Using strong mode with a mix of opted in and opted out libraries may result in a static call with bad arguments (missing required named argument) that may crash the VM with an assert fault. Ideally, we should emit code to throw a NoSuchMethodError when the code is executed. For now, we report the error as an informative exception during compilation instead of aborting.
The name of the enclosing function needs to be canonical when creating an instance of _FallThroughError when compiling a switch statement.
Change-Id: I50349fcc5b5036fbdec4697238b1e840de15ec07
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144498
Commit-Queue: Régis Crelier <regis@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
This reverts commit 367ee916d6.
Reason for revert: Build failures
Original change's description:
> [ DartDev ] Generate dartdev.dart.snapshot for runtime builds and place it in the root of the build output directory
>
> VM developers rarely run the create_sdk build as it's significantly
> slower than just building the VM. Since DartDev is coupled to the VM, we
> should build it to ensure that there's no snapshot errors due to old
> snapshots from previous create_sdk builds.
>
> Change-Id: I7626e0f3c791f2a6e69830641d9abf043ec5138f
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144465
> Reviewed-by: Jaime Wren <jwren@google.com>
> Commit-Queue: Ben Konyi <bkonyi@google.com>
TBR=jwren@google.com,bkonyi@google.com,asiva@google.com
Change-Id: I36754f7830e637313a390218d16869a8528f75ff
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144622
Reviewed-by: Ben Konyi <bkonyi@google.com>
Commit-Queue: Ben Konyi <bkonyi@google.com>
VM developers rarely run the create_sdk build as it's significantly
slower than just building the VM. Since DartDev is coupled to the VM, we
should build it to ensure that there's no snapshot errors due to old
snapshots from previous create_sdk builds.
Change-Id: I7626e0f3c791f2a6e69830641d9abf043ec5138f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144465
Reviewed-by: Jaime Wren <jwren@google.com>
Commit-Queue: Ben Konyi <bkonyi@google.com>
In most cases, Code objects in the static calls tables are reachable
through other means. Thus, only add the overhead of walking the static
calls tables when we are in precompiled mode and
--retain-function-objects is disabled.
Also, when deduplicating Instructions objects in Code objects, only
adjust the entry points for the owning function if the Code object is
the current code for that function.
Fixes https://github.com/dart-lang/sdk/issues/41619
Bug: 41619
Change-Id: I51621ccfe1d1c8402088a1c24c56e0a74a02e1ed
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144560
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Symbolic and non-symbolic stack traces had separate ToCString
implementations with the same general structure, but those
implementations diverged enough to cause issues in certain async stack
modes. They are now merged, with the --dwarf-stack-traces-mode flag
checked where appropriate within the merged method.
Also, now non-symbolic stack traces do not include frames for Code with
invisible Function owners unless the flag --show-invisible-frames is
enabled. With this change, pkg/native_stack_traces no longer needs to
guess at which frames corresponding to Dart code should be treated as
internal.
Tested by adding --dwarf-stack-traces version of the tests
in runtime/tests/vm/dart/causal_stacks.
Fixes https://github.com/dart-lang/sdk/issues/41578
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-obfuscate-linux-release-x64-try
Change-Id: I41a887129616c88acd7729492addf7364d95df33
Bug: 41578
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143816
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Clement Skau <cskau@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Never type in the VM isolate causes bootstrapping problems when
null-safety mode is auto-detected. We cannot initialize its type
testing stub without knowing null-safety mode and we cannot patch it
after figuring out null-safety mode because VM isolate is read-only.
The solution is to move Never type (and Never class) out of the VM
isolate and make them per isolate.
Issue: https://github.com/dart-lang/sdk/issues/41206
Change-Id: Iede937aa5becd5f1f7b029db836c25a1b20fbc5c
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144488
Reviewed-by: Régis Crelier <regis@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Added a private RPC, _yieldControlToDDS, which DDS invokes once
connected to the VM service. Once this method is called, the following
happens:
- The VM service stops accepting new web socket requests.
- The VM service enters single client mode, where DDS is the single
client.
DDS can only place the VM service in single client mode when it is the only
VM service client. If other clients are connected, DDS will fail to start and
throw an exception.
The purpose of this change is to simplify state management between the
VM service and DDS. The plan is to delegate the following features to
DDS:
- Service extension management
- Custom client name support
- Isolate resume coordination for multiple clients
Change-Id: I0b3143710d94ee9444a4894e24555f7ed1caf5ad
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/142564
Commit-Queue: Ben Konyi <bkonyi@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
When using C++ value types to represent tagged pointers, we cannot use forward declarations for tagged pointers. This helps to break include cycles when attempting to keep IsXYZ predicates as tagged pointer member functions.
class_id.h also seems like a more natural place for these predicates, which were written before there was a class_id.h.
