Fangyi Zhou 6078f5eb21
Reland [Clang][analyzer] replace Stmt* with ConstCFGElement in SymbolConjured (#137355)
Closes #57270.

This PR changes the `Stmt *` field in `SymbolConjured` with
`CFGBlock::ConstCFGElementRef`. The motivation is that, when conjuring a
symbol, there might not always be a statement available, causing
information to be lost for conjured symbols, whereas the CFGElementRef
can always be provided at the callsite.

Following the idea, this PR changes callsites of functions to create
conjured symbols, and replaces them with appropriate `CFGElementRef`s.

There is a caveat at loop widening, where the correct location is the
CFG terminator (which is not an element and does not have a ref). In
this case, the first element in the block is passed as a location.

Previous PR #128251, Reverted at #137304.
2025-05-12 14:19:44 +02:00

84 lines
3.5 KiB
C++

//===--- LoopWidening.cpp - Widen loops -------------------------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
///
/// This file contains functions which are used to widen loops. A loop may be
/// widened to approximate the exit state(s), without analyzing every
/// iteration. The widening is done by invalidating anything which might be
/// modified by the body of the loop.
///
//===----------------------------------------------------------------------===//
#include "clang/StaticAnalyzer/Core/PathSensitive/LoopWidening.h"
#include "clang/AST/AST.h"
#include "clang/ASTMatchers/ASTMatchFinder.h"
#include "clang/StaticAnalyzer/Core/PathSensitive/ExplodedGraph.h"
using namespace clang;
using namespace ento;
using namespace clang::ast_matchers;
const auto MatchRef = "matchref";
namespace clang {
namespace ento {
ProgramStateRef getWidenedLoopState(ProgramStateRef PrevState,
const LocationContext *LCtx,
unsigned BlockCount,
ConstCFGElementRef Elem) {
// Invalidate values in the current state.
// TODO Make this more conservative by only invalidating values that might
// be modified by the body of the loop.
// TODO Nested loops are currently widened as a result of the invalidation
// being so inprecise. When the invalidation is improved, the handling
// of nested loops will also need to be improved.
ASTContext &ASTCtx = LCtx->getAnalysisDeclContext()->getASTContext();
const StackFrameContext *STC = LCtx->getStackFrame();
MemRegionManager &MRMgr = PrevState->getStateManager().getRegionManager();
const MemRegion *Regions[] = {MRMgr.getStackLocalsRegion(STC),
MRMgr.getStackArgumentsRegion(STC),
MRMgr.getGlobalsRegion()};
RegionAndSymbolInvalidationTraits ITraits;
for (auto *Region : Regions) {
ITraits.setTrait(Region,
RegionAndSymbolInvalidationTraits::TK_EntireMemSpace);
}
// References should not be invalidated.
auto Matches = match(
findAll(stmt(hasDescendant(
varDecl(hasType(hasCanonicalType(referenceType()))).bind(MatchRef)))),
*LCtx->getDecl()->getBody(), ASTCtx);
for (BoundNodes Match : Matches) {
const VarDecl *VD = Match.getNodeAs<VarDecl>(MatchRef);
assert(VD);
const VarRegion *VarMem = MRMgr.getVarRegion(VD, LCtx);
ITraits.setTrait(VarMem,
RegionAndSymbolInvalidationTraits::TK_PreserveContents);
}
// 'this' pointer is not an lvalue, we should not invalidate it. If the loop
// is located in a method, constructor or destructor, the value of 'this'
// pointer should remain unchanged. Ignore static methods, since they do not
// have 'this' pointers.
const CXXMethodDecl *CXXMD = dyn_cast<CXXMethodDecl>(STC->getDecl());
if (CXXMD && CXXMD->isImplicitObjectMemberFunction()) {
const CXXThisRegion *ThisR =
MRMgr.getCXXThisRegion(CXXMD->getThisType(), STC);
ITraits.setTrait(ThisR,
RegionAndSymbolInvalidationTraits::TK_PreserveContents);
}
return PrevState->invalidateRegions(Regions, Elem, BlockCount, LCtx, true,
nullptr, nullptr, &ITraits);
}
} // end namespace ento
} // end namespace clang