This is to make debugging serialization easier by having a way
to inspect the object graph. Serializer::PrintStack() can be
invoked from the debugger to print the current traversal path.
R=jgruber@chromium.org
Change-Id: Ie67408ade7989dc559904821b3f009fdfe3e459f
Reviewed-on: https://chromium-review.googlesource.com/571219
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46662}
The new message lets us know which directory is involved and what we can
do to fix it.
Bug:
Change-Id: Icfcb92b35b1ef2644649789b1e8473cb6ae50336
Reviewed-on: https://chromium-review.googlesource.com/571702
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46659}
Bug: v8:6000
Change-Id: Ia50108ebbf838e210d95cb268858394e6a66c88d
Reviewed-on: https://chromium-review.googlesource.com/567990
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46658}
It is not clear why we need this BuildTranslation. The intent of
this commit is to check if any tests fail.
Bug:
Change-Id: Ibc8a0914a1483138cfb6234e0fdf87acc0681408
Reviewed-on: https://chromium-review.googlesource.com/567922
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Juliana Patricia Vicente Franco <jupvfranco@google.com>
Cr-Commit-Position: refs/heads/master@{#46657}
Currently every VirtualMemory allocation on 64-bit systems
uses a random 46-bit address hint for ASLR.
This leads to wired page leak on MacOS discovered by Erik Chen (see
crbug.com/700928 and https://chromium-review.googlesource.com/c/557958/):
"The Darwin kernel [as of macOS 10.12.5] does not clean up page directory
entries [PDE] created from mmap or mach_vm_allocate, even after
the region is destroyed. Using a virtual address space that is too large
causes a leak of about 1 wired [can never be paged out] page per call to
mmap(). The page is only reclaimed when the process is killed."
This patch changes VirtualMemory to accept the hint parameter explicitly.
On MacOS the hints are confined to 4GB contiguous region. Algorithm:
- On startup, set heap.mmap_region_base_ to a random address.
- For each mmap use heap.mmap_region_base_ + (random_offset % (4*GB)).
BUG=chromium:700928
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: I2ae6a024e02fbe63f940105d7920b57c19abacc6
Reviewed-on: https://chromium-review.googlesource.com/558876
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46656}
This CL inlines the following builtins into TurboFan
- %MapIteratorPrototype%.next
- %SetIteratorPrototype%.next
following the design that we are using for Array iteration already
(different instance types for the different kinds of iterators). Details
can be found in the relevant design document at:
https://docs.google.com/document/d/13z1fvRVpe_oEroplXEEX0a3WK94fhXorHjcOMsDmR-8
The key to great performance here is to ensure that the inlined code
allows escape analysis and scalar replacement of aggregates to remove
the allocations for the iterator itself as well as the iterator results
and potential key/value arrays in the simple case of a for-of loop (and
by extension also in other constructs that reduce to for-of loops
internally), i.e.:
const s = new Set;
// ... do something with s
for (const x of s) {
// ...
}
Here the for-of loop shouldn't perform any allocations of helper
objects.
Drive-by-fix: Replace the ExistsJSMapWithness in JSBuiltinReducer with a more
general HasInstanceTypeWitness, similar to what's in JSCallReducer. Also
migrate the {Map,Set}.prototype.size getter inlining to the
JSBuiltinReducer, so that everything is in a single place.
R=jgruber@chromium.org
Bug: v8:6344, v8:6571, chromium:740122
Change-Id: I09cb506fe26ed3e10d7dcb2f95ec4415e639582d
Reviewed-on: https://chromium-review.googlesource.com/570159
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46655}
Only allow BindingIdentifier in BindingRestPattern and
ValidReferenceExpression in AssignmentRestPattern.
Also updated to a better, actionable error message.
Bug: v8:6500, v8:6513
Cq-Include-Trybots: master.tryserver.v8:v8_linux_noi18n_rel_ng
Change-Id: Ifaba2f85c20bc20e263267e8c76d50a27075b87d
Reviewed-on: https://chromium-review.googlesource.com/550559
Commit-Queue: Sathya Gunasekaran <gsathya@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46653}
The tail call implementation is hidden behind the --harmony-tailcalls
flag, which is off-by-default (and has been unstaged since February).
It is known to be broken in a variety of cases, including clusterfuzz
security issues (see sample Chromium issues below). To avoid letting
the implementation bitrot further on trunk, this patch removes it.
