[Sema] Fix errors due to implicitly deleted copy constructor. #292

dcci · 2015-12-07T05:04:01Z

No description provided.

dcci · 2015-12-07T05:04:25Z

This fixes a build failure on FreeBSD with recent clang.

nadavrot · 2015-12-07T05:27:53Z

@dcci What is the error message on FreeBSD?

Can you use "=default " as the body?

dcci · 2015-12-07T05:31:42Z

Hi Nadav, I tried with =default but I keep getting the same error (which is weird). Anyway, looking at it again I'm not entirely convinced it's the right fix.
As per your request, here's the error message:

/exps/swift/swift/include/swift/Sema/TypeCheckRequestKinds.def:29:1: error: call to implicitly-deleted copy constructor of 'swift::TypeCheckRequest'
TYPE_CHECK_REQUEST(QualifiedLookupInDeclContext, DeclContextLookup)
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/exps/swift/swift/include/swift/Sema/TypeCheckRequest.h:154:10: note: expanded from macro 'TYPE_CHECK_REQUEST'
return TypeCheckRequest(TypeCheckRequest::Request, payload);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/exps/swift/swift/include/swift/Sema/TypeCheckRequest.h:79:5: note: copy constructor of 'TypeCheckRequest' is implicitly deleted because field 'Payload' has a deleted copy constructor
} Payload;
^
/exps/swift/swift/include/swift/Sema/TypeCheckRequest.h:73:5: note: explicitly defaulted function was implicitly deleted here
PayloadType(const PayloadType &) = default;
^
/exps/swift/swift/include/swift/Sema/TypeCheckRequestPayloads.def:37:28: note: copy constructor of 'PayloadType' is implicitly deleted because variant field 'InheritedClauseEntry' has a non-trivial
copy constructor
TYPE_CHECK_REQUEST_PAYLOAD(InheritedClauseEntry, std::pair<llvm::PointerUnion<TypeDecl *, ExtensionDecl *>, unsigned>)
^
/exps/swift/swift/include/swift/Sema/TypeCheckRequest.h:76:17: note: expanded from macro 'TYPE_CHECK_REQUEST_PAYLOAD'
VA_ARGS PayloadName;
^

Thanks!

Davide

dcci · 2015-12-07T05:36:24Z

Clang seems to DTRT:

http://en.cppreference.com/w/cpp/language/copy_constructor
the implicitly-declared or defaulted copy constructor for class T is defined as deleted if any of the following conditions are true:
T is a union and has a variant member with non-trivial copy constructor

tbartelmess · 2015-12-07T12:31:25Z

@dcci What version of clang are you using to compile?

jckarter · 2015-12-07T15:58:42Z

Does FreeBSD use an outdated Clang by default? There was another PR floating around to make us explicitly use clang35 or later IIRC.

tbartelmess · 2015-12-07T16:06:52Z

@jckarter

$ clang++ --version
FreeBSD clang version 3.4.1 (tags/RELEASE_34/dot1-final 208032) 20140512
Target: x86_64-unknown-freebsd10.2
Thread model: posix
$

gribozavr · 2015-12-07T18:39:03Z

Pull request #169 discusses that. I think Clang 3.4 is just too old, and there will be more issues once we get past this one.

dcci · 2015-12-07T18:44:52Z

This is clang 3.7 on FreeBSD -CURRENT and not 3.4 on stable-10

gribozavr · 2015-12-07T18:49:23Z

@dcci I know. What I was suggesting in #169 is that we use "clang35" on stable-10 as a special case, and "clang" on FreeBSD versions we don't special case.