Change-Id: I0677560a794ed084d10f844606e202feb0c3820a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144321
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
This allows the program to get at the singleton component context
for the currently running component. The component context can
be meaningfully instantiated only once; and since dart doesn't
have a predefined way to wire platform-specific values into
platform-specific code, we instantiate a singleton here, so all
code that is interested can use it.
The users of this functionality will be the dart and flutter
fuchsia runners. See issue
https://github.com/dart-lang/sdk/issues/41523 for details.
Tested:
manual compile for fuchsia, check behavior.
Change-Id: Ib4e15af376ecb932c2d2b28d6f099383d53ccc84
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/136524
Commit-Queue: Filip Filmar <fmil@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
When an isolate dies it has to close all it's open ports. Right now this
is done by walking all ports in the system.
=> O(#num-isolates x #num-ports) complexity
Instead we let each isolate remember it's own open ports: When the
isolate dies we iterate over those ports that particular isolate has
open and close them.
=> O(#num-isolates + #num-ports) complexity
On a many-isolate test of the form:
foo(int n) {
if (n == 1) return 1;
return foo(n-1) + 1;
}
for `foo(50k)` this reduces the runtime from 90 seconds to 30 seconds.
Issue https://github.com/dart-lang/sdk/issues/36097
Change-Id: I3ac68c5334a1d5e8cac47e89e15cbb50c26b65ac
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144261
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
For some reason constexpr seems to need a definition, because the linker
complains about not being able to find the symbol:
% tools/gn.py -ax64 -mdebug -d0
% tools/build.py -ax64 -mdebug
...
ld.lld: error: undefined symbol: dart::FlowGraphBuilderHelper::kPhiSelfReference
>>> referenced by constant_propagator_test.cc:53 (../../runtime/vm/compiler/backend/constant_propagator_test.cc:53)
...
We may also want to consider passing `-fmerge-all-constants` explicitly
in the future - though it is not standards compliant.
See also http://crbug.com/829795.
Change-Id: I1cc125a01751c8a83092b033e80afdcc591c7b92
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144287
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
This CL extracts the hashset implementation from PortMap into
it's own (template) class into port_set.h:PortSet.
This splits the functionality (a nice cleanup in itself) and will
make it re-usable (via template) for a follow-up CL which makes
each message handler aware of it's own open ports.
Issue https://github.com/dart-lang/sdk/issues/36097
Change-Id: I4a56c997414672a4efc9330077110ac5f54d03d0
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144260
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
As part of adding support for brew installations, a bug was introduced
which resulted in the executable directory path not being resolved
correctly if the process was launched via a symlink. This resulted in
the script path passed to the CFE not being fully qualified relative to
the executable location which would cause lookup failures for relative
script paths.
Fixes https://github.com/dart-lang/sdk/issues/41329
Change-Id: I31813ccbc024eb0e4dd6dc265986ba3aefd9898e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144005
Commit-Queue: Ben Konyi <bkonyi@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
The asynchronous waiting mechanism used by the eventhandler thread
on Fuchsia is inherently level-triggered. However, since the upper
layers of the dart:io socket implementation expect edge-triggering,
the Fuchsia eventhandler has to emulate edge-triggering.
Previously, this edge-triggering was emulated for kInEvent and
kOutEvent, but not kCloseEvent. It was not thought that simulating
edge-triggering for kCloseEvent was necessary since the socket would
soon be going down, and that a few redundant messages would be
harmless. This turned out not to be the case. In practice, the
eventhandler thread may flood the Dart thread with kCloseEvent
messages.
This CL adds the edge-triggering emulation for kCloseEvent. The
Dart thread will no longer be flooded with kCloseEvent messages.
Unfortunately, the eventhandler thread may still spin on Fuchsia's
equivalent of POLLHUP signals until the Dart thread destroys the
socket, as these cannot be unsubscribed from.
Fixes b/150902962
Change-Id: I2784220197771508689504156a761842ac164c8e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143888
Reviewed-by: Zichang Guo <zichangguo@google.com>
Commit-Queue: Zach Anderson <zra@google.com>
Adds another interface target to instance calls, pointing to the torn
off method for tearoff getter calls. This is then used to direct the
call to the method extractor for the method.
Increases the fraction of dynamically dispatched instance calls in
Flutter Gallery that are handled by the dispatch table from 95.8% to
98.4%. The remaining 92 IC calls are all on static receiver type
dynamic or Function (34 function calls, 22 method calls, 21 operator
calls, 14 getter calls and 1 setter call).
