v.g.vassilev created this revision.
v.g.vassilev added reviewers: rsmith, rjmccall, lhames, sgraenitz.
Herald added a subscriber: StephenFan.
Herald added a project: All.
v.g.vassilev requested review of this revision.
This patch teaches clang to parse statements on the global scope to allow:
./bin/clang-repl
clang-repl> int i = 12;
clang-repl> ++i;
clang-repl> extern "C" int printf(const char*,...);
clang-repl> printf("%d\n", i);
13
clang-repl> quit
The patch conceptually models the possible "top-level" statements between two
top-level declarations into a block which is emitted as part of the global init
function.
The current implementation in CodeGen is a placeholder allowing us to discuss
what would be the best approach to tackle this usecase. CodeGen works with
"GlobalDecl"s and it is awkward to model a block of statements in that way.
Currently, we attach each statement as an initializer of a global VarDecl.
However, that granularity is not desirable.
Repository:
rC Clang
https://reviews.llvm.org/D127284
Files:
clang/include/clang/AST/ASTConsumer.h
clang/include/clang/Parse/Parser.h
clang/lib/CodeGen/CodeGenAction.cpp
clang/lib/CodeGen/ModuleBuilder.cpp
clang/lib/Interpreter/IncrementalParser.cpp
clang/lib/Parse/ParseDecl.cpp
clang/lib/Parse/ParseTentative.cpp
clang/lib/Parse/Parser.cpp
clang/test/Interpreter/execute-stmts.cpp
Index: clang/test/Interpreter/execute-stmts.cpp
===================================================================
--- /dev/null
+++ clang/test/Interpreter/execute-stmts.cpp
@@ -0,0 +1,18 @@
+// REQUIRES: host-supports-jit
+// UNSUPPORTED: system-aix
+// RUN: cat %s | clang-repl | FileCheck %s
+
+int i = 12;
+++i;
+extern "C" int printf(const char*,...);
+printf("i = %d\n", i);
+// CHECK: i = 13
+
+// FIXME: isConstructorDeclarator thinks these are ctors:
+// namespace Ns { void f(){} }
+// Ns::f();
+
+// void g() {}
+// g();
+
+quit
Index: clang/lib/Parse/Parser.cpp
===================================================================
--- clang/lib/Parse/Parser.cpp
+++ clang/lib/Parse/Parser.cpp
@@ -582,13 +582,14 @@
///
/// Note that in C, it is an error if there is no first declaration.
bool Parser::ParseFirstTopLevelDecl(DeclGroupPtrTy &Result,
- Sema::ModuleImportState &ImportState) {
+ Sema::ModuleImportState &ImportState,
+ StmtVector *Stmts) {
Actions.ActOnStartOfTranslationUnit();
// For C++20 modules, a module decl must be the first in the TU. We also
// need to track module imports.
ImportState = Sema::ModuleImportState::FirstDecl;
- bool NoTopLevelDecls = ParseTopLevelDecl(Result, ImportState);
+ bool NoTopLevelDecls = ParseTopLevelDecl(Result, ImportState, Stmts);
// C11 6.9p1 says translation units must have at least one top-level
// declaration. C++ doesn't have this restriction. We also don't want to
@@ -609,7 +610,8 @@
/// declaration
/// [C++20] module-import-declaration
bool Parser::ParseTopLevelDecl(DeclGroupPtrTy &Result,
- Sema::ModuleImportState &ImportState) {
+ Sema::ModuleImportState &ImportState,
+ StmtVector *Stmts /*=nullptr*/) {
DestroyTemplateIdAnnotationsRAIIObj CleanupRAII(*this);
// Skip over the EOF token, flagging end of previous input for incremental
@@ -724,6 +726,23 @@
ParsedAttributes attrs(AttrFactory);
MaybeParseCXX11Attributes(attrs);
+ // FIXME: Remove the incremental processing pre-condition and verify clang
+ // still can pass its test suite, which will harden `isDeclarationStatement`.
+ // It is known to have several weaknesses, for example in
+ // isConstructorDeclarator, infinite loop in c-index-test, etc..
+ // Parse a block of top-level-stmts.
+ while (PP.isIncrementalProcessingEnabled() && Stmts &&
+ !isDeclarationStatement()) {
+ //isStmtExpr ? ParsedStmtContext::InStmtExpr : ParsedStmtContext()
+ ParsedStmtContext SubStmtCtx = ParsedStmtContext();
+ auto R = ParseStatementOrDeclaration(*Stmts, SubStmtCtx);
+ if (!R.isUsable())
+ return true;
+ Stmts->push_back(R.get());
+ if (Tok.is(tok::eof))
+ return false;
+ }
+
Result = ParseExternalDeclaration(attrs);
// An empty Result might mean a line with ';' or some parsing error, ignore
// it.
