clang不止是前端编译器，更是连接了LLVM整个编译过程和其他工具的一个驱动程序。
在 clang/include/clang/Basic 目录下定义了众多td模版文件，例如DiagnosticDriverKinds.td就是Driver的相关诊断信息。对应的类型则是在 clang/include/clang/Basic/DiagnosticIDs.h 文件中定义了6个类型：
/// Used for handling and querying diagnostic IDs.
///
/// Can be used and shared by multiple Diagnostics for multiple translation units.
class DiagnosticIDs : public RefCountedBase {
public:
/// The level of the diagnostic, after it has been through mapping.
enum Level {
Ignored, Note, Remark, Warning, Error, Fatal
};
clang driver
clang Driver 负责拼接编译器命令和 ld 命令。
他的处理原理如下：
- Parse：解析传入的参数
- Pipeline：根据每个输入的文件和类型，组建action，具体类型可以查看ActionClass枚举类型，对应到具体的JobAction
- Bind：根据action选择对应的工具和文件名信息，具体可以通过
clang -ccc-print-bindings进行查看 - Translate：将输入的参数转换为不同的tool的参数。如clang -cc1 -arch arm64，在clang中使用的是-triple arm64-apple-ios14，而ld则会使用-arch arm64
- Execute：调用tool执行任务。该步骤会通过创建子进程方式调用tool。
举个实例：
/Test » xcrun -l clang test.c -v -O2 -o Test n14637@GIH-D-21687
env SDKROOT=/Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang test.c -v -O2 -o Test
Apple clang version 12.0.0 (clang-1200.0.32.29)
Target: x86_64-apple-darwin19.6.0
Thread model: posix
InstalledDir: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin
"/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang" -cc1 -triple x86_64-apple-macosx10.15.0 -Wdeprecated-objc-isa-usage -Werror=deprecated-objc-isa-usage -Werror=implicit-function-declaration -emit-obj -disable-free -disable-llvm-verifier -discard-value-names -main-file-name test.c -mrelocation-model pic -pic-level 2 -mthread-model posix -mframe-pointer=all -fno-strict-return -masm-verbose -munwind-tables -target-sdk-version=10.15.6 -fcompatibility-qualified-id-block-type-checking -target-cpu penryn -dwarf-column-info -debugger-tuning=lldb -target-linker-version 609.8 -v -resource-dir /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/clang/12.0.0 -isysroot /Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk -I/usr/local/include -internal-isystem /Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/usr/local/include -internal-isystem /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/clang/12.0.0/include -internal-externc-isystem /Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/usr/include -internal-externc-isystem /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/include -O2 -Wno-reorder-init-list -Wno-implicit-int-float-conversion -Wno-c99-designator -Wno-final-dtor-non-final-class -Wno-extra-semi-stmt -Wno-misleading-indentation -Wno-quoted-include-in-framework-header -Wno-implicit-fallthrough -Wno-enum-enum-conversion -Wno-enum-float-conversion -fdebug-compilation-dir /Test -ferror-limit 19 -fmessage-length 164 -stack-protector 1 -fstack-check -mdarwin-stkchk-strong-link -fblocks -fencode-extended-block-signature -fregister-global-dtors-with-atexit -fgnuc-version=4.2.1 -fobjc-runtime=macosx-10.15.0 -fmax-type-align=16 -fdiagnostics-show-option -fcolor-diagnostics -vectorize-loops -vectorize-slp -o /var/folders/71/830m4bcj2_v1ly5qbtsbhxm4w50c2h/T/test-72900e.o -x c test.c
clang -cc1 version 12.0.0 (clang-1200.0.32.29) default target x86_64-apple-darwin19.6.0
ignoring nonexistent directory "/Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/usr/local/include"
ignoring nonexistent directory "/Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/Library/Frameworks"
#include "..." search starts here:
#include <...> search starts here:
/usr/local/include
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/clang/12.0.0/include
/Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/usr/include
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/include
/Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/System/Library/Frameworks (framework directory)
End of search list.
