Add abilities to generate bitcode buffers from munged bitcode.

lib/Bitcode/NaCl/TestUtils/NaClBitcodeMunge.cpp

Index: lib/Bitcode/NaCl/TestUtils/NaClBitcodeMunge.cpp
diff --git a/lib/Bitcode/NaCl/TestUtils/NaClBitcodeMunge.cpp b/lib/Bitcode/NaCl/TestUtils/NaClBitcodeMunge.cpp
index 2d88585ab5c81813db593a15b3473cc1d74627f9..147cee6c1d1f12e8d79767adf71fe45902f24602 100644
--- a/lib/Bitcode/NaCl/TestUtils/NaClBitcodeMunge.cpp
+++ b/lib/Bitcode/NaCl/TestUtils/NaClBitcodeMunge.cpp
@@ -15,7 +15,6 @@
 
 #include "llvm/Bitcode/NaCl/NaClBitcodeHeader.h"
 #include "llvm/Bitcode/NaCl/NaClBitcodeParser.h"
-#include "llvm/Bitcode/NaCl/NaClBitstreamWriter.h"
 #include "llvm/Bitcode/NaCl/NaClCompress.h"
 #include "llvm/Bitcode/NaCl/NaClReaderWriter.h"
 #include "llvm/IR/LLVMContext.h"
@@ -27,9 +26,8 @@
 
 using namespace llvm;
 
-// For debugging. When true, shows each PNaCl record that is
-// emitted to the bitcode file.
-static bool DebugEmitRecord = false;
+// For debugging. When true, shows each test being run.
+static bool TraceTestRuns = false;
 
 void NaClBitcodeMunger::setupTest(
     const char *TestName, const uint64_t Munges[], size_t MungesSize,
@@ -42,18 +40,18 @@ void NaClBitcodeMunger::setupTest(
   std::string DumpBuffer;
   DumpStream = new raw_string_ostream(DumpResults);
   MungedInputBuffer.clear();
-  NaClBitstreamWriter OutStream(MungedInputBuffer);
-  Writer = &OutStream;
 
-  if (DebugEmitRecord) {
+  if (TraceTestRuns) {
     errs() << "*** Run test: " << TestName << "\n";
   }
 
-  WriteBlockID = -1;
-  SetBID = -1;
-
   MungedBitcode.munge(Munges, MungesSize, RecordTerminator);
-  writeMungedBitcode(MungedBitcode, AddHeader);
+  NaClMungedBitcode::WriteResults Results =
+      MungedBitcode.writeMaybeRepair(MungedInputBuffer, AddHeader, WriteFlags);
+  if (Results.HasErrors
+      && (!WriteFlags.TryToRecover || !Results.HasRepairs)

[jvoung (off chromium), 2015/05/06 22:08:09]

Hmm might be easier to read with a round of Captain DeMorgan's.

HasErrors
&& !(WriteFlags.TryToRecover && Results.HasRepairs)
&& !(FlagSaysOkayForX && ResultWasX)

?

[Karl, 2015/05/07 20:09:18]

Done.

+      && (!WriteFlags.SaveBadAbbrevIndices || !Results.SavedBadAbbrevIndices))
+    report_fatal_error("Unable to generate bitcode file due to write errors");
 
   // Add null terminator, so that we meet the requirements of the
   // MemoryBuffer API.
@@ -71,7 +69,6 @@ void NaClBitcodeMunger::cleanupTest() {
   DumpStream->flush();
   delete DumpStream;
   DumpStream = nullptr;
-  Writer = nullptr;
 }
 
 // Return the next line of input (including eoln), starting from
@@ -108,219 +105,6 @@ getLinesWithTextMatch(const std::string &Substring, bool MustBePrefix) const {
   return Messages;
 }
 
