[llvm-cov] Read in function names for filtering from a text file.

[seaneveson]

Summary

Add a -name-whitelist option, which behaves in the same way as -name, but it reads in multiple function names from the given input file(s).

[vsk]

test/tools/llvm-cov/Inputs/name_whitelist.cpp

18

Just a reminder: please make sure that name_whitelist.covmapping is derived from an object built with -mllvm -enable-name-compression=false. I've been bitten by that before :) (not all the bots have zlib installed).

tools/llvm-cov/CodeCoverage.cpp

656

In the future, we might benefit from the flexibility offered by llvm::SpecialCaseList. Could you use that to parse the whitelist?

[seaneveson]

tools/llvm-cov/CodeCoverage.cpp

656

Thanks for the review!

Having to specify a section for each item seems redundant for an option that only filters on function names, e.g.

fun:foo
fun:bar

Are you considering having a more general -whitelist rather than a -name-whitelist? Otherwise I'm not sure what other sections there would be. The wildcard handling would be useful though.

[vsk]

tools/llvm-cov/CodeCoverage.cpp

656

I agree that it's a little redundant, but think that it shouldn't be too onerous. E.g the section names can be added in by a quick shell one-liner. The potential benefit is that you can have one file for {white, black}listing {functions, files, directories}. To that end, I suggest using more specific section names like "whitelist_fun:".

[seaneveson]

Change to use SpecialCaseList and add documentation of new option.

[vsk]

Thanks!

tools/llvm-cov/CoverageFilters.h

64 ↗

(On Diff #113235)

nitpick: Can this method be marked const?

test/tools/llvm-cov/Inputs/name_whitelist.cpp

+int func1() {
+    return 1;
+}
+int func2() {
+    return 1;
+}
+int func3() {
+    return 1;
+}
+int func4() {
+    return 1;
+}
+int func5() {
+    return 1;
+}
+int func6() {
+    return 1;
+}

test/tools/llvm-cov/Inputs/name_whitelist.proftext

+_Z5func1v
+# Func Hash:
+0
+# Num Counters:
+1
+# Counter Values:
+0
+
+_Z5func2v
+# Func Hash:
+0
+# Num Counters:
+1
+# Counter Values:
+0
+
+_Z5func3v
+# Func Hash:
+0
+# Num Counters:
+1
+# Counter Values:
+0
+
+_Z5func4v
+# Func Hash:
+0
+# Num Counters:
+1
+# Counter Values:
+0
+
+main
+# Func Hash:
+0
+# Num Counters:
+1
+# Counter Values:
+1
+
+_Z5func5v
+# Func Hash:
+0
+# Num Counters:
+1
+# Counter Values:
+0
+
+_Z5func6v
+# Func Hash:
+0
+# Num Counters:
+1
+# Counter Values:
+0
+

test/tools/llvm-cov/Inputs/whitelist1.txt

+# Comment
+
+func1 # Comment
+  func2  # Comment

test/tools/llvm-cov/Inputs/whitelist2.txt

+func3
+func4

test/tools/llvm-cov/name_whitelist.test

+RUN: llvm-profdata merge %S/Inputs/name_whitelist.proftext -o %t.profdata
+
+RUN: llvm-cov show %S/Inputs/name_whitelist.covmapping -instr-profile=%t.profdata -path-equivalence=/tmp,%S/Inputs -name-whitelist=%S/Inputs/whitelist1.txt %S/Inputs/name_whitelist.cpp > %t.one_list
+RUN: FileCheck -input-file=%t.one_list -check-prefix=ONE_WHITELIST %s
+RUN: FileCheck -input-file=%t.one_list -check-prefix=ONE_WHITELIST_NEG %s
+ONE_WHITELIST: _Z5func1v:
+ONE_WHITELIST: _Z5func2v:
+ONE_WHITELIST_NEG-NOT: _Z5func3v:
+ONE_WHITELIST_NEG-NOT: _Z5func4v:
+ONE_WHITELIST_NEG-NOT: _Z5func5v:
+ONE_WHITELIST_NEG-NOT: _Z5func6v:
+
+RUN: llvm-cov show %S/Inputs/name_whitelist.covmapping -instr-profile=%t.profdata -path-equivalence=/tmp,%S/Inputs -name-whitelist=%S/Inputs/whitelist1.txt -name-whitelist=%S/Inputs/whitelist2.txt %S/Inputs/name_whitelist.cpp  > %t.two_list
+RUN: FileCheck -input-file=%t.two_list -check-prefix=TWO_WHITELIST %s
+RUN: FileCheck -input-file=%t.two_list -check-prefix=TWO_WHITELIST_NEG %s
+TWO_WHITELIST: _Z5func1v:
+TWO_WHITELIST: _Z5func2v:
+TWO_WHITELIST: _Z5func3v:
+TWO_WHITELIST: _Z5func4v:
+TWO_WHITELIST_NEG-NOT: _Z5func5v:
+TWO_WHITELIST_NEG-NOT: _Z5func6v:

tools/llvm-cov/CodeCoverage.cpp

   std::mutex LoadedSourceFilesLock;
   std::vector<std::pair<std::string, std::unique_ptr<MemoryBuffer>>>
       LoadedSourceFiles;
+
+  /// Function names from -name-whitelist to be used for filtering.
+  std::vector<std::string> FilteredFunctionNamesFromFile;
 };
 }
 

       cl::desc("Show code coverage only for functions with the given name"),
       cl::ZeroOrMore, cl::cat(FilteringCategory));
 
+  cl::list<std::string> NameFilterFiles(
+      "name-whitelist", cl::Optional,
+      cl::desc("Show code coverage only for functions listed in the given "
+               "file"),
+      cl::ZeroOrMore, cl::cat(FilteringCategory));
+
   cl::list<std::string> NameRegexFilters(
       "name-regex", cl::Optional,
       cl::desc("Show code coverage only for functions that match the given "

       ViewOpts.DemanglerOpts.swap(DemanglerOpts);
     }
 
+    // Read in -name-whitelist files.
+    for (const auto &NameFile : NameFilterFiles) {
+      auto BufOrError = MemoryBuffer::getFile(NameFile);
+      if (!BufOrError) {
+        error(NameFile, BufOrError.getError().message());
+        continue;
+      }
+      for (line_iterator LI(**BufOrError, /*SkipBlanks=*/true,
+                            /*CommentMarker=*/'#');
+           !LI.is_at_eof(); ++LI) {
+        auto Name =
+            LI->take_until([](char C) -> bool { return C == '#'; }).trim();
+        if (!Name.empty())
+          FilteredFunctionNamesFromFile.push_back(Name);
+      }
+    }
+
     // Create the function filters
-    if (!NameFilters.empty() || !NameRegexFilters.empty()) {
+    if (!NameFilters.empty() || !FilteredFunctionNamesFromFile.empty() ||
+        !NameRegexFilters.empty()) {
       auto NameFilterer = llvm::make_unique<CoverageFilters>();
       for (const auto &Name : NameFilters)
         NameFilterer->push_back(llvm::make_unique<NameCoverageFilter>(Name));
+      for (const auto &Name : FilteredFunctionNamesFromFile)
+        NameFilterer->push_back(llvm::make_unique<NameCoverageFilter>(Name));
       for (const auto &Regex : NameRegexFilters)
         NameFilterer->push_back(
             llvm::make_unique<NameRegexCoverageFilter>(Regex));