1//===--- SarifDiagnostics.cpp - Sarif Diagnostics for Paths -----*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This file defines the SarifDiagnostics object.
10//
11//===----------------------------------------------------------------------===//
12
13#include "clang/Analysis/MacroExpansionContext.h"
14#include "clang/Analysis/PathDiagnostic.h"
15#include "clang/Basic/FileManager.h"
16#include "clang/Basic/Sarif.h"
17#include "clang/Basic/SourceManager.h"
18#include "clang/Basic/Version.h"
19#include "clang/Lex/Preprocessor.h"
20#include "clang/StaticAnalyzer/Core/PathDiagnosticConsumers.h"
21#include "llvm/ADT/STLExtras.h"
22#include "llvm/ADT/StringMap.h"
23#include "llvm/Support/ConvertUTF.h"
24#include "llvm/Support/JSON.h"
25#include "llvm/Support/Path.h"
26
27using namespace llvm;
28using namespace clang;
29using namespace ento;
30
31namespace {
32class SarifDiagnostics : public PathDiagnosticConsumer {
33 std::string OutputFile;
34 const LangOptions &LO;
35 SarifDocumentWriter SarifWriter;
36
37public:
38 SarifDiagnostics(const std::string &Output, const LangOptions &LO,
39 const SourceManager &SM)
40 : OutputFile(Output), LO(LO), SarifWriter(SM) {}
41 ~SarifDiagnostics() override = default;
42
43 void FlushDiagnosticsImpl(std::vector<const PathDiagnostic *> &Diags,
44 FilesMade *FM) override;
45
46 StringRef getName() const override { return "SarifDiagnostics"; }
47 PathGenerationScheme getGenerationScheme() const override { return Minimal; }
48 bool supportsLogicalOpControlFlow() const override { return true; }
49 bool supportsCrossFileDiagnostics() const override { return true; }
50};
51} // end anonymous namespace
52
53void ento::createSarifDiagnosticConsumer(
54 PathDiagnosticConsumerOptions DiagOpts, PathDiagnosticConsumers &C,
55 const std::string &Output, const Preprocessor &PP,
56 const cross_tu::CrossTranslationUnitContext &CTU,
57 const MacroExpansionContext &MacroExpansions) {
58
59 // TODO: Emit an error here.
60 if (Output.empty())
61 return;
62
63 C.push_back(
64 x: new SarifDiagnostics(Output, PP.getLangOpts(), PP.getSourceManager()));
65 createTextMinimalPathDiagnosticConsumer(Diagopts: std::move(DiagOpts), C, Prefix: Output, PP,
66 CTU, MacroExpansions);
67}
68
69static StringRef getRuleDescription(StringRef CheckName) {
70 return llvm::StringSwitch<StringRef>(CheckName)
71#define GET_CHECKERS
72#define CHECKER(FULLNAME, CLASS, HELPTEXT, DOC_URI, IS_HIDDEN) \
73 .Case(FULLNAME, HELPTEXT)
74#include "clang/StaticAnalyzer/Checkers/Checkers.inc"
75#undef CHECKER
76#undef GET_CHECKERS
77 ;
78}
79
80static StringRef getRuleHelpURIStr(StringRef CheckName) {
81 return llvm::StringSwitch<StringRef>(CheckName)
82#define GET_CHECKERS
83#define CHECKER(FULLNAME, CLASS, HELPTEXT, DOC_URI, IS_HIDDEN) \
84 .Case(FULLNAME, DOC_URI)
85#include "clang/StaticAnalyzer/Checkers/Checkers.inc"
86#undef CHECKER
87#undef GET_CHECKERS
88 ;
89}
90
91static ThreadFlowImportance
92calculateImportance(const PathDiagnosticPiece &Piece) {
93 switch (Piece.getKind()) {
94 case PathDiagnosticPiece::Call:
95 case PathDiagnosticPiece::Macro:
96 case PathDiagnosticPiece::Note:
97 case PathDiagnosticPiece::PopUp:
98 // FIXME: What should be reported here?
99 break;
100 case PathDiagnosticPiece::Event:
101 return Piece.getTagStr() == "ConditionBRVisitor"
102 ? ThreadFlowImportance::Important
103 : ThreadFlowImportance::Essential;
104 case PathDiagnosticPiece::ControlFlow:
105 return ThreadFlowImportance::Unimportant;
106 }
107 return ThreadFlowImportance::Unimportant;
108}
109
110/// Accepts a SourceRange corresponding to a pair of the first and last tokens
111/// and converts to a Character granular CharSourceRange.
112static CharSourceRange convertTokenRangeToCharRange(const SourceRange &R,
113 const SourceManager &SM,
114 const LangOptions &LO) {
115 // Caret diagnostics have the first and last locations pointed at the same
116 // location, return these as-is.
