[llvm-cov] Add support for gcov --hash-filenames option.

[iignatev]

Summary

The patch adds support for --hash-filenames to llvm-cov. This option adds md5 hash of the source path to the name of the generated .gcov file. The option is crucial for cases where you have multiple files with the same name but can't use --preserve-paths as resulting filenames exceed the limit.

from gcov(1):

 -x
--hash-filenames
    By default, gcov uses the full pathname of the source files to to
    create an output filename.  This can lead to long filenames that
    can overflow filesystem limits.  This option creates names of the
    form source-file##md5.gcov, where the source-file component is
    the final filename part and the md5 component is calculated from
    the full mangled name that would have been used otherwise.

[iignatev]

forgot to add test/tools/llvm-cov/Inputs/test_hash.output

[vsk]

Hi @iignatev, thanks for the patch. LGTM.

In the future you might get a faster review by setting assigning a few reviewers on Phab (typically, the last few people who've touched the code).

docs/CommandGuide/llvm-cov.rst

156 ↗

(On Diff #187315)

nit, calculated

[iignatev]

Hi @vsk,

thanks for your review. I've uploaded the patch w/ the typo in llvm-cov.rst fixed. as I don't have commit access, could you please help me w/ integration of this patch?

In the future you might get a faster review by setting assigning a few reviewers on Phab (typically, the last few people who've touched the code).

thanks for the tip, will definitely do next time.

llvm/trunk/docs/CommandGuide/llvm-cov.rst

 
  Include unconditional branches in the output for the --branch-probabilities
  option.
 
 .. option:: -version
 
  Display the version of llvm-cov.
 
+.. option:: -x, --hash-filenames
+
+ Use md5 hash of file name when naming the coverage output files. The source
+ file name will be suffixed by ``##`` followed by MD5 hash calculated for it.
+
 EXIT STATUS
 ^^^^^^^^^^^
 
 :program:`llvm-cov gcov` returns 1 if it cannot read input files.  Otherwise,
 it exits with zero.
 
 
 .. program:: llvm-cov show

llvm/trunk/include/llvm/ProfileData/GCOV.h

 class FileInfo;
 
 namespace GCOV {
 
 enum GCOVVersion { V402, V404, V704 };
 
 /// A struct for passing gcov options between functions.
 struct Options {
-  Options(bool A, bool B, bool C, bool F, bool P, bool U, bool L, bool N)
+  Options(bool A, bool B, bool C, bool F, bool P, bool U, bool L, bool N, bool X)
       : AllBlocks(A), BranchInfo(B), BranchCount(C), FuncCoverage(F),
-        PreservePaths(P), UncondBranch(U), LongFileNames(L), NoOutput(N) {}
+        PreservePaths(P), UncondBranch(U), LongFileNames(L), NoOutput(N),
+        HashFilenames(X) {}
 
   bool AllBlocks;
   bool BranchInfo;
   bool BranchCount;
   bool FuncCoverage;
   bool PreservePaths;
   bool UncondBranch;
   bool LongFileNames;
   bool NoOutput;
+  bool HashFilenames;
 };
 
 } // end namespace GCOV
 
 /// GCOVBuffer - A wrapper around MemoryBuffer to provide GCOV specific
 /// read operations.
 class GCOVBuffer {
 public:

llvm/trunk/lib/ProfileData/GCOV.cpp

 
 #include "llvm/ProfileData/GCOV.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/Config/llvm-config.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/FileSystem.h"
 #include "llvm/Support/Format.h"
 #include "llvm/Support/Path.h"
+#include "llvm/Support/MD5.h"
 #include "llvm/Support/raw_ostream.h"
 #include <algorithm>
 #include <system_error>
 
 using namespace llvm;
 
 //===----------------------------------------------------------------------===//
 // GCOVFile implementation.
if (Options.NoOutput)
     // mangled at all, and the -l and -p options are ignored. Here, we do the
     // same.
     return Filename;
 
   std::string CoveragePath;
   if (Options.LongFileNames && !Filename.equals(MainFilename))
     CoveragePath =
         mangleCoveragePath(MainFilename, Options.PreservePaths) + "##";
-  CoveragePath += mangleCoveragePath(Filename, Options.PreservePaths) + ".gcov";
+  CoveragePath += mangleCoveragePath(Filename, Options.PreservePaths);
+  if (Options.HashFilenames) {
+    MD5 Hasher;
+    MD5::MD5Result Result;
+    Hasher.update(Filename.str());
+    Hasher.final(Result);
+    CoveragePath += "##" + Result.digest().str().str();
+  }
+  CoveragePath += ".gcov";
   return CoveragePath;
 }
 