Bug: v8:4698, chromium:636914, chromium:724746
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng;master.tryserver.v8:v8_linux_noi18n_rel_ng
Change-Id: I9cb547101456a582374fdf7b1a3f044a9ef33e5c
Reviewed-on: https://chromium-review.googlesource.com/569069
Commit-Queue: Adam Klein <adamk@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46651}
Commit f37d726435 limited inner function
parsing to function declarations, to allow function expressions to
be eagerly-compiled if the parser discovered that they are immediately
invoked. But it's not only declarations that won't be immediately invoked:
methods and accessors are in the same boat, and should be treated the same.
This patch reverses the logic to exclude function expressions from inner
lazy treatment, thus making both function declarations and methods/accessors
inner-lazy-parseable.
Bug: v8:5501
Change-Id: I71a57667e52fcb917362ba629667c4c84ae29011
Reviewed-on: https://chromium-review.googlesource.com/569180
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46650}
We only optimize functions which are already compiled, so there is no need
to ensure baseline in pipeline.cc, and since ast-graph-builder doesn't
do inlining, there are no other uses.
BUG=v8:5203, v8:6409
Change-Id: I830c8868d50363f61193a96d9a5774e059a1af0e
Reviewed-on: https://chromium-review.googlesource.com/570033
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46649}
These are no longer necessary since we only have one optimizing compiler.
Also avoid changing --turbo-filter when --no-opt is set, and instead
explicitly check both the FLAG_opt and FLAG_turbo_filter in
GetOptimizedCode to check whether optimization is disabled.
BUG=v8:6408
Change-Id: I0948f788e8ff111c08022270d86c22f848da300a
Reviewed-on: https://chromium-review.googlesource.com/568484
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46648}
The WhichPowerOf2 function now takes arbitrary integer arguments, and
hence subsumes the WhichPowerOf2_64 function.
R=mstarzinger@chromium.org
Change-Id: I0692a17bbbe0074425d7c330342bc6fa34fb9d5c
Reviewed-on: https://chromium-review.googlesource.com/568495
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46646}
Found issue with BuildChangeEndianness. There is difference between load and store case.
Load depends primarily on MachineRepresentation, while store depends on ValueType.
TEST=wasm-spec-tests/tests/endianness
TEST=wasm-spec-tests/tests/memory
TEST=wasm-spec-tests/tests/memory_trap
Bug:
Change-Id: I437f611107daad2f425a67fcc068e874822e0f58
Reviewed-on: https://chromium-review.googlesource.com/558882
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Ivica Bogosavljevic <ivica.bogosavljevic@imgtec.com>
Cr-Commit-Position: refs/heads/master@{#46645}
In my previous CL, I completely disabled the processing of inlining stacks
when generating code comments. This CL restores the feature at least for
runs without concurrent recompilation.
In other words, if you use --code-comments and want to see nice source positions,
you must also use --no-concurrent-recompilation.
Bug: v8:6048
Change-Id: Ief7e52a4046cb74e02b1783dbace19534093c588
Reviewed-on: https://chromium-review.googlesource.com/570021
Commit-Queue: Georg Neis <neis@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46643}
This also makes sure 6.0 and 6.1 have the same constants exposed
in v8.h.
Bug: v8:6592
R=bmeurer@chromium.org
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: I3afc0e5e4495594c76229555aab148ac78388f80
Reviewed-on: https://chromium-review.googlesource.com/569618
Commit-Queue: Yang Guo <yangguo@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46642}
In debug mode (SLOW_DCHECK), Handle<T>::cast accesses the object to
check its type. Obviously we can no longer do that now that we run
on a background thread.
NOTE: I think there are other parts of TF that suffer from the same
problem. I will look into fixing those as well.
Bug: v8:6048, v8:6590
Change-Id: I9abfdf30f1899cdb0c8b9078b0cf71463d608251
Reviewed-on: https://chromium-review.googlesource.com/570054
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46641}
This is the x64 implementation of the CL
https://codereview.chromium.org/2763593002
Original message:
[wasm][arm] Add an additional stack check for functions with big frames.
Stack overflow checks are typically implemented as part of the TurboFan
graph of a function. This means that the stack check code is executed
after frame construction. When a frame is too big, though, there may not
be enough space on the stack anymore to throw the stack overflow
exception after frame construction. With this CL we do an additional
stack check before frame construction for functions with big frames.
As discussed offline with mstarzinger, I do this change currently only
for WebAssembly.
This CL contains only the changes for arm. I will do the other platforms
in separate CLs.
R=titzer@chromium.org
Bug: v8:6318
Change-Id: Id4a8ea3ee76c37132e86a7c4e5d05d3bd86df44a
Reviewed-on: https://chromium-review.googlesource.com/565562
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Reviewed-by: Ben Titzer <titzer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46640}
... that have computed name and/or require home object.
This should give us the opportunity to implement initialization
of name and home object values in a stub.
