Parser: allow on-demand member decl parsing for other nominal types. #19260

nkcsgexi · 2018-09-11T23:25:36Z

No description provided.

nkcsgexi · 2018-09-11T23:25:46Z

@swift-ci please smoke test

The lexer diagnostics have already been emitted during token skipping.

nkcsgexi · 2018-09-11T23:31:41Z

@swift-ci please smoke test

slavapestov · 2018-09-11T23:33:10Z

Can you add -parse to scale_test and write some tests for this as we discussed?

nkcsgexi · 2018-09-11T23:39:23Z

@slavapestov sure, I'm looking into that.

slavapestov · 2018-09-12T00:09:34Z

lib/Parse/ParseDecl.cpp

@@ -186,7 +186,9 @@ void PersistentParserState::parseMembers(IterableDeclContext *IDC) {
  SourceFile &SF = *IDC->getDecl()->getDeclContext()->getParentSourceFile();
  assert(!SF.hasInterfaceHash() &&
    "Cannot delay parsing if we care about the interface hash.");
+  assert(SF.Kind != SourceFileKind::SIL && "cannot delay parsing SIL");


We need to have a way to keep SILParserState around till delayed parsing, which requires non-trivial work. I don't think there're fundamental blockers there, just we need to keep track more stuff.

Ok, fair enough. Can you add a comment somewhere to that effect? We don’t care about SIL parsing performance much anyway, I was just curious.

nkcsgexi · 2018-09-12T00:18:38Z

@swift-ci please smoke test

nkcsgexi · 2018-09-12T02:18:45Z

@swift-ci please test

swift-ci · 2018-09-12T02:20:35Z

Build failed
Swift Test Linux Platform
Git Sha - 8a28fec56efc38a45141f0b28b9e5a7a9e37d348

swift-ci · 2018-09-12T02:20:40Z

Build failed
Swift Test OS X Platform
Git Sha - 8a28fec56efc38a45141f0b28b9e5a7a9e37d348

…embers.

…parsed.

nkcsgexi · 2018-09-12T21:21:23Z

@swift-ci please smoke test

nkcsgexi · 2018-09-12T21:24:58Z

@slavapestov I've added a scale test. It passes when delayed body parsing is enabled and fails when it's not.

nkcsgexi · 2018-09-12T21:31:04Z

@swift-ci please test

swift-ci · 2018-09-12T21:34:26Z

Build failed
Swift Test OS X Platform
Git Sha - 8cdfddcafc7a95b2cb522ca2c15329277bbc1d7c

swift-ci · 2018-09-12T21:52:19Z

Build failed
Swift Test Linux Platform
Git Sha - 8cdfddcafc7a95b2cb522ca2c15329277bbc1d7c

… tree or verifying parse diagnostics.

nkcsgexi · 2018-09-13T02:42:14Z

@swift-ci please smoke test

nkcsgexi · 2018-09-13T03:10:30Z

@swift-ci please test

nkcsgexi · 2018-09-13T03:11:03Z

@swift-ci Please Test Source Compatibility

swift-ci · 2018-09-13T03:11:04Z

Build failed
Swift Test OS X Platform
Git Sha - 3f03202

swift-ci · 2018-09-13T03:12:11Z

Build failed
Swift Test Linux Platform
Git Sha - 3f03202

nkcsgexi · 2018-09-13T05:28:14Z

All tests passed. Merging

dcci · 2018-09-13T17:03:10Z

@nkcsgexi This seems to have broken a lldb test that runs only in the full run. Mind if we take a look together?

dcci · 2018-09-13T17:15:09Z

Trying to get a reduced testcase.

nkcsgexi · 2018-09-13T17:56:38Z

@dcci Thanks for working on a reduced test case! I think we specifically disabled delayed parsing for REPL input here: https://github.com/apple/swift/blob/master/lib/Parse/ParseDecl.cpp#L3372.

dcci · 2018-09-13T18:07:24Z

In the meanwhile, @vedantk / @fredriss there's a bug in the test suite, as this is a REPL inline test, and should be run as part of the "regular" PR cycle. [cc:ing you as you did some work in the last month in the area]

dcci · 2018-09-13T18:09:47Z

This doesn't reproduce on the REPL from the cmdline, it could be a bug in the SBAPI.