Index: clang/lib/Parse/ParseTentative.cpp
===================================================================
--- clang/lib/Parse/ParseTentative.cpp
+++ clang/lib/Parse/ParseTentative.cpp
@@ -22,6 +22,7 @@
///
/// declaration-statement:
/// block-declaration
+/// template-declaration
///
/// block-declaration:
/// simple-declaration
@@ -46,12 +47,24 @@
/// 'using' 'namespace' '::'[opt] nested-name-specifier[opt]
/// namespace-name ';'
///
+/// template-declaration:
+/// 'template' declaration
+/// 'template' concept-definition
+///
bool Parser::isCXXDeclarationStatement() {
switch (Tok.getKind()) {
- // asm-definition
+ // OpenCL
+ case tok::kw___kernel:
+ return getLangOpts().OpenCL;
+
case tok::kw_asm:
+ case tok::kw_export:
+ case tok::kw_import:
// namespace-alias-definition
+ case tok::kw_inline: // inline namespace
case tok::kw_namespace:
+ // template-declaration
+ case tok::kw_template:
// using-declaration
// using-directive
case tok::kw_using:
@@ -59,7 +72,20 @@
case tok::kw_static_assert:
case tok::kw__Static_assert:
return true;
+
+ case tok::identifier:
+ if (getLangOpts().CPlusPlus) {
+ // FIXME: This is quite broken as it considers f() from void f(){}; f();
+ // to be a constructor declarator...
+ if (isConstructorDeclarator(/*Unqualified=*/false))
+ return true;
+ // Check if this is a dtor.
+ if (!TryAnnotateTypeOrScopeToken() && Tok.is(tok::annot_cxxscope) &&
+ NextToken().is(tok::tilde))
+ return true;
+ }
// simple-declaration
+ LLVM_FALLTHROUGH;
default:
return isCXXSimpleDeclaration(/*AllowForRangeDecl=*/false);
}
@@ -1397,6 +1423,12 @@
case tok::kw_volatile:
return TPResult::True;
+ // ObjC
+ case tok::at:
+ if (getLangOpts().ObjC)
+ return TPResult::True;
+ LLVM_FALLTHROUGH;
+
// OpenCL address space qualifiers
case tok::kw_private:
if (!getLangOpts().OpenCL)
Index: clang/lib/Parse/ParseDecl.cpp
===================================================================
--- clang/lib/Parse/ParseDecl.cpp
+++ clang/lib/Parse/ParseDecl.cpp
@@ -5243,11 +5243,22 @@
switch (Tok.getKind()) {
default: return false;
+ // ObjC
+ case tok::at:
+ return getLangOpts().ObjC;
+
+ // OpenCL
+ case tok::kw___kernel:
+ return getLangOpts().OpenCL;
+
// OpenCL 2.0 and later define this keyword.
case tok::kw_pipe:
return getLangOpts().OpenCL &&
getLangOpts().getOpenCLCompatibleVersion() >= 200;
+ case tok::kw_asm:
+ return true;
+
case tok::identifier: // foo::bar
// Unfortunate hack to support "Class.factoryMethod" notation.
if (getLangOpts().ObjC && NextToken().is(tok::period))
@@ -5500,6 +5511,7 @@
}
ConsumeParen();
+ // FIXME: this returns true for NS::f();
// A right parenthesis, or ellipsis followed by a right parenthesis signals
// that we have a constructor.
if (Tok.is(tok::r_paren) ||
Index: clang/lib/Interpreter/IncrementalParser.cpp
===================================================================
--- clang/lib/Interpreter/IncrementalParser.cpp
+++ clang/lib/Interpreter/IncrementalParser.cpp
@@ -165,11 +165,17 @@
Parser::DeclGroupPtrTy ADecl;
Sema::ModuleImportState ImportState;
- for (bool AtEOF = P->ParseFirstTopLevelDecl(ADecl, ImportState); !AtEOF;
- AtEOF = P->ParseTopLevelDecl(ADecl, ImportState)) {
+ Parser::StmtVector Stmts;
+ for (bool AtEOF = P->ParseFirstTopLevelDecl(ADecl, ImportState, &Stmts);
+ !AtEOF; AtEOF = P->ParseTopLevelDecl(ADecl, ImportState, &Stmts)) {
// If we got a null return and something *was* parsed, ignore it. This
// is due to a top-level semicolon, an action override, or a parse error
// skipping something.
+ if (!Stmts.empty() && !Consumer->HandleTopLevelStmts(Stmts))
+ return llvm::make_error<llvm::StringError>("Parsing failed. "
+ "The consumer rejected a stmt",
+ std::error_code());
+
if (ADecl && !Consumer->HandleTopLevelDecl(ADecl.get()))
return llvm::make_error<llvm::StringError>("Parsing failed. "
"The consumer rejected a decl",
Index: clang/lib/CodeGen/ModuleBuilder.cpp
===================================================================
--- clang/lib/CodeGen/ModuleBuilder.cpp
+++ clang/lib/CodeGen/ModuleBuilder.cpp
@@ -168,7 +168,30 @@
Builder->HandleCXXStaticMemberVarInstantiation(VD);
}
+ bool HandleTopLevelStmts(const llvm::SmallVectorImpl<Stmt*> &Stmts) override {
+ if (Diags.hasErrorOccurred())
+ return false; // Abort parsing.