Bug: v8:6459
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: I47a1a2c185e120e86c793733cce737811f895291
Reviewed-on: https://chromium-review.googlesource.com/512802
Reviewed-by: Ben Titzer <titzer@chromium.org>
Reviewed-by: Andreas Rossberg <rossberg@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46638}
Removes the --ignition flag which is now on by default. Adds a
--stress-fullcodegen flag which enables running all functions supported
by fullcodegen to be compiled by fullcodegen.
This will enable moving parser internalization later when we are not
stressing fullcodegen or compiling asm.js functions.
BUG=v8:5203, v8:6409, v8:6589
Change-Id: I7fa68016d4e734755434ec0b4e749ef65ffa7f4e
Reviewed-on: https://chromium-review.googlesource.com/565569
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46635}
Additionally, drive-by add a test inspired by a worrying TODO
in the parser.
Cq-Include-Trybots: master.tryserver.v8:v8_linux_noi18n_rel_ng,master.tryserver.v8:v8_linux64_msan_rel,v8_linux64_tsan_rel,v8_linux_arm64_dbg,v8_linux_gc_stress_dbg,v8_linux_noi18n_rel_ng,v8_linux_nosnap_rel
Change-Id: I693a858a638753f8c1091f2f9086710f628b40dd
Reviewed-on: https://chromium-review.googlesource.com/553677
Reviewed-by: Georg Neis <neis@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Commit-Queue: Daniel Ehrenberg <littledan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46634}
This reverts commit 5648aad553.
Reason for revert: Compile error on mips:
https://build.chromium.org/p/client.v8.ports/builders/V8%20Mips%20-%20builder/builds/10732
Original change's description:
> [wasm] Don't store global handles in the interpreter
>
> Storing global handles in the interpreter is dangerous, because the
> global handles are strong roots into the heap. The interpreter itself is
> referenced from the heap via a Managed. Hence the interpreter keeps the
> instance alive, while the instance keeps the Managed alive. So the GC
> will never collect them.
>
> This CL refactors this to only store the handle to the instance object
> while executing in the interpreter, and clearing it when returning.
> It also removes the cache of import wrappers, as it should not be
> performance critical, but keeps lots of objects alive. If it turns out
> to be performance critical, we will have to reintroduce such a cache
> stored in the WasmDebugInfo object.
>
> R=titzer@chromium.org
> CC=ahaas@chromium.org
>
> Bug: chromium:610330
> Change-Id: I54b489dadc16685887c0c1a98da6fd0df5ad7cbb
> Reviewed-on: https://chromium-review.googlesource.com/567058
> Reviewed-by: Ben Titzer <titzer@chromium.org>
> Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#46629}
TBR=titzer@chromium.org,clemensh@chromium.org
Change-Id: Ifadfb885f937f37bb3eab4732a97f20ff40c2583
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: chromium:610330
Reviewed-on: https://chromium-review.googlesource.com/569962
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46630}
Storing global handles in the interpreter is dangerous, because the
global handles are strong roots into the heap. The interpreter itself is
referenced from the heap via a Managed. Hence the interpreter keeps the
instance alive, while the instance keeps the Managed alive. So the GC
will never collect them.
This CL refactors this to only store the handle to the instance object
while executing in the interpreter, and clearing it when returning.
It also removes the cache of import wrappers, as it should not be
performance critical, but keeps lots of objects alive. If it turns out
to be performance critical, we will have to reintroduce such a cache
stored in the WasmDebugInfo object.
R=titzer@chromium.orgCC=ahaas@chromium.org
Bug: chromium:610330
Change-Id: I54b489dadc16685887c0c1a98da6fd0df5ad7cbb
Reviewed-on: https://chromium-review.googlesource.com/567058
Reviewed-by: Ben Titzer <titzer@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46629}
When disassembling some instructions we would print `r12`, which can be
confusing when the rest of the disassembly consistently uses `ip`.
Bug:
Change-Id: Id4cfc5805ef102a0845cdaaa8390e618ee981b19
Reviewed-on: https://chromium-review.googlesource.com/570038
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Pierre Langlois <pierre.langlois@arm.com>
Cr-Commit-Position: refs/heads/master@{#46628}
There is just one version now, called IsPowerOfTwo. It accepts any
integral type.
There is one slight semantical change: Called with kMinInt, it
previously returned true, because the argument was implicitly casted to
an unsigned. It's now (correctly) returning false, so I had to add
special handlings of kMinInt in machine-operator-reducer before calling
IsPowerOfTwo on that value.