Davides-Mac-Pro:bin davide$ ./lldb --repl
Welcome to Swift version 4.2-dev (LLVM 7b8e4a2b59, Clang 61c26a2ab2, Swift 1e499bf5a9).
Type :help for assistance.
  1> class A {init(a: Int) {}}
  2> class B : A {let x: Int; init() { x = 5 + 5; super.init(a: x) } }
  3> B().x
$R0: Int = 10
  4> extension B : CustomStringConvertible { public var description:String { return "class B\(x) is a subclass of class A"} }
  5> B().description
$R1: String = "class B10 is a subclass of class A"

vedantk · 2018-09-13T18:21:12Z

This test is in test/testcases/repl/subclassing/TestREPLSubclassing.py. It doesn't run because we pass --test-subdir lang/swift to dotest. I'll move it.

vedantk · 2018-09-13T18:22:34Z

(Well, I'll set up a PR to move it; done correctly, it won't pass smoke testing until we fix the underlying issue here.)

dcci · 2018-09-13T19:00:52Z

@vedantk thanks for the explanation!

nkcsgexi added 2 commits September 11, 2018 16:19

parser: on demand parsing of class decl members.

8cac553

parser: on demand parsing of struct decl members.

1798317

nkcsgexi added 3 commits September 11, 2018 16:30

parser: on demand parsing of protocol decl members.

b02c7b1

Parser: disable lexer diagnostics during delayed parsing.

0a03259

The lexer diagnostics have already been emitted during token skipping.

Fix an unused variable warning in no-assertion build.

e0c1fea

nkcsgexi force-pushed the parse-other-type branch from 8a28fec to e0c1fea Compare September 11, 2018 23:31

slavapestov reviewed Sep 12, 2018

View reviewed changes

nkcsgexi added 2 commits September 12, 2018 12:42

Parser: disable delayed member decl parsing when parsing SIL.

7fbd10f

swift-ide-test: input complete test is for REPL.

c0d52fd

nkcsgexi force-pushed the parse-other-type branch from 8cdfddc to c0d52fd Compare September 12, 2018 19:43

nkcsgexi added 3 commits September 12, 2018 14:11

ASTVerifier: avoid verifying IterableDeclContext if it has unparsed m…

033732b

…embers.

scale-test: teach the test driver to test parse only.

53121af

test: add a scale test for the number of IterableDeclContext getting …

3f03202

…parsed.

nkcsgexi force-pushed the parse-other-type branch from b232e91 to 3f03202 Compare September 12, 2018 21:21

FrontendTool: parse all delayed member decls when we're dumping parse…

a7a06d7

… tree or verifying parse diagnostics.

nkcsgexi merged commit c94bc1d into swiftlang:master Sep 13, 2018

vedantk mentioned this pull request Sep 13, 2018

[test] Move Swift REPL tests into lang/swift apple/swift-lldb#891

Closed

Parser: allow on-demand member decl parsing for other nominal types. #19260

Parser: allow on-demand member decl parsing for other nominal types. #19260

Uh oh!

Conversation

nkcsgexi commented Sep 11, 2018

Uh oh!

nkcsgexi commented Sep 11, 2018

Uh oh!

nkcsgexi commented Sep 11, 2018

Uh oh!

slavapestov commented Sep 11, 2018

Uh oh!

nkcsgexi commented Sep 11, 2018

Uh oh!

slavapestov Sep 12, 2018

Choose a reason for hiding this comment

Uh oh!

nkcsgexi Sep 12, 2018

Choose a reason for hiding this comment

Uh oh!

slavapestov Sep 12, 2018

Choose a reason for hiding this comment

Uh oh!

nkcsgexi commented Sep 12, 2018

Uh oh!

nkcsgexi commented Sep 12, 2018

Uh oh!

swift-ci commented Sep 12, 2018

Uh oh!

swift-ci commented Sep 12, 2018

Uh oh!

nkcsgexi commented Sep 12, 2018

Uh oh!

nkcsgexi commented Sep 12, 2018

Uh oh!

nkcsgexi commented Sep 12, 2018

Uh oh!

swift-ci commented Sep 12, 2018

Uh oh!

swift-ci commented Sep 12, 2018

Uh oh!

nkcsgexi commented Sep 13, 2018

Uh oh!

nkcsgexi commented Sep 13, 2018

Uh oh!

nkcsgexi commented Sep 13, 2018

Uh oh!

swift-ci commented Sep 13, 2018

Uh oh!

swift-ci commented Sep 13, 2018

Uh oh!

nkcsgexi commented Sep 13, 2018

Uh oh!

dcci commented Sep 13, 2018

Uh oh!

dcci commented Sep 13, 2018

Uh oh!

nkcsgexi commented Sep 13, 2018

Uh oh!

dcci commented Sep 13, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dcci commented Sep 13, 2018

Uh oh!

vedantk commented Sep 13, 2018

Uh oh!

vedantk commented Sep 13, 2018

Uh oh!

dcci commented Sep 13, 2018

Uh oh!

Uh oh!

dcci commented Sep 13, 2018 •

edited

Loading