Change-Id: I0bde1a8271c8563f4631534bd39d1254cd779380
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/142155
Commit-Queue: Aske Simon Christensen <askesc@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Currently, turning on --dwarf-stack-traces can cause the flow graph
compiler to fail in JIT mode. In addition, non-symbolic stack traces can
be useless in JIT mode if recompilation happens and thus there's no
DWARF information for the new code.
Thus, change code that uses the flag to also check for precompiled mode,
so it doesn't change compilation or serialization for JIT snapshots, and
only produce non-symbolic stack traces in the precompiled runtime when
it is on.
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try
Change-Id: Iedae16d1391416de63bc2bdbd19428f607c34fed
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143802
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Virtual addresses are only guaranteed to be valid for the original
shared library that was loaded as a snapshot. This can lead to confusion
if someone attempts to use the virtual address with separately saved
debugging information. This information may not include all the same
sections as the snapshot or include them in the same order and thus may
have different virtual addresses for the instructions sections.
In fact, this happens right now: we don't include a .bss section in the
debugging information, as there's no code in the text sections that
needs relocation. Since the .bss section comes before the .text
sections, this means the (empty) .text sections in the debugging
information have different virtual offsets than the .text sections in
the snapshot.
At this point, we print enough information in non-symbolic stack traces
that the virtual address can be calculated from the absolute address if
truly needed, whether the snapshot was loaded by the OS or by our
non-native ELF loader, so stop printing it to avoid this confusion.
Change-Id: I7949524cc3bb2f510d6812d226b66806939d0da7
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143801
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
This reverts commit 393468dc89.
Reason for revert: This was a trial submit and so reverting it.
Original change's description:
> This is a trial submit of the unfork CL, it will be reverted immediately.
> This commit is being done so we can roll it into Flutter and run some device lab testing to see the effect this CL will have on Flutter.
>
> [BUILD] - Initial CL to unfork the NNBD Dart SDK
>
> - Flips the flag from --nnbd to --no-nnbd so that by default it builds
> the NNBD version
> - using the --no-nnbd flag results in the SDK being built in a
> directory which has the 'Legacy' suffix added to it
> (e.g: out/DebugX64Legacy)
> - the '--enable-experiment=non-nullable' flag still needs to be passed
> in during execution so that CFE runs in that mode. This is different
> from the 'null_safety' flag
>
> Change-Id: I83c813aee2a5b61ad876d8a7b0988705bb24c839
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144082
> Reviewed-by: Siva Annamalai <asiva@google.com>
TBR=asiva@google.com,liama@google.com
Change-Id: I7b50fae71764eceb17893338d1981e40ea2de520
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144083
Reviewed-by: Siva Annamalai <asiva@google.com>
Commit-Queue: Siva Annamalai <asiva@google.com>
This commit is being done so we can roll it into Flutter and run some device lab testing to see the effect this CL will have on Flutter.
[BUILD] - Initial CL to unfork the NNBD Dart SDK
- Flips the flag from --nnbd to --no-nnbd so that by default it builds
the NNBD version
- using the --no-nnbd flag results in the SDK being built in a
directory which has the 'Legacy' suffix added to it
(e.g: out/DebugX64Legacy)
- the '--enable-experiment=non-nullable' flag still needs to be passed
in during execution so that CFE runs in that mode. This is different
from the 'null_safety' flag
Change-Id: I83c813aee2a5b61ad876d8a7b0988705bb24c839
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144082
Reviewed-by: Siva Annamalai <asiva@google.com>
Special sentinel values which are used for lazy initialization can be
stored in variables of any type. In NNBD world Null is no longer a
bottom type as null cannot be assigned into variables of non-nullable
types. The new universal bottom type is Never, so type of sentinel
values is changed from Null to Never.
This also allows to make late fields non-nullable in the type
propagation and avoid extra updates to field guards.
This change also adds special handling of comparisons with sentinel
values into constant propagation as it can constant fold comparisons
of values with different cids.
Fixes https://github.com/dart-lang/sdk/issues/40796
Change-Id: I8266cfba0cc434b78ffa9570c30d434d4380d4af
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143864
Reviewed-by: Liam Appelbe <liama@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This reverts commit 922ea3e9b6 in patchset 1, fix for assertion triggered in https://ci.chromium.org/b/8883214567628884960 in patchset 2, fix for deadlock around symbols table mutex in patchset 4.
Original commit description:
Speed up is achieved by sharing most of the dart code, object store
and class table between isolates in single isolate group. So
instead of bootstrapping isolate from the snapshot, isolate is
initialized by setting pointers to existing data structures already
set up for first isolate, and only few isolate-specific structures (moved
to newly introducted isolate_object_store) are created.
To allow for safe cross-isolate switchable call site, type test cache
mutations additional synchronization via RunWithStoppedMutators(that
relies on safepoints) was added.
