Parser creates a FunctionState for default ctors, which affects the
next_function_is_likely_called logic. PreParser needs to match that logic, so
that Parser and PreParser agree about which functions are skippable.
BUG=v8:5515, chromium:773576
Change-Id: I96cb6f5aa68e74389a863355f70a34693a2d1329
Reviewed-on: https://chromium-review.googlesource.com/712579
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48511}
The catch variable is a special VAR-mode variable which is not in a declaration
scope. Normally creating such a variable is not possible with DeclareVariable,
but Parser bypasses it by calling DeclareLocal directly (which doesn't have the
hoisting check).
PreParser used to cut corners and declare the catch variable as a LET-mode
variable to prevent hoisting.
But since LET and VAR variables behave differently when deciding whether they
block sloppy block function hoisting, that approach doesn't fly.
BUG=v8:5516,chromium:771474
Change-Id: Ic6f5f4996416c9fa59132725c8b0b6b570c72f48
Reviewed-on: https://chromium-review.googlesource.com/700634
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48308}
Track whether the async arrow func parameter list was simple or not; the
information is already there, we just didn't pipe it through correctly. It's
needed by PreParser so that it can create the correct Scope structure.
Implementation notes:
- I could've used async_classifier for transmitting the "is_simple" bit, but I
made it explicit (it would be unnecessary to use ExpressionClassifier for
this, as we're not classifying any expressions) instead.
- I'm also moving work (setting parameter_list.is_simple) from Parser to
ParserBase, and adding a DCHECK in Parser to assert that the work was indeed
already done.
BUG=v8:5516,chromium:765532
Change-Id: Iacf91b150d1b57996544b5e64baa7d91ac134445
Reviewed-on: https://chromium-review.googlesource.com/674695
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48132}
PreParser and Parser didn't agree whether a generator in a sloppy block is a
sloppy block function or not, and thus the data generated by PreParser was
inconsistent with what the Parser wanted to restore.
BUG=v8:5516, chromium:760116
Change-Id: I0fd3c267691b8afd63a1336774769caf551c143e
Reviewed-on: https://chromium-review.googlesource.com/642886
Reviewed-by: Adam Klein <adamk@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47727}
Makes ClusterFuzz start fuzzing with the flag on.
BUG=v8:5516
Change-Id: Ia80f7d22f12fe25efb226102a896e8b0e3537947
Reviewed-on: https://chromium-review.googlesource.com/610000
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47366}
In some cases, PreParser cannot replicate the Scope structure created by
Parser. It happens esp. with arrow function parameters, since the relevant
information is already lost by the time we figure out it's an arrow function.
In these cases, PreParser should bail out of trying to create data for skipping
inner functions.
Implementation notes:
- The arrow function case is more fundamental; the non-arrow case could be
hacked together somehow if we implemented tracking is_simple for each param
separately; but now that it's possible to bail out consistently from both
cases, I don't think the is_simple complication is worth it.
- The added mjsunit test cases are based on the test262 test cases which exposed
the problem.
- cctest/preparser/PreParserScopeAnalysis was exercising similar cases, but the
problem didn't show up because the function parameters didn't contain
skippable functions. Those test cases have been repurposed for testing the
bailout.
- Extra precaution: the bailout tests are in a separate file, to guard from the
bug that a bailout case results in bailing out of *all* data creation, which
would make all skipping tests in the same file useless.
BUG=v8:5516
Change-Id: I4324749a5ec602fa5d7dc27647ade0284a6842fe
Reviewed-on: https://chromium-review.googlesource.com/599849
Reviewed-by: Adam Klein <adamk@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47170}
This way, each lazy function needs to handle only the data relevant to
itself. This reduced data handling overheads.
Other changes:
1) Don't deserialize the data; once it's on the heap, it can stay there. Lazy
function compilation is only done in the main thread.
2) Separate ProducedPreParsedScopeData and ConsumedPreParsedScopeData. It's clearer, because:
- The data looks fundamentally different when we're producing it and when we're
consuming it.
- Cleanly separates the operations we can do in the "producing phase" and in the
"consuming phase".
Bug: v8:5516
Change-Id: I6985a6621f71b348a55155724765624b5d5f7c33
Reviewed-on: https://chromium-review.googlesource.com/528094
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46347}
In the failing case (see test), the loop variable (which should be context
allocated) is in a hidden scope, so we need to save and restore data for hidden
scopes too.
The !is_hidden() check was overly limiting - NeedsScopeData already handles the
"hidden leaf scope" case which is the one we want to avoid.
(Btw, this also means that the previous assumption "variables in hidden scopes
are not context allocated" was wrong.)
BUG=v8:5516
Change-Id: I1c6116654b19ef0cfd64e8a743b46af683a9fcd5
Reviewed-on: https://chromium-review.googlesource.com/544938
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Daniel Vogelheim <vogelheim@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46136}
let f = function g() { ... } declares "g" inside the function. This
CL makes the preparser declare it too, and saves + restores the scope data for
it.
BUG=v8:5516
Change-Id: Id4c64f446d30f5252038cfb0f0f473b85ba24a9b
Reviewed-on: https://chromium-review.googlesource.com/544816
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Daniel Vogelheim <vogelheim@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46133}
- Enable aggressive lazy inner funcs (make non-declaration funcs lazy, ie let f =
function() { ... } when --experimental-preparser-scope-analysis is on.
- Turn on variable tracking for lazy top level functions: this makes their inner
functions skippable.
- Test fix for an testing bug uncovered by this work: when restoring the data
for the relevant scope, don't assume it's the outermost scope for which we
have data.
- Fix: if we abort lazy parsing a function, we shouldn't produce any data for
it.
BUG=v8:5516
Change-Id: I0606fbabb5886dc57dbb53ab5f3fb894ff5d032e
Reviewed-on: https://chromium-review.googlesource.com/518165
Reviewed-by: Daniel Vogelheim <vogelheim@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45615}
The feature is not quite ready for getting ClusterFuzzed.
BUG=v8:5516
Change-Id: I90a42f950727c8ecf46cb2987c9a459b2ba1f5a7
Reviewed-on: https://chromium-review.googlesource.com/480400
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Daniel Vogelheim <vogelheim@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44693}
Unfortunately, this test cannot test that a function was really skipped (i.e.,
not parsed).
BUG=v8:5516
Change-Id: I8db5027d2216a95cc012ceae8e17554095cc1d4f
Reviewed-on: https://chromium-review.googlesource.com/457037
Reviewed-by: Daniel Vogelheim <vogelheim@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44615}