"/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ld" -demangle -lto_library /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/libLTO.dylib -dynamic -arch x86_64 -platform_version macos 10.15.0 10.15.6 -syslibroot /Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk -o Test -L/usr/local/lib /var/folders/71/830m4bcj2_v1ly5qbtsbhxm4w50c2h/T/test-72900e.o -lSystem /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/clang/12.0.0/lib/darwin/libclang_rt.osx.a
可以看到，执行xcrun之后，实际上clang构造了两个job，一个是编译任务，一个是链接任务，最后根据job创建两个进程执行任务。
看到源码部分，在driver.cpp：
//解析参数，该方法会调用DriverOptTable中的getDriverOptTable方法判断clang driver支持的所有参数类型
//其内部的OptTable::Info InfoTable[]是通过clang/Driver/Options.inc生成的，而inc是由tablegen将optins.td转换的
IntrusiveRefCntPtr DiagOpts =
CreateAndPopulateDiagOpts(argv);
//诊断引擎绑定到一个翻译单元和一个SourceManager。
DiagnosticsEngine Diags(DiagID, &*DiagOpts, DiagClient);
//并将diagnostics传递给DiagnosticConsumer，以便向用户报告
if (!DiagOpts->DiagnosticSerializationFile.empty()) {
auto SerializedConsumer =
clang::serialized_diags::create(DiagOpts->DiagnosticSerializationFile,
&*DiagOpts, /*MergeChildRecords=*/true);
Diags.setClient(new ChainedDiagnosticConsumer(
Diags.takeClient(), std::move(SerializedConsumer)));
}
ProcessWarningOptions(Diags, *DiagOpts, /*ReportDiags=*/false);
//创建driver实例
Driver TheDriver(Path, llvm::sys::getDefaultTargetTriple(), Diags);
SetInstallDir(argv, TheDriver, CanonicalPrefixes);
TheDriver.setTargetAndMode(TargetAndMode);
//xxx
//1.构造需要执行的命令，跳转到最下面的方法
std::unique_ptr C(TheDriver.BuildCompilation(argv));
//2.构建action
if (TC.getTriple().isOSBinFormatMachO())
BuildUniversalActions(*C, C->getDefaultToolChain(), Inputs);
else
BuildActions(*C, C->getArgs(), Inputs, C->getActions());
//3.构建jobs
BuildJobs(*C);
//4.执行任务
Driver::ExecuteCompilation -> Compilation::ExecuteJobs -> Compilation::ExecuteCommand-> Command::Execute -> llvm::sys::ExecuteAndWait
Compilation *Driver::BuildCompilation(ArrayRef ArgList) {
//xxx
//1.1.根据命令行指定的参数进行解析
CLOptions = std::make_unique(
ParseArgStrings(ArgList.slice(1), IsCLMode(), ContainsError));
//1.2.获取triple并通过gettoolchain获取对应的toolchain
const ToolChain &TC = getToolChain(
*UArgs, computeTargetTriple(*this, TargetTriple, *UArgs));
//1.3.Compilation类接管参数
Compilation *C = new Compilation(*this, TC, UArgs.release(), TranslatedArgs,
ContainsError);
//1.4.获取输入文件进行编译，该方法会通过输入的文件扩展后缀获取文件类型
BuildInputs(C->getDefaultToolChain(), *TranslatedArgs, Inputs);
}
//1.1.1.