R=mlippautz@chromium.org,mstarzinger@chromium.org,jgruber@chromium.org,ishell@chromium.org,yangguo@chromium.org
Change-Id: Idc112a89034cdc8c03365b778b33b1c29fefb38d
Reviewed-on: https://chromium-review.googlesource.com/568140
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46627}
This CL changes the USE macro to accept more than one parameter.
This is particularly interesting for calling a method on each type in a
parameter pack, as in:
template<typename... T>
void foo(T&&... ts) {
USE(do_something(ts)...);
}
Drive-by fix: Allow to pass arbitrary types to USE, including
references. This might prevent a copy for pass-by-value.
R=ishell@chromium.org, tebbi@chromium.org
Also-by: tebbi@chromium.org
Change-Id: I544e83bb996aaa638e7512295973dd3e742254bc
Reviewed-on: https://chromium-review.googlesource.com/567507
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46626}
This is for backwards ABI compatibility to version 6.0.
R=bmeurer@chromium.org
Bug: v8:6592
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: I20f3fa79324511c6e393ddaad50880f22ef6df77
Reviewed-on: https://chromium-review.googlesource.com/569758
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46625}
It's ok that the instance of the called code object is different from
the caller instance. This happens if one instance calls an exported
function of another instance.
R=ahaas@chromium.org
Bug: chromium:739768
Change-Id: I6afa8332a9b33fe32e9332cdca573053f058421d
Reviewed-on: https://chromium-review.googlesource.com/568494
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46624}
It used to be module-relative, but since we switched to decoding
sections individually, the offsets would be section-relative.
This CL fixes this by adding the buffer offset to the reported
locations.
R=ahaas@chromium.org
Change-Id: I27240832e5cbd625d6b952678503252017dd83f3
Reviewed-on: https://chromium-review.googlesource.com/568488
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46623}
The shortcuts ensure that field type generalization is properly
propagated in the transition graph.
Bug: chromium:738763
Change-Id: Id701a6f95ed6ea093c707fbe0bac228f1f856e9f
Reviewed-on: https://chromium-review.googlesource.com/567992
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46622}
The DCHECK is supposed to check that the read in the next line does not
read out of bounds. Hence we have to use {pc} and not {pc_}.
I accidentally triggered this while writing an unrelated test case, but
now fail to reproduce.
R=ahaas@chromium.org
Change-Id: Ia3fa9a9874866dc14180b22b9af526459fbda495
Reviewed-on: https://chromium-review.googlesource.com/568487
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46621}
This correctly types values loaded via {LoadElement} nodes from arrays
of HOLEY_DOUBLE_ELEMENTS elements kind as {Type::NumberOrHole}. Even
though "the hole" is still encoded as a tagged NaN, the type system
still needs to consider it as a potential hole value.
R=bmeurer@chromium.org
TEST=mjsunit/regress/regress-crbug-736575
BUG=chromium:736575
Change-Id: Ib869284900a4affb2ddaa1d2a96df9443dba6921
Reviewed-on: https://chromium-review.googlesource.com/567180
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46620}
The instruction sequence consumes a significant amount of memory, so it
should be deallocated as early as possible. This CL separates the zone
which is used by the code generator from the zone which is used by the
instruction selector. Thereby we can delete the instruction selector
zone, which contains the instruction sequence, already after
AssembleCode, and not only after FinalizeCode.
For WebAssembly this means that the instruction sequence gets deleted
on the background tasks already and does not stay alive until the
main threads deletes it.
R=bmeurer@chromium.org, neis@chromium.orgCC=mtrofin@chromium.org
Change-Id: I090a2140ca05ae3bcc66268b0eddb08846fea690
Reviewed-on: https://chromium-review.googlesource.com/566831
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Reviewed-by: Mircea Trofin <mtrofin@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46619}
Enabling background code assembly broke --code-comments. This CL "fixes"
it for now by simply skipping the printing of inlining stack info.
R=bmeurer@chromium.org
Bug: v8:6048
Change-Id: I8b9ea629c7c302198484510a7234014bfbfd890b
Reviewed-on: https://chromium-review.googlesource.com/569163
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46617}
Previously we enforced that all lengths for ArrayLike objects must
be within Smi range, but all negative numbers should actually be first
converted to +0.
Bug: chromium:740372
Change-Id: If50de9ce0eeb7cb09e14b8e8803f434350d00508
Reviewed-on: https://chromium-review.googlesource.com/566867
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46615}
Insertion into a collection changes the map because of the addition of
the hash value property. Check the root map, not the current map.
Fixes: https://github.com/nodejs/node/issues/14139
Change-Id: Iabcea5337323b9b6deffa1a06892c1cb749f2065
Reviewed-on: https://chromium-review.googlesource.com/566833
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46614}
