Remove PolymorphicFunctionType #5935

slavapestov · 2016-11-27T09:20:04Z

For those following along at home, this is a continuation of the work in #4849 and earlier commits like da4af64. PolymorphicFunctionType is the contextual type equivalent of GenericFunctionType, a polymorphic function signature but written in terms of archetypes. It didn't make much sense in the first place, and GenericFunctionType had a FIXME comment explaining it should replace PolymorphicFunctionType -- since 2013 :-)

There are three big pieces here:

Ensuring all ValueDecls have an interface type assigned, instead of getInterfaceType() punting to the contextual type if an interface type is not available
Ensuring getType() is not called on polymorphic declarations at all, with getInterfaceType() used everywhere instead
Removing the code that sets the contextual type on polymorphic declarations, which means that PolymorphicFunctionType is now dead and fully subsumed by GenericFunctionType

slavapestov · 2016-11-27T09:20:19Z

@swift-ci Please smoke test

DougGregor · 2016-11-28T05:43:08Z

validation-test/IDE/crashers_fixed/087-swift-declcontext-getparentmodule.swift

@@ -0,0 +1,3 @@
+// RUN: %target-swift-ide-test -code-completion -code-completion-token=A -source-filename=%s
+// REQUIRES: asserts


doesn't require asserts any more

DougGregor

I agree that the uses of SubstitutedType are fragile. In the first case, getMemberSubstitutions() seems like a far simpler and more correct solution.

For accessibility and availability checking, maintaining proper sugar in the Type (with all of the appropriate declaration references as written) would make it far easier. Or, better yet, use the TypeRepr for these checks: then we have all of the source-location information.

DougGregor

Kinda surprised we can do this, because IIRC we haven't done anything to preclude this recursion elsewhere.

DougGregor · 2016-11-28T06:01:34Z

I browsed through everything thus far, and it's looking great. I hadn't realized we were relying on SubstitutedType for semantics in some places, which is... a bit disturbing. Definitely some cleanup to be done there.

slavapestov · 2016-11-28T07:52:14Z

@swift-ci Please smoke test Linux

slavapestov · 2016-11-28T07:52:22Z

@swift-ci Please smoke test Linux

slavapestov · 2016-11-28T07:52:31Z

@swift-ci Please test macOS

slavapestov · 2016-11-28T07:52:42Z

@swift-ci Please test macOS

swift-ci · 2016-11-28T08:09:03Z

Build failed
Jenkins build - Swift Test OS X Platform
Git Commit - 33da79f8524acbc4300543514b2bde18a09a7c26
Test requested by - @slavapestov

slavapestov · 2016-11-28T08:22:39Z

@DougGregor This is now at the point where check-swift-validation passes on my Mac, however it looks like I broke something in SwiftPM, and lldb needs changes as usual. Progress :)

slavapestov · 2016-11-28T23:15:39Z

@swift-ci Please test macOS

swift-ci · 2016-11-28T23:16:28Z

Build failed
Jenkins build - Swift Test OS X Platform
Git Commit - 33da79f8524acbc4300543514b2bde18a09a7c26
Test requested by - @slavapestov

slavapestov · 2016-11-28T23:18:39Z

@swift-ci Please test macOS

slavapestov · 2016-11-28T23:44:54Z

@swift-ci Please smoke test macOS

slavapestov · 2016-11-28T23:46:18Z

@swift-ci Please smoke test macOS

swift-ci · 2016-11-28T23:47:13Z

Build failed
Jenkins build - Swift Test OS X Platform
Git Commit - 035dff452808df850649051f2324fbac606f7126
Test requested by - @slavapestov

slavapestov · 2016-11-29T02:13:01Z

@swift-ci Please smoke test macOS

…tDecl These can contain PolymorphicFunctionTypes, and should no longer be accessed.

Instead of checking if the function has a contextual type, the more accurate predicate here is if it already has a generic environment.

…tDecl

slavapestov · 2016-11-29T10:09:39Z

@swift-ci Please smoke test macOS

slavapestov · 2016-11-29T10:09:48Z

@swift-ci Please smoke test macOS

slavapestov · 2016-11-29T10:10:00Z

@swift-ci Please smoke test Linux

slavapestov · 2016-11-29T10:10:34Z

@swift-ci Please smoke test Linux

slavapestov · 2016-11-29T10:26:38Z

@swift-ci Please smoke test Linux

slavapestov · 2016-11-29T10:26:44Z

@swift-ci Please smoke test macOS

slavapestov · 2016-11-29T10:27:03Z

@swift-ci Please smoke test Linux

slavapestov · 2016-11-29T10:27:18Z

@swift-ci Please smoke test macOS

slavapestov · 2016-11-29T10:33:35Z

@swift-ci Please test Linux

slavapestov · 2016-11-29T10:33:42Z

@swift-ci Please test macOS

swift-ci · 2016-11-29T10:34:39Z

Build failed
Jenkins build - Swift Test OS X Platform
Git Commit - 035dff452808df850649051f2324fbac606f7126
Test requested by - @slavapestov

swift-ci · 2016-11-29T10:34:53Z

Build failed
Jenkins build - Swift Test Linux Platform
Git Commit - f5757e77892c31cb843ba965fe9270da52d1a41e
Test requested by - @slavapestov