dcci · 2015-12-08T01:26:26Z

So, I can confirm my suspicion that this fix isn't quite right.
While this is able to build swift the generated executable crashes while compiling the standard library:
Top of the stack:
Core was generated by `swift'.
Program terminated with signal SIGBUS, Bus error.
#0 llvm::PointerIntPair<swift::Type, 1u, bool, llvm::PointerLikeTypeTraitsswift::Type >::getPointer (this=0x89f07d99f836d848) at /exps/swi
ft/llvm/include/llvm/ADT/PointerIntPair.h:77
77 reinterpret_cast<void*>(Value & PointerBitMask));
(gdb) bt
#0 llvm::PointerIntPair<swift::Type, 1u, bool, llvm::PointerLikeTypeTraitsswift::Type >::getPointer (this=0x89f07d99f836d848) at /exps/swi
ft/llvm/include/llvm/ADT/PointerIntPair.h:77
#1 0x0000000000435c75 in swift::TypeLoc::getType (this=0x89f07d99f836d848) at /exps/swift/swift/include/swift/AST/TypeLoc.h:66
#2 0x00000000014bb7c0 in swift::IterativeTypeChecker::isResolveInheritedClauseEntrySatisfied (this=0x7fffffff1578, payload=...) at /exps/swi
ft/swift/lib/Sema/ITCDecl.cpp:85
#3 0x00000000014ba0cf in swift::IterativeTypeChecker::isSatisfied (this=0x7fffffff1578, request=...) at /exps/swift/swift/include/swift/Sema
/TypeCheckRequestKinds.def:37
#4 0x00000000014ba3b9 in swift::IterativeTypeChecker::satisfy (this=0x7fffffff1578, request=...) at /exps/swift/swift/lib/Sema/IterativeType
Checker.cpp:70

@DougGregor who originally wrote this code and may have insights on the right fix.

GuillaumeBibaut · 2015-12-08T11:36:59Z

I'm getting the same compilation error than @dcci.
I'm on FreeBSD 10.2-RELEASE with clang 3.7 installed and used by the building script. Mainly:

apple/swift/include/swift/Sema/TypeCheckRequestKinds.def:54:1: error: call to implicitly-deleted copy constructor of 'swift::TypeCheckRequest'
TYPE_CHECK_REQUEST(ResolveTypeDecl, TypeDeclResolution)
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/home/yom/Projects/apple/swift/include/swift/Sema/TypeCheckRequest.h:154:10: note: expanded from macro 'TYPE_CHECK_REQUEST'
  return TypeCheckRequest(TypeCheckRequest::Request, payload);                 \
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/home/yom/Projects/apple/swift/include/swift/Sema/TypeCheckRequest.h:79:5: note: copy constructor of 'TypeCheckRequest' is implicitly deleted because field 'Payload' has a deleted copy constructor
  } Payload;
    ^
/usr/home/yom/Projects/apple/swift/include/swift/Sema/TypeCheckRequest.h:73:5: note: explicitly defaulted function was implicitly deleted here
    PayloadType(const PayloadType &) = default;
    ^
/usr/home/yom/Projects/apple/swift/include/swift/Sema/TypeCheckRequestPayloads.def:37:28: note: copy constructor of 'PayloadType' is implicitly deleted because variant field 'InheritedClauseEntry' has a non-trivial copy constructor
TYPE_CHECK_REQUEST_PAYLOAD(InheritedClauseEntry, std::pair<llvm::PointerUnion<TypeDecl *, ExtensionDecl *>, unsigned>)
                           ^
/usr/home/yom/Projects/apple/swift/include/swift/Sema/TypeCheckRequest.h:76:17: note: expanded from macro 'TYPE_CHECK_REQUEST_PAYLOAD'
    __VA_ARGS__ PayloadName;
                ^

dcci · 2015-12-10T21:15:35Z

I think that a way to fix this might be implementing copy-constructor and operator= so that it picks the correct underlying constructor depending on the type. Something like this, modulo mistakes. Thoughts?

diff --git a/include/swift/Sema/TypeCheckRequest.h b/include/swift/Sema/TypeCheckRequest.h
index b5767e5..38c9fb0 100644
--- a/include/swift/Sema/TypeCheckRequest.h
+++ b/include/swift/Sema/TypeCheckRequest.h
@@ -114,7 +114,45 @@ public:
   }