 std::unique_ptr<raw_ostream>
 FileInfo::openCoveragePath(StringRef CoveragePath) {
   if (Options.NoOutput)
     return llvm::make_unique<raw_null_ostream>();
 

llvm/trunk/test/tools/llvm-cov/Inputs/test_hash.output

+File 'srcdir/./nested_dir/../test.cpp'
+Lines executed:84.21% of 38
+srcdir/./nested_dir/../test.cpp:creating 'test.cpp##a806e5b3093cd6f683da88c0da150daf.gcov'
+
+File 'srcdir/./nested_dir/../test.h'
+Lines executed:100.00% of 1
+srcdir/./nested_dir/../test.h:creating 'test.h##0cbee7e2421fa4517420ac4f935620ca.gcov'
+

llvm/trunk/test/tools/llvm-cov/llvm-cov.test

 RUN: diff -aub test_paths.cpp.gcov test_paths.cpp##test.cpp.gcov
 RUN: diff -aub test_paths.h.gcov test_paths.cpp##test.h.gcov
 
 # Long file names and preserve paths.
 RUN: llvm-cov gcov -lp -gcno test_paths.gcno -gcda test_paths.gcda srcdir/../test_paths.cpp | diff -u test_long_paths.output -
 RUN: diff -aub test_paths.cpp.gcov srcdir#^#test_paths.cpp##srcdir#nested_dir#^#test.cpp.gcov
 RUN: diff -aub test_paths.h.gcov srcdir#^#test_paths.cpp##srcdir#nested_dir#^#test.h.gcov
 
+# Hash pathnames.
+RUN: llvm-cov gcov -x -gcno test_paths.gcno -gcda test_paths.gcda srcdir/../test_paths.cpp | diff -u test_hash.output -
+RUN: diff -aub test_paths.cpp.gcov test.cpp##a806e5b3093cd6f683da88c0da150daf.gcov
+RUN: diff -aub test_paths.h.gcov test.h##0cbee7e2421fa4517420ac4f935620ca.gcov
+
 # Function summaries. This changes stdout, but not the gcov files.
 RUN: llvm-cov gcov test.c -f | diff -u test_-f.output -
 RUN: diff -aub test_no_options.cpp.gcov test.cpp.gcov
 RUN: diff -aub test_no_options.h.gcov test.h.gcov
 
 # All blocks. This doesn't affect stdout, only the gcov files.
 RUN: llvm-cov gcov test.c -a | diff -u test_no_options.output -
 RUN: diff -aub test_-a.cpp.gcov test.cpp.gcov

llvm/trunk/tools/llvm-cov/gcov.cpp

cl::opt<bool> PreservePaths("p", cl::Grouping, cl::init(false),
                               cl::desc("Preserve path components"));
   cl::alias PreservePathsA("preserve-paths", cl::aliasopt(PreservePaths));
 
   cl::opt<bool> UncondBranch("u", cl::Grouping, cl::init(false),
                              cl::desc("Display unconditional branch info "
                                       "(requires -b)"));
   cl::alias UncondBranchA("unconditional-branches", cl::aliasopt(UncondBranch));
 
+  cl::opt<bool> HashFilenames("x", cl::Grouping, cl::init(false),
+                              cl::desc("Hash long pathnames"));
+  cl::alias HashFilenamesA("hash-filenames", cl::aliasopt(HashFilenames));
+
+
   cl::OptionCategory DebugCat("Internal and debugging options");
   cl::opt<bool> DumpGCOV("dump", cl::init(false), cl::cat(DebugCat),
                          cl::desc("Dump the gcov file to stderr"));
   cl::opt<std::string> InputGCNO("gcno", cl::cat(DebugCat), cl::init(""),
                                  cl::desc("Override inferred gcno file"));
   cl::opt<std::string> InputGCDA("gcda", cl::cat(DebugCat), cl::init(""),
                                  cl::desc("Override inferred gcda file"));
 
   cl::ParseCommandLineOptions(argc, argv, "LLVM code coverage tool\n");
 
   GCOV::Options Options(AllBlocks, BranchProb, BranchCount, FuncSummary,
-                        PreservePaths, UncondBranch, LongNames, NoOutput);
+                        PreservePaths, UncondBranch, LongNames, NoOutput,
+                        HashFilenames);
 
   for (const auto &SourceFile : SourceFiles)
     reportCoverage(SourceFile, ObjectDir, InputGCNO, InputGCDA, DumpGCOV,
                    Options);
   return 0;
 }