DougGregor · 2016-11-29T14:45:15Z

Waahoooooooo!

slavapestov changed the title ~~[Not for checking] Remove polymorphic function type wip~~ [Not for checkin] Remove polymorphic function type wip Nov 27, 2016

slavapestov force-pushed the remove-polymorphic-function-type-wip-november-2016 branch 4 times, most recently from b827fba to fd839e2 Compare November 27, 2016 18:22

slavapestov changed the title ~~[Not for checkin] Remove polymorphic function type wip~~ [Not for checkin] Remove polymorphic function type Nov 27, 2016

BasThomas mentioned this pull request Nov 27, 2016

[48] Issue #48, Dec 1st SwiftWeekly/swiftweekly.github.io#142

Closed

DougGregor reviewed Nov 28, 2016

View reviewed changes

slavapestov force-pushed the remove-polymorphic-function-type-wip-november-2016 branch 2 times, most recently from 4087297 to 33da79f Compare November 28, 2016 07:47

slavapestov force-pushed the remove-polymorphic-function-type-wip-november-2016 branch from 33da79f to 035dff4 Compare November 28, 2016 23:13

slavapestov force-pushed the remove-polymorphic-function-type-wip-november-2016 branch from 035dff4 to 78b4c9e Compare November 28, 2016 23:44

slavapestov force-pushed the remove-polymorphic-function-type-wip-november-2016 branch from 78b4c9e to 92727d5 Compare November 29, 2016 02:12

slavapestov added 13 commits November 29, 2016 03:05

IDE: hasType() => hasInterfaceType()

7b59e75

Serialization: hasType() => hasInterfaceType()

0dd6cf3

AST: Assert if getType() called on AbstractFunctionDecl or EnumElemen…

8272fc1

…tDecl These can contain PolymorphicFunctionTypes, and should no longer be accessed.

IDE: Rework TypeReconstruction to not rely on PolymorphicFunctionType

087fe17

Sema: Tweak a circularity check

82b764d

Instead of checking if the function has a contextual type, the more accurate predicate here is if it already has a generic environment.

AST: Don't call setType() on AbstractFunctionDecls and EnumElementDecls

0f7a455

AST: Assert if setType() called on AbstractFunctionDecl or EnumElemen…

fd0f468

…tDecl

AST: Assert if hasType() called on AbstractFunctionDecl or EnumElemen…

c284bd4

…tDecl

Sema: configureImplicitSelf() no longer needs to return a type

2877006

AST: Remove PolymorphicFunctionType

835472b

AST: Prohibit archetypes in setInterfaceType()

01bb403

AST: Map ParamDecl types out of context when mangling

b43c395

AST: Map closure types out of context when mangling

1809fc8

slavapestov force-pushed the remove-polymorphic-function-type-wip-november-2016 branch from 92727d5 to 1809fc8 Compare November 29, 2016 10:09

Non-deterministic crasher

3749e78

slavapestov changed the title ~~[Not for checkin] Remove polymorphic function type~~ Remove PolymorphicFunctionType Nov 29, 2016

slavapestov merged commit 33c322f into swiftlang:master Nov 29, 2016

		@@ -0,0 +1,3 @@
		// RUN: %target-swift-ide-test -code-completion -code-completion-token=A -source-filename=%s
		// REQUIRES: asserts

Remove PolymorphicFunctionType #5935

Remove PolymorphicFunctionType #5935

Uh oh!

Conversation

slavapestov commented Nov 27, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

slavapestov commented Nov 27, 2016

Uh oh!

DougGregor Nov 28, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

DougGregor left a comment

Choose a reason for hiding this comment

Uh oh!

DougGregor left a comment

Choose a reason for hiding this comment

Uh oh!

DougGregor commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

swift-ci commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

swift-ci commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 28, 2016

Uh oh!

swift-ci commented Nov 28, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

slavapestov commented Nov 29, 2016

Uh oh!

swift-ci commented Nov 29, 2016

Uh oh!

swift-ci commented Nov 29, 2016

Uh oh!

DougGregor commented Nov 29, 2016

Uh oh!

Uh oh!

slavapestov commented Nov 27, 2016 •

edited

Loading

DougGregor Nov 28, 2016 •

edited

Loading