 #include "swift/Sema/TypeCheckRequestPayloads.def"
-
+
+  TypeCheckRequest(const TypeCheckRequest &T) { *this = T; }
+
+  TypeCheckRequest& operator=(const TypeCheckRequest &T) {
+    switch (getPayloadKind(getKind())) {
+    case PayloadKind::Class:
+      Payload.Class = T.Payload.Class;
+      break;
+    case PayloadKind::Enum:
+      Payload.Enum = T.Payload.Enum;
+      break;
+    case PayloadKind::InheritedClauseEntry:
+      // XXX: FIXME
+      new (&Payload.InheritedClauseEntry) std::pair<llvm::PointerUnion<TypeDecl *, ExtensionDecl *>, unsigned>();
+      Payload.InheritedClauseEntry = T.Payload.InheritedClauseEntry;
+      break;
+    case PayloadKind::Protocol:
+      Payload.Protocol = T.Payload.Protocol;
+      break;
+    case PayloadKind::DeclContextLookup:
+      new (&Payload.DeclContextLookup) DeclContextLookupInfo();
+      Payload.DeclContextLookup = T.Payload.DeclContextLookup;
+      break;
+    case PayloadKind::TypeResolution:
+      // XXX: FIXME
+      new (&Payload.InheritedClauseEntry) std::tuple<TypeRepr *, DeclContext *, unsigned>();
+      Payload.TypeResolution = T.Payload.TypeResolution;
+      break;
+    case PayloadKind::TypeDeclResolution:
+      Payload.TypeDeclResolution = T.Payload.TypeDeclResolution;
+      break;
+#if 0
+    default:
+      llvm_unreachable("unknown kind");
+#endif
+    }
+    return *this;
+  }
+
   /// Determine the kind of type check request.
   Kind getKind() const { return TheKind; }

…zed. Added a check for whether the escape use is due to an UncheckedTakeEnumDataAddrInst, which accesses the memory to grab the .Some part of an Optional during “?.” or “!”. In that case, use the generic used-before-initialized diagnostic, since it isn’t actually because of closure capture. Added test cases, including specifically testing that capturing in a closure in order to unwrap the optional is still ok.

Update the README.md instructions for utils/update-checkout option. Should read `--clone-with-ssh`.

Update README.md to show the correct utils/update-checkout option.

Fix typo in UnicodeNormalization.cpp

Update getting started instructions

[build-script] cmark c flags should derive from cmark build type

…umTag() witness We were checking if the tag was <= numPayloadCases, not <. Fix this and swap the order of the basic blocks so that the no-tag case is the fall through for consistency with getEnumTag().

We weren't actually successfully mutating ParentType, so fix that, and also ensure we set ParentType together with the rest of the state. Discovered while writing resilient enum tests.

Move these to SILDeclRef, maybe not the best place but a good home for now. Factor out a new requiresForeignToNativeThunk() function, which cleans up some code duplication introduced by the following patch: 478e1c7 This is a small step towards consolidating duplicated logic for figuring out method dispatch semantics and emitting curry thunks.

This time, the issue is that TypeNullifier skips bodies of multi-statement closures. However, ExprRewriter will type happily pass them on to typeCheckClosureBody(). This could trigger assertions. Fix this by skipping type checking of multi-statement closures when diagnosing. There seems to be a minor QoI regression in some test cases that already looked pretty dodgy and/or had FIXMEs. However I think its worth fixing a crash.

Fix typos

This caused a linker failure on linux for some stdlib unit tests. rdar://problem/23849414

[SR-184] Better diagnose error via unwrapping optional before initialized.

It's useful for debugging.

dcci · 2015-12-16T19:00:21Z

I have a new pull request for this (on a different branch) that I'll submit shortly.

