office-gobmx/compilerplugins/clang/loopvartoosmall.cxx
Luboš Luňák adb08e892b make (some) clang plugins share the same RecursiveASTVisitor
Each plugin currently uses its own recursive AST run, which adds up.
This patch adds another shared plugin which internally contains all
(suitable) plugins and dispatches to them from the same one recursive
run. This patch converts ~25 plugins and for starmath's accessibility.cxx
reduces clang build time from 5.43s to 5.14s (and it's 4.39s without any
plugins). As there are almost 50 more plugins to go, this can theoretically
result in 4.56s final time, although probably not all plugins can be
that easily converted, if at all.

This mostly requires very little change in many plugins (see e.g.
BadStatics), some even work without any functionality change (e.g.
CharRightShift). Traverse* calls require some changes but are often
not that difficult. WalkUp* probably can't be supported, although some
plugins can(?) possibly be adjusted to not rely on them. And of course
some plugins can be left as they are, using their own recursive run.
See description at the top of generator.cxx for description of how to
convert a plugin.

The sharedvisitor.cxx source is generated based on scanning relevant
plugin sources using a clang-based scanner/generator. The generated
source is intentionally included instead of getting always generated,
as the generating currently takes some time, so it should get updated
in git whenever a change in a plugin triggers a source change in it.

Change-Id: Ia0d2e3a5a464659503dbb4ed6c20b6cc89b4de01
Reviewed-on: https://gerrit.libreoffice.org/68026
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2019-03-12 12:36:57 +01:00