-void NaClBitcodeMunger::writeMungedBitcode(const NaClMungedBitcode &Bitcode,
-                                           bool AddHeader) {
-  if (AddHeader) {
-    NaClWriteHeader(*Writer, true);
-  }
-  for (const NaClBitcodeAbbrevRecord &Record : Bitcode) {
-    emitRecord(Record);
-  }
-}
-
-void NaClBitcodeMunger::emitRecord(const NaClBitcodeAbbrevRecord &Record) {
-  if (DebugEmitRecord) {
-    errs() << "Emit " << Record.Abbrev << ": <" << Record.Code;
-    for (size_t i = 0, e = Record.Values.size(); i < e; ++i) {
-      errs() << ", " << Record.Values[i];
-    }
-    errs() << ">\n";
-  }
-
-  switch (Record.Code) {
-  case naclbitc::BLK_CODE_ENTER: {
-    unsigned NumBits = naclbitc::DEFAULT_MAX_ABBREV;
-    WriteBlockID = -1;
-    if (Record.Abbrev != naclbitc::ENTER_SUBBLOCK) {
-      Fatal() << "Enter block record code " << Record.Code
-              << " uses illegal abbreviation index " << Record.Abbrev << "\n";
-      ReportFatalError();
-    }
-    if (Record.Values.size() == 2) {
-      WriteBlockID = Record.Values[0];
-      NumBits = Record.Values[1];
-      if (NumBits > 32) {
-        Fatal() << "Error: Bit size " << NumBits
-                << " for record should be <= 32\n";
-        ReportFatalError();
-      }
-      if (NumBits < 2) {
-        Fatal() << "Error: Bit size " << NumBits
-                << " for record should be >= 2\n";
-        ReportFatalError();
-      }
-    } else {
-      Fatal() << "Error: Values for enter record should be of size 2. Found: "
-              << Record.Values.size();
-      ReportFatalError();
-    }
-    uint64_t MaxAbbrev = (static_cast<uint64_t>(1) << NumBits) - 1;
-    AbbrevIndexLimitStack.push_back(MaxAbbrev);
-    if (WriteBlockID == naclbitc::BLOCKINFO_BLOCK_ID) {
-      if (NumBits != naclbitc::DEFAULT_MAX_ABBREV) {
-        Fatal()
-            << "Error: Numbits entry for abbreviations record not 2. Found: "
-            << NumBits << "\n";
-        ReportFatalError();
-      }
-      Writer->EnterBlockInfoBlock();
-    } else {
-      NaClBitcodeSelectorAbbrev CurCodeLen(MaxAbbrev);
-      Writer->EnterSubblock(WriteBlockID, CurCodeLen);
-    }
-    return;
-  }
-  case naclbitc::BLK_CODE_EXIT:
-    if (Record.Abbrev != naclbitc::END_BLOCK) {
-      Fatal() << "Error: Exit block record code " << Record.Code
-              << " uses illegal abbreviation index " << Record.Abbrev << "\n";
-      ReportFatalError();
-    }
-    if (!Record.Values.empty()) {
-      Fatal() << "Error: Exit block should not have values. Found: "
-              << Record.Values.size() << "\n";
-      ReportFatalError();
-    }
-    if (!AbbrevIndexLimitStack.empty())
-      AbbrevIndexLimitStack.pop_back();
-    Writer->ExitBlock();
-    return;
-  case naclbitc::BLK_CODE_DEFINE_ABBREV: {
-    if (Record.Abbrev != naclbitc::DEFINE_ABBREV) {
-      Fatal() << "Error: Define abbreviation record code " << Record.Code
-               << " uses illegal abbreviation index " << Record.Abbrev << "\n";
-      ReportFatalError();
-    }
-    NaClBitCodeAbbrev *Abbrev = buildAbbrev(Record);
-    if (Abbrev == NULL) return;
-    if (WriteBlockID == naclbitc::BLOCKINFO_BLOCK_ID) {
-      Writer->EmitBlockInfoAbbrev(SetBID, Abbrev);
-    } else {
-      Writer->EmitAbbrev(Abbrev);
-    }
-    return;
-  }
-  case naclbitc::BLK_CODE_HEADER:
-    // Note: There is no abbreviation index here. Ignore.
-    for (uint64_t Value : Record.Values)
-      Writer->Emit(Value, 8);
-    return;
-  default:
-    if ((Record.Abbrev != naclbitc::UNABBREV_RECORD
-         && !Writer->isUserRecordAbbreviation(Record.Abbrev))) {
-      uint64_t BlockAbbrevIndexLimit = 0;
-      if (!AbbrevIndexLimitStack.empty())
-        BlockAbbrevIndexLimit = AbbrevIndexLimitStack.back();
-      if (Record.Abbrev > BlockAbbrevIndexLimit) {
-        Fatal() << "Error: Record code " << Record.Code
-                << " uses illegal abbreviation index " << Record.Abbrev
-                << ". Must not exceed " << BlockAbbrevIndexLimit << "\n";