Revert "linux: update header used for `major` macro"

[Sema] Fix errors due to implicitly deleted copy constructor.

f4e946c

gregomni and others added 16 commits December 10, 2015 23:26

Fix typos

e799d3e

cmark c flags should derive from cmark build type

37840db

Update README.md

7730b49

Update the README.md instructions for utils/update-checkout option. Should read `--clone-with-ssh`.

Update getting started instructions

4dad7c7

Merge pull request #427 from djwbrown/djwbrown-swift-readme

43e4f17

Update README.md to show the correct utils/update-checkout option.

Merge pull request #433 from nuts23/patch-1

dfe1544

Fix typo in UnicodeNormalization.cpp

Merge pull request #429 from garriguv/readme-clone-typo

c15e2a3

Update getting started instructions

Merge pull request #422 from halfnelson/patch-1

3a9c52b

[build-script] cmark c flags should derive from cmark build type

IRGen: Fix incorrect conditional in multi-payload destructiveInjectEn…

2adabe5

…umTag() witness We were checking if the tag was <= numPayloadCases, not <. Fix this and swap the order of the basic blocks so that the no-tag case is the fall through for consistency with getEnumTag().

Sema: Fix assertion failure if EnumElementPattern is re-typechecked

0269e02

We weren't actually successfully mutating ParentType, so fix that, and also ensure we set ParentType together with the rest of the state. Discovered while writing resilient enum tests.

Interpreter tests for resilient enums

c31b209

Merge pull request #380 from jutaz/bugfix/typos-take-3

3e94baa

Fix typos

Fix some compiler_crashers that regressed recently

2a34a62

eeckstein and others added 5 commits December 11, 2015 10:00

[StackPromotion] fix the declaration of swift_buffer[De]AllocateOnStack.

fb9f564

This caused a linker failure on linux for some stdlib unit tests. rdar://problem/23849414

Remove a wrong TODO comment.

94d2c2f

Merge pull request #418 from gregomni/sr-184

cdaff82

[SR-184] Better diagnose error via unwrapping optional before initialized.

Resurrect call-graph dot viewing.

689db17

It's useful for debugging.

Merge branch 'master' of https://github.com/dcci/swift

8bd49d6

dcci closed this Dec 16, 2015

slavapestov pushed a commit to slavapestov/swift that referenced this pull request Nov 27, 2018

Merge pull request swiftlang#292 from apple/revert-286-major-update

837f2bd

Revert "linux: update header used for `major` macro"

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[Sema] Fix errors due to implicitly deleted copy constructor. #292

[Sema] Fix errors due to implicitly deleted copy constructor. #292

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

nadavrot commented Dec 7, 2015

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

tbartelmess commented Dec 7, 2015

Uh oh!

jckarter commented Dec 7, 2015

Uh oh!

tbartelmess commented Dec 7, 2015

Uh oh!

gribozavr commented Dec 7, 2015

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

gribozavr commented Dec 7, 2015

Uh oh!

dcci commented Dec 8, 2015

Uh oh!

GuillaumeBibaut commented Dec 8, 2015

Uh oh!

dcci commented Dec 10, 2015

Uh oh!

dcci commented Dec 16, 2015

Uh oh!

Uh oh!

[Sema] Fix errors due to implicitly deleted copy constructor. #292

[Sema] Fix errors due to implicitly deleted copy constructor. #292

Uh oh!

Conversation

dcci commented Dec 7, 2015

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

nadavrot commented Dec 7, 2015

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

tbartelmess commented Dec 7, 2015

Uh oh!

jckarter commented Dec 7, 2015

Uh oh!

tbartelmess commented Dec 7, 2015

Uh oh!

gribozavr commented Dec 7, 2015

Uh oh!

dcci commented Dec 7, 2015

Uh oh!

gribozavr commented Dec 7, 2015

Uh oh!

dcci commented Dec 8, 2015

Uh oh!

GuillaumeBibaut commented Dec 8, 2015

Uh oh!

dcci commented Dec 10, 2015

Uh oh!

dcci commented Dec 16, 2015

Uh oh!

Uh oh!