245 lines
7.2 KiB
C++

/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
/*
* This file is part of the LibreOffice project.
*
* Based on LLVM/Clang.
*
* This file is distributed under the University of Illinois Open Source
* License. See LICENSE.TXT for details.
*
*/
#ifndef LO_CLANG_SHARED_PLUGINS
#include <algorithm>
#include <cassert>
#include <list>
#include <map>
#include "plugin.hxx"
//#include "clang/AST/CXXInheritance.h"
// Idea from bubli. Check that the index variable in a for loop is able to cover the range
// revealed by the terminating condition.
// If not, we might end up in an endless loop, or just not process certain parts.
namespace
{
class LoopVarTooSmall:
public loplugin::FilteringPlugin<LoopVarTooSmall>
{
public:
explicit LoopVarTooSmall(loplugin::InstantiationData const & data):
FilteringPlugin(data) {}
virtual void run() override {
TraverseDecl(compiler.getASTContext().getTranslationUnitDecl());
}
bool VisitForStmt( const ForStmt* stmt ) {
checkExpr(stmt->getCond());
return true;
}
bool VisitWhileStmt(WhileStmt const * stmt) {
checkExpr(stmt->getCond());
return true;
}
bool VisitDoStmt(DoStmt const * stmt) {
checkExpr(stmt->getCond());
return true;
}
private:
unsigned getIntValueWidth(QualType type) const;
void checkSubExpr(Expr const * expr, bool positive);
void checkExpr(Expr const * expr);
struct Comparison {
BinaryOperator const * op;
unsigned rhsWidth;
};
struct Comparisons {
std::list<Comparison> comparisons;
unsigned lhsWidth;
};
std::map<Decl const *, Comparisons> comparisons_;
};
unsigned LoopVarTooSmall::getIntValueWidth(QualType type) const {
if (auto const et = type->getAs<EnumType>()) {
auto const ed = et->getDecl();
if (!ed->isFixed()) {
unsigned pos = ed->getNumPositiveBits();
unsigned neg = ed->getNumNegativeBits();
return neg == 0 ? std::max(pos, 1U) : std::max(pos + 1, neg);
}
}
return compiler.getASTContext().getIntWidth(type);
}
void LoopVarTooSmall::checkSubExpr(Expr const * expr, bool positive) {
auto const e = expr->IgnoreImplicit()->IgnoreParenImpCasts();
if (auto const uo = dyn_cast<UnaryOperator>(e)) {
if (uo->getOpcode() == UO_LNot) {
checkSubExpr(uo->getSubExpr(), !positive);
}
return;
}
const BinaryOperator* binOp = dyn_cast<BinaryOperator>(e);
if (!binOp)
return;
bool less;
if (positive) {
switch (binOp->getOpcode()) {
case BO_LAnd:
checkSubExpr(binOp->getLHS(), true);
checkSubExpr(binOp->getRHS(), true);
return;
case BO_LT:
case BO_NE:
less = true;
break;
case BO_LE:
less = false;
break;
default:
return;
}
} else {
switch (binOp->getOpcode()) {
case BO_LOr:
checkSubExpr(binOp->getLHS(), false);
checkSubExpr(binOp->getRHS(), false);
return;
case BO_GE:
case BO_EQ:
less = true;
break;
case BO_GT:
less = false;
break;
default:
return;
}
}
auto lhs = dyn_cast<DeclRefExpr>(binOp->getLHS()->IgnoreParenImpCasts());
if (!lhs)
return;
QualType qt = lhs->getType();
if (!qt->isIntegralOrEnumerationType())
return;
unsigned qt1BitWidth = getIntValueWidth(qt);
auto lhsDecl = lhs->getDecl();
if (!isa<VarDecl>(lhsDecl)) {
if (auto fd = dyn_cast<FieldDecl>(lhsDecl)) {
if (fd->isBitField()) {
qt1BitWidth = std::max(
qt1BitWidth,
fd->getBitWidthValue(compiler.getASTContext()));
}
} else {
return;
}
}
const Expr* binOpRHS = binOp->getRHS()->IgnoreParenImpCasts();
QualType qt2 = binOpRHS->getType();
if (!qt2->isIntegralType(compiler.getASTContext()))
return;
unsigned qt2BitWidth;
llvm::APSInt aIntResult;
// Work around missing Clang 3.9 fix <https://reviews.llvm.org/rL271762>
// "Sema: do not attempt to sizeof a dependent type", causing Clang 3.8 to
// crash during EvaluateAsInt() on expressions of the form
//
// sizeof (T)
//
// with dependent type T:
if (!binOpRHS->isValueDependent()
&& compat::EvaluateAsInt(binOpRHS, aIntResult, compiler.getASTContext()))
{
if (less && aIntResult.isStrictlyPositive()) {
--aIntResult;
}
qt2BitWidth = aIntResult.isUnsigned() || !aIntResult.isNegative()
? std::max(aIntResult.getActiveBits(), 1U)
: aIntResult.getBitWidth() - aIntResult.countLeadingOnes() + 1;
} else {
// Ignore complex expressions for now, promotion rules on conditions
// like "i < (size()+1)" make it hard to guess at a correct type:
if (isa<BinaryOperator>(binOpRHS) || isa<ConditionalOperator>(binOpRHS))
{
return;
}
qt2BitWidth = getIntValueWidth(qt2);
if (auto dre = dyn_cast<DeclRefExpr>(binOpRHS)) {
if (auto fd = dyn_cast<FieldDecl>(dre->getDecl())) {
if (fd->isBitField()) {
qt2BitWidth = std::max(
qt2BitWidth,
fd->getBitWidthValue(compiler.getASTContext()));
}
}
}
}
auto i = comparisons_.find(lhsDecl);
if (i == comparisons_.end()) {
i = (comparisons_.insert(
decltype(comparisons_)::value_type(lhsDecl, {{}, qt1BitWidth}))
.first);
} else {
assert(i->second.lhsWidth == qt1BitWidth);
}
bool ins = true;
for (auto j = i->second.comparisons.begin();
j != i->second.comparisons.end();)
{
if (qt2BitWidth > j->rhsWidth) {
ins = false;
break;
} else if (qt2BitWidth < j->rhsWidth) {
j = i->second.comparisons.erase(j);
} else {
++j;
}
}
if (ins) {
i->second.comparisons.push_back({binOp, qt2BitWidth});
}
}
void LoopVarTooSmall::checkExpr(Expr const * expr) {
if (expr != nullptr && !ignoreLocation(expr)) {
assert(comparisons_.empty());
checkSubExpr(expr, true);
for (auto const & i: comparisons_) {
for (auto const & j: i.second.comparisons) {
if (i.second.lhsWidth < j.rhsWidth) {
report(
DiagnosticsEngine::Warning,
"loop index type %0 is narrower than length type %1",
j.op->getExprLoc())
<< j.op->getLHS()->IgnoreImpCasts()->getType()
<< j.op->getRHS()->IgnoreImpCasts()->getType()
<< j.op->getSourceRange();
}
}
}
comparisons_.clear();
}
}
loplugin::Plugin::Registration< LoopVarTooSmall > loopvartoosmall("loopvartoosmall");
}
#endif // LO_CLANG_SHARED_PLUGINS
/* vim:set shiftwidth=4 softtabstop=4 expandtab: */