InputArgList Driver::ParseArgStrings(ArrayRef ArgStrings,
bool IsClCompatMode,
bool &ContainsError) {
//调用getOpts获取支持的所有参数，然后调用parseArgs对命令行参数进行解析
//解析的规则则是通过调用OptTable的ParseOneArg方法对字符串进行遍历解析
//ParseOneArg内部调用了accept方法，该方法对参数别名会进行判断和特殊处理，如下:
InputArgList Args =
getOpts().ParseArgs(ArgStrings, MissingArgIndex, MissingArgCount,
IncludedFlagsBitmask, ExcludedFlagsBitmask);
//判断参数是否支持，是否支持通过Optins.td文件进行查找，否则抛出异常
for (const Arg *A : Args) {
if (A->getOption().hasFlag(options::Unsupported)) {
DiagID = diag::err_drv_unsupported_opt;
Diag(DiagID) << ArgString;
}
}
Arg *Option::accept(const ArgList &Args,
unsigned &Index,
unsigned ArgSize) const {
//解析参数
std::unique_ptr A(acceptInternal(Args, Index, ArgSize));
if (!A)
return nullptr;
//判断是否有别名
const Option &UnaliasedOption = getUnaliasedOption();
if (getID() == UnaliasedOption.getID())
return A.release();
//从unalias选项中获取拼写，对应的关系在Optins.td中有声明
StringRef UnaliasedSpelling = Args.MakeArgString(
Twine(UnaliasedOption.getPrefix()) + Twine(UnaliasedOption.getName()));
}
void Driver::BuildUniversalActions(Compilation &C, const ToolChain &TC,
const InputList &BAInputs) const {
//2.1.根据-arch参数生成需要处理的的Archs，构建actions
DerivedArgList &Args = C.getArgs();
ActionList &Actions = C.getActions();
for (Arg *A : Args) {
if (A->getOption().matches(options::OPT_arch)) {
//xxx
if (ArchNames.insert(A->getValue()).second)
Archs.push_back(A->getValue());
}
}
//如果没有传入 -arch 参数，则获取 triple 对应的架构
if (!Archs.size())
Archs.push_back(Args.MakeArgString(TC.getDefaultUniversalArchName()));
ActionList SingleActions;
//调用 Driver::handleArguments 方法对参数进行处理
BuildActions(C, Args, BAInputs, SingleActions);
// 构建 offloading actions.
OffloadingActionBuilder OffloadBuilder(C, Args, Inputs);
// 构造要执行的actions
HeaderModulePrecompileJobAction *HeaderModuleAction = nullptr;
ActionList LinkerInputs;
ActionList MergerInputs;
for (auto &I : Inputs) {
types::ID InputType = I.first;
const Arg *InputArg = I.second;
//根据输入源码文件inputs获取需要处理的 phase 数组
//phase其实就是一个枚举类型，包含Preprocess Precompile Compile Backend Assemble Link IfsMerge
//在 Types.def 文件维护了不同文件类型默认情况下需要经历的 phase
llvm::SmallVector PL;
types::getCompilationPhases(*this, Args, InputType, PL);
if (PL.empty())
continue;
//xxx
}
//2.2.
void Driver::BuildActions(Compilation &C, DerivedArgList &Args,
const InputList &Inputs, ActionList &Actions) const {
//将每个phase转化为一个action
for (phases::ID Phase : PL) {
//xxx
// Otherwise construct the appropriate action.
//该方法内部根据Phase的类型进行action的构建，详情在个下方法：
Action *NewCurrent = ConstructPhaseAction(C, Args, Phase, Current);
if (auto *HMA = dyn_cast(NewCurrent))
HeaderModuleAction = HMA;
Current = NewCurrent;
// Use the current host action in any of the offloading actions, if
// required.
if (OffloadBuilder.addHostDependenceToDeviceActions(Current, InputArg))
break;
if (Current->getType() == types::TY_Nothing)
break;
}
}
Action *Driver::ConstructPhaseAction(
Compilation &C, const ArgList &Args, phases::ID Phase, Action *Input,
Action::OffloadKind TargetDeviceOffloadKind) const {
switch (Phase) {
case phases::Preprocess: {
//通过 Input 和 OutputTy 构建 PreprocessJobAction
return C.MakeAction(Input, OutputTy);
}
return C.MakeAction(Input, OutputTy);
}
case phases::Compile: {
return C.MakeAction(Input, types::TY_LLVM_BC);
}
case phases::Backend: {
if (isUsingLTO() && TargetDeviceOffloadKind == Action::OFK_None) {
types::ID Output =