-        ReportFatalError();
-      }
-      // Note: If this point is reached, the abbreviation is
-      // bad. However, that may be the point of munge being
-      // applied. Hence, emit the bad abbreviation and the data so
-      // that the reader can be tested on this bad input.  For
-      // simplicity, we output the record data using the default
-      // abbreviation pattern.
-      errs() << "Warning: Record code " << Record.Code
-             << " uses illegal abbreviation index " << Record.Abbrev << "\n";
-      Writer->EmitCode(Record.Abbrev);
-      Writer->EmitVBR(Record.Code, 6);
-      uint32_t NumValues = static_cast<uint32_t>(Record.Values.size());
-      Writer->EmitVBR(NumValues, 6);
-      for (uint32_t i = 0; i < NumValues; ++i) {
-        Writer->EmitVBR64(Record.Values[i], 6);
-      }
-      return;
-    }
-    if (Record.Abbrev == naclbitc::UNABBREV_RECORD)
-      Writer->EmitRecord(Record.Code, Record.Values);
-    else
-      Writer->EmitRecord(Record.Code, Record.Values, Record.Abbrev);
-    return;
-  }
-  Fatal() << "emitRecord on unimplemented code" << "\n";
-  ReportFatalError();
-}
-
-NaClBitCodeAbbrev *NaClBitcodeMunger::buildAbbrev(
-    const NaClBitcodeAbbrevRecord &Record) {
-  NaClBitCodeAbbrev *Abbrev = new NaClBitCodeAbbrev();
-  size_t Index = 0;
-  if (Record.Values.empty()) {
-    Fatal() << "Empty abbreviation record not allowed\n";
-    ReportFatalError();
-  }
-  size_t NumAbbreviations = Record.Values[Index++];
-  size_t NumValues = Record.Values.size();
-  if (NumAbbreviations == 0) {
-    Fatal() << "Abbreviation must contain at least one operator\n";
-    ReportFatalError();
-  }
-  for (size_t Count = 0; Count < NumAbbreviations; ++Count) {
-    if (Index >= NumValues) {
-      Fatal() << "Malformed abbreviation found. Expects "
-              << NumAbbreviations << " operands. Found: "
-              << Count << "\n";
-      ReportFatalError();
-    }
-    switch (Record.Values[Index++]) {
-    case 1:
-      if (Index >= NumValues) {
-        Fatal() << "Malformed literal abbreviation.\n";
-        ReportFatalError();
-      }
-      Abbrev->Add(NaClBitCodeAbbrevOp(Record.Values[Index++]));
-      break;
-    case 0: {
-      if (Index >= NumValues) {
-        Fatal() << "Malformed abbreviation found.\n";
-        ReportFatalError();
-      }
-      unsigned Kind = Record.Values[Index++];
-      switch (Kind) {
-      case NaClBitCodeAbbrevOp::Fixed:
-        if (Index >= NumValues) {
-          Fatal() << "Malformed fixed abbreviation found.\n";
-          ReportFatalError();
-        }
-        Abbrev->Add(NaClBitCodeAbbrevOp(NaClBitCodeAbbrevOp::Fixed,
-                                        Record.Values[Index++]));
-        break;
-      case NaClBitCodeAbbrevOp::VBR:
-        if (Index >= NumValues) {
-          Fatal() << "Malformed vbr abbreviation found.\n";
-          ReportFatalError();
-        }
-        Abbrev->Add(NaClBitCodeAbbrevOp(NaClBitCodeAbbrevOp::VBR,
-                                        Record.Values[Index++]));
-        break;
-      case NaClBitCodeAbbrevOp::Array:
-        if (Index >= NumValues) {
-          Fatal() << "Malformed array abbreviation found.\n";
-          ReportFatalError();
-        }
-        Abbrev->Add(NaClBitCodeAbbrevOp(NaClBitCodeAbbrevOp::Array));
-        break;
-      case NaClBitCodeAbbrevOp::Char6:
-        Abbrev->Add(NaClBitCodeAbbrevOp(NaClBitCodeAbbrevOp::Char6));
-        break;
-      default:
-        Fatal() << "Unknown abbreviation kind. Found: " << Kind << "\n";
-        ReportFatalError();
-      }
-      break;
-    }
-    default:
-      Fatal() << "Error: Bad literal flag in abbreviation. Found: "
-              << Record.Values[Index];
-      ReportFatalError();
-    }
-  }
-  return Abbrev;
-}
-
 bool NaClObjDumpMunger::runTestWithFlags(
     const char *Name, const uint64_t Munges[], size_t MungesSize,
     bool AddHeader, bool NoRecords, bool NoAssembly) {