Fixes https://github.com/dart-lang/sdk/issues/36857
> Issue https://github.com/dart-lang/sdk/issues/37030
>
> Change-Id: I1cc410c94c0f4b229413e793728a261afcb10aaf
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/103130
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
> Commit-Queue: Martin Kustermann <kustermann@google.com>
TBR=kustermann@google.com,rmacnak@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: I49715d2400f4a5c8806b7d6a2912b7258f671a0a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104343
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
Auto-Submit: Alexander Markov <alexmarkov@google.com>
OK, this time the patch was applied.
However, there are still compile-time errors as HttpClientResponseCompressionState
was not imported into create_test.dart:
org-dartlang-app:///test/commands/create_test.dart:1176:3: Error: Type 'HttpClientResponseCompressionState' not found.
HttpClientResponseCompressionState get compressionState {
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
org-dartlang-app:///test/commands/create_test.dart:1177:12: Error: Getter not found: 'HttpClientResponseCompressionState'.
return HttpClientResponseCompressionState.decompressed;
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
org-dartlang-app:///test/commands/create_test.dart:1177:12: Error: The getter 'HttpClientResponseCompressionState' isn't defined for the class 'MockHttpClientResponse'.
- 'MockHttpClientResponse' is from 'org-dartlang-app:///test/commands/create_test.dart'.
Try correcting the name to the name of an existing getter, or defining a getter or field named 'HttpClientResponseCompressionState'.
return HttpClientResponseCompressionState.decompressed;
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
This change should fix these errors.
Change-Id: I58b7ca97b2181fe04b4169a11a47173e63c3b865
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104381
Reviewed-by: Todd Volkert <tvolkert@google.com>
Paul Berry