Args.hasArg(options::OPT_S) ? types::TY_LTO_IR : types::TY_LTO_BC;
return C.MakeAction(Input, Output);
}
if (Args.hasArg(options::OPT_emit_llvm)) {
types::ID Output =
Args.hasArg(options::OPT_S) ? types::TY_LLVM_IR : types::TY_LLVM_BC;
return C.MakeAction(Input, Output);
}
return C.MakeAction(Input, types::TY_PP_Asm);
}
case phases::Assemble:
return C.MakeAction(std::move(Input), types::TY_Object);
}
}
//3.构建jobs
void Driver::BuildJobs(Compilation &C) const {
//1.收集需要处理的架构
llvm::StringSet<> ArchNames;
if (C.getDefaultToolChain().getTriple().isOSBinFormatMachO())
for (const Arg *A : C.getArgs())
if (A->getOption().matches(options::OPT_arch))
ArchNames.insert(A->getValue());
//2.缓存action和inputinfo的映射
//BuildJobsForAction 方法会先查找缓存，如果缓存中不存在则再调用 BuildJobsForActionNoCache 方法创建 InputInfo
std::map, InputInfo> CachedResults;
BuildJobsForAction(C, A, &C.getDefaultToolChain(),
/*BoundArch*/ StringRef(),
/*AtTopLevel*/ true,
/*MultipleArchs*/ ArchNames.size() > 1,
/*LinkingOutput*/ LinkingOutput, CachedResults,
/*TargetDeviceOffloadKind*/ Action::OFK_None);
}
InputInfo Driver::BuildJobsForActionNoCache(
Compilation &C, const Action *A, const ToolChain *TC, StringRef BoundArch,
bool AtTopLevel, bool MultipleArchs, const char *LinkingOutput,
std::map, InputInfo> &CachedResults,
Action::OffloadKind TargetDeviceOffloadKind) const {
if (const BindArchAction *BAA = dyn_cast(A)) {
const ToolChain *TC;
StringRef ArchName = BAA->getArchName();
//3.通过 computeTargetTriple 计算 triple，然后获取合适的工具链
// computeTargetTriple会获取-target参数更新TargetTriple字符串，然后根据其生产Triple实例
if (!ArchName.empty())
TC = &getToolChain(C.getArgs(),
computeTargetTriple(*this, TargetTriple,
C.getArgs(), ArchName));
else
TC = &C.getDefaultToolChain();
//4.随后会以 BindArchAction 持有的第一个 input（类型是 LinkJobAction）为参数再次调用 BuildJobsForAction 方法
return BuildJobsForAction(C, *BAA->input_begin(), TC, ArchName, AtTopLevel,
MultipleArchs, LinkingOutput, CachedResults,
TargetDeviceOffloadKind);
}
//5.获取 LinkJobAction 的 Inputs
ActionList Inputs = A->getInputs();
const JobAction *JA = cast(A);
ActionList CollapsedOffloadActions;
//6.创建 ToolSelector 的实例 TS，并调用 ToolSelector::getTool 获取支持 link 的工具
ToolSelector TS(JA, *TC, C, isSaveTempsEnabled(),
embedBitcodeInObject() && !isUsingLTO());
const Tool *T = TS.getTool(Inputs, CollapsedOffloadActions);
//7.通过 BuildJobsForAction 处理 Inputs
for (const auto *OA : CollapsedOffloadActions)
cast(OA)->doOnEachDependence(
/*IsHostDependence=*/BuildingForOffloadDevice,
[&](Action *DepA, const ToolChain *DepTC, const char *DepBoundArch) {
OffloadDependencesInputInfo.push_back(BuildJobsForAction(
C, DepA, DepTC, DepBoundArch, /* AtTopLevel */ false,
/*MultipleArchs=*/!!DepBoundArch, LinkingOutput, CachedResults,
DepA->getOffloadingDeviceKind()));
});
//xxx
//调用 Compilation::getArgsForToolChain 进行参数转换
llvm::Triple EffectiveTriple;
const ToolChain &ToolTC = T->getToolChain();
const ArgList &Args =
C.getArgsForToolChain(TC, BoundArch, A->getOffloadingDeviceKind());
//xxx
if (UnbundlingResults.empty())
//调用 darwin::Linker 的 ConstructJob 方法构建 Job
T->ConstructJob(
C, *JA, Result, InputInfos,
C.getArgsForToolChain(TC, BoundArch, JA->getOffloadingDeviceKind()),
LinkingOutput);
else
T->ConstructJobMultipleOutputs(
C, *JA, UnbundlingResults, InputInfos,
C.getArgsForToolChain(TC, BoundArch, JA->getOffloadingDeviceKind()),
LinkingOutput);
}