117 if (R.getBegin() == R.getEnd())
118 return CharSourceRange::getCharRange(R);
119
120 SourceLocation BeginCharLoc = R.getBegin();
121 // For token ranges, the raw end SLoc points at the first character of the
122 // last token in the range. This must be moved to one past the end of the
123 // last character using the lexer.
124 SourceLocation EndCharLoc =
125 Lexer::getLocForEndOfToken(Loc: R.getEnd(), /* Offset = */ 0, SM, LangOpts: LO);
126 return CharSourceRange::getCharRange(B: BeginCharLoc, E: EndCharLoc);
127}
128
129static SmallVector<ThreadFlow, 8> createThreadFlows(const PathDiagnostic *Diag,
130 const LangOptions &LO) {
131 SmallVector<ThreadFlow, 8> Flows;
132 const PathPieces &Pieces = Diag->path.flatten(ShouldFlattenMacros: false);
133 for (const auto &Piece : Pieces) {
134 auto Range = convertTokenRangeToCharRange(
135 R: Piece->getLocation().asRange(), SM: Piece->getLocation().getManager(), LO);
136 auto Flow = ThreadFlow::create()
137 .setImportance(calculateImportance(Piece: *Piece))
138 .setRange(Range)
139 .setMessage(Piece->getString());
140 Flows.push_back(Elt: Flow);
141 }
142 return Flows;
143}
144
145static StringMap<uint32_t>
146createRuleMapping(const std::vector<const PathDiagnostic *> &Diags,
147 SarifDocumentWriter &SarifWriter) {
148 StringMap<uint32_t> RuleMapping;
149 llvm::StringSet<> Seen;
150
151 for (const PathDiagnostic *D : Diags) {
152 StringRef CheckName = D->getCheckerName();
153 std::pair<llvm::StringSet<>::iterator, bool> P = Seen.insert(key: CheckName);
154 if (P.second) {
155 auto Rule = SarifRule::create()
156 .setName(CheckName)
157 .setRuleId(CheckName)
158 .setDescription(getRuleDescription(CheckName))
159 .setHelpURI(getRuleHelpURIStr(CheckName));
160 size_t RuleIdx = SarifWriter.createRule(Rule);
161 RuleMapping[CheckName] = RuleIdx;
162 }
163 }
164 return RuleMapping;
165}
166
167static SarifResult createResult(const PathDiagnostic *Diag,
168 const StringMap<uint32_t> &RuleMapping,
169 const LangOptions &LO) {
170
171 StringRef CheckName = Diag->getCheckerName();
172 uint32_t RuleIdx = RuleMapping.lookup(Key: CheckName);
173 auto Range = convertTokenRangeToCharRange(
174 R: Diag->getLocation().asRange(), SM: Diag->getLocation().getManager(), LO);
175
176 SmallVector<ThreadFlow, 8> Flows = createThreadFlows(Diag, LO);
177 auto Result = SarifResult::create(RuleIdx)
178 .setRuleId(CheckName)
179 .setDiagnosticMessage(Diag->getVerboseDescription())
180 .setDiagnosticLevel(SarifResultLevel::Warning)
181 .setLocations({Range})
182 .setThreadFlows(Flows);
183 return Result;
184}
185
186void SarifDiagnostics::FlushDiagnosticsImpl(
187 std::vector<const PathDiagnostic *> &Diags, FilesMade *) {
188 // We currently overwrite the file if it already exists. However, it may be
189 // useful to add a feature someday that allows the user to append a run to an
190 // existing SARIF file. One danger from that approach is that the size of the
191 // file can become large very quickly, so decoding into JSON to append a run
192 // may be an expensive operation.
193 std::error_code EC;
194 llvm::raw_fd_ostream OS(OutputFile, EC, llvm::sys::fs::OF_TextWithCRLF);
195 if (EC) {
196 llvm::errs() << "warning: could not create file: " << EC.message() << '\n';
197 return;
198 }
199
200 std::string ToolVersion = getClangFullVersion();
201 SarifWriter.createRun(ShortToolName: "clang", LongToolName: "clang static analyzer", ToolVersion);
202 StringMap<uint32_t> RuleMapping = createRuleMapping(Diags, SarifWriter);
203 for (const PathDiagnostic *D : Diags) {
204 SarifResult Result = createResult(Diag: D, RuleMapping, LO);
205 SarifWriter.appendResult(SarifResult: Result);
206 }
207 auto Document = SarifWriter.createDocument();
208 OS << llvm::formatv(Fmt: "{0:2}\n", Vals: json::Value(std::move(Document)));
209}
210