+
+ // FIXME: We should reimplement this in a proper way where we append the
+ // statements to the global init function.
+ static unsigned id = 0;
+ // Make sure to emit all elements of a Decl.
+ for (Stmt *S : Stmts) {
+ Expr *E = cast<Expr>(S);
+ IdentifierInfo* name = &Ctx->Idents.get("_stmt" + std::to_string(id++));
+ VarDecl *VD = VarDecl::Create(*Ctx, Ctx->getTranslationUnitDecl(),
+ E->getBeginLoc(), SourceLocation(),
+ name, E->getType(), /*TSI*/nullptr,
+ SC_None);
+ VD->setInit(E);
+ HandleTopLevelDecl(DeclGroupRef(VD));
+ }
+
+ return true;
+ }
+
bool HandleTopLevelDecl(DeclGroupRef DG) override {
+ // FIXME: Why not return false and abort parsing?
if (Diags.hasErrorOccurred())
return true;
Index: clang/lib/CodeGen/CodeGenAction.cpp
===================================================================
--- clang/lib/CodeGen/CodeGenAction.cpp
+++ clang/lib/CodeGen/CodeGenAction.cpp
@@ -237,6 +237,9 @@
return true;
}
+ bool HandleTopLevelStmts(const llvm::SmallVectorImpl<Stmt*> &Stmts) override {
+ return Gen->HandleTopLevelStmts(Stmts);
+ }
void HandleInlineFunctionDefinition(FunctionDecl *D) override {
PrettyStackTraceDecl CrashInfo(D, SourceLocation(),
Index: clang/include/clang/Parse/Parser.h
===================================================================
--- clang/include/clang/Parse/Parser.h
+++ clang/include/clang/Parse/Parser.h
@@ -462,6 +462,11 @@
typedef Sema::FullExprArg FullExprArg;
+ /// A SmallVector of statements, with stack size 32 (as that is the only one
+ /// used.)
+ typedef SmallVector<Stmt*, 32> StmtVector;
+
+
// Parsing methods.
/// Initialize - Warm up the parser.
@@ -470,12 +475,14 @@
/// Parse the first top-level declaration in a translation unit.
bool ParseFirstTopLevelDecl(DeclGroupPtrTy &Result,
- Sema::ModuleImportState &ImportState);
+ Sema::ModuleImportState &ImportState,
+ StmtVector *Stmts = nullptr);
/// ParseTopLevelDecl - Parse one top-level declaration. Returns true if
/// the EOF was encountered.
bool ParseTopLevelDecl(DeclGroupPtrTy &Result,
- Sema::ModuleImportState &ImportState);
+ Sema::ModuleImportState &ImportState,
+ StmtVector *Stmts = nullptr);
bool ParseTopLevelDecl() {
DeclGroupPtrTy Result;
Sema::ModuleImportState IS = Sema::ModuleImportState::NotACXX20Module;
@@ -2059,9 +2066,6 @@
//===--------------------------------------------------------------------===//
// C99 6.8: Statements and Blocks.
- /// A SmallVector of statements, with stack size 32 (as that is the only one
- /// used.)
- typedef SmallVector<Stmt*, 32> StmtVector;
/// A SmallVector of expressions, with stack size 12 (the maximum used.)
typedef SmallVector<Expr*, 12> ExprVector;
/// A SmallVector of types.
Index: clang/include/clang/AST/ASTConsumer.h
===================================================================
--- clang/include/clang/AST/ASTConsumer.h
+++ clang/include/clang/AST/ASTConsumer.h
@@ -13,6 +13,10 @@
#ifndef LLVM_CLANG_AST_ASTCONSUMER_H
#define LLVM_CLANG_AST_ASTCONSUMER_H
+namespace llvm {
+ template <typename T> class SmallVectorImpl;
+}
+
namespace clang {
class ASTContext;
class CXXMethodDecl;
@@ -22,6 +26,7 @@
class ASTMutationListener;
class ASTDeserializationListener; // layering violation because void* is ugly
class SemaConsumer; // layering violation required for safe SemaConsumer
+ class Stmt;
class TagDecl;
class VarDecl;
class FunctionDecl;
@@ -46,6 +51,15 @@
/// ASTContext.
virtual void Initialize(ASTContext &Context) {}
+ /// HandleTopLevelStmts - Handle the specified top-level statements. This is
+ /// called by the parser to process every top-level Stmt* in incremental
+ /// compilation mode.
+ ///
+ /// \returns true to continue parsing, or false to abort parsing.
+ virtual bool HandleTopLevelStmts(const llvm::SmallVectorImpl<Stmt*>& Stmts) {
+ return true;
+ }
+
/// HandleTopLevelDecl - Handle the specified top-level declaration. This is
/// called by the parser to process every top-level Decl*.
///
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