Besides switchable call sites, no other mutation to the dart code is
done in AOT, which allows such sharing.
Bug: https://github.com/dart-lang/sdk/issues/37835
Bug: https://github.com/dart-lang/sdk/issues/36097
Change-Id: I655e337198214c9dfacbe76f7852b941b5a7e910
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143462
Commit-Queue: Alexander Aprelev <aam@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Given cyclic phi which participates in strict or equality comparison
with its own argument
y <- phi(x, y);
z <- y == x
CP pass would sometimes fail to converge because it would mark the phi
as unwrapped and then add it to the worklist, which would cause the
phi uses to be revisited: equality would mark the phi again and then
phi will be visited as its own use, which would add it to the worklist
because it is marked and so on.
The fix is to only add marked phis to worklist when reachability of
one of the predecessor blocks changes and that affects whether phi
can be unwrapped or not.
Test is included in a separate CL to make this fix easy to cherry pick.
Fixes https://github.com/flutter/flutter/issues/53903
Change-Id: I9e09a155d7109b7e19ba3ae5aff3fab57dc67fb8
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143581
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
We already have some scratch space in the Instruction for
storing either place id for load optimizer or lifetime
position for register allocator. Some passes also use
the same scratch space.
This change removes pass specific getter/setters from
the Instruction and adds a tag to stored id - to make
sure that different passes don't get confused by
each others ids.
Change-Id: Ie88b6c9cf60604d3ef5d289c23363f18bdb0d014
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143524
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
#if defined(DART_PRECOMPILED_RUNTIME)
#error "AOT runtime should not use compiler sources (including header files)"
#endif // defined(DART_PRECOMPILED_RUNTIME)
Remove #if defined(DART_PRECOMPILED_RUNTIME) from most compiler sources.
Change-Id: Id175c83fdbea38d9d5e1371ff433e3888f2afe8a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143523
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Make them form its own source set (libdart_compiler) and completely exclude
them from AOT runtime targets.
Previously we had some inconsistencies, with some files were using
DART_PRECOMPILED_RUNTIME to fully or partially exclude either contents
from their headers or from implementation, while other files did nothing
and relied on linker to throw their contents away.
This change tries to address this inconsistency.
A follow up change would include a check in most compiler headers which
would prohibit to use them while building AOT runtime.
Change-Id: Ief11b11cbc518b301d3e93fce80580a31bbad151
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/142993
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
This relands commit ff34fd8110,
but puts tree shaking of write-only fields under the flag which is
disabled by default.
So far tree shaking was removing fields which are not used at all.
This change improves tree shaking of fields so fields
which are only written or used as interface targets can be removed.
The following limitations apply:
* Field is not removed if there is a constant object with that field, as
it may impact identity of constant objects which is an observable
behavior.
* Instance field is not removed if it has a non-trivial initializer as
it may have side-effects when executed by constructors.
* Late final fields are not removed, as writing such fields may have
side-effect.
* When field is removed, we may need to introduce an abstract getter
or abstract setter if field is used as a target of an interface call.
If a field was written, then setter would be non-abstract (but empty).
Issue https://github.com/dart-lang/sdk/issues/35310
Change-Id: Iec75b8301892664f4f955a01e5960b17e6620531
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143286
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Before, use of a function in the static calls table of a Code object
would cause it to be retained. Now, like dispatch table use, we only
retain such functions if there are other uses like dynamic function
lookup that needs the function object or if the
--retain-function-objects flag is enabled.
In most modes, --retain-function-objects is enabled by default.
It is only disabled by default in product mode when
--dwarf-stack-traces is enabled, since otherwise the removed
function objects may be needed for debugging platforms like the
Observatory or when creating symbolic stack traces.
Changes on flutter gallery in release mode:
Default:
arm7: isolate: +6.65%, total: +1.30%
arm8: isolate: +6.67%, total: +1.28%
(The increase is due to changing back to all function objects
being retained as the default when using symbolic stack traces.)
With --no-retain-function-objects:
arm7: isolate: -9.95%, total: -1.73%
arm8: isolate: -10.04%, total: -1.77%
(This measures how much dropping static call function objects
when possible affects the size compared to just dropping
function objects for dispatch table entries when possible.)
Bug: https://github.com/dart-lang/sdk/issues/41052
Change-Id: I3c834b14b0c58ccfdbaca3f154df536df29c13ef
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/142146
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Once upon a time, Observatory was built via pub and needed to cache the result of `pub get` to have stable / reproducible builds. Now it is built directly with dart2js and no longer uses any old packages, so the copies of packages brought in via DEPS are fine.
Change-Id: I7b32910570cd21fbf42cae6ac25b57c72a134c81
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143403
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Ben Konyi <bkonyi@google.com>
Martin Kustermann