[Driver] Use VFS for GetProgramPath() (NFCI)

[nikic]

Summary

To allow unit testing this code path, switch to using VFS instead of the real file system.

[MaskRay]

Looks great!

clang/lib/Driver/Driver.cpp

5979

Change to lower case while updating the signature.

[nikic]

Unfortunately, it looks like there are legitimate failures of flang tests on Windows in pre-merge checks, thanks to this bit of magic in the Windows implementation of can_execute: https://github.com/llvm/llvm-project/blob/f0687b47a0ce82da07127fee4fe6af801df54ca6/llvm/lib/Support/Windows/Path.inc#L643-L646 So when checking whether flang-new is executable, it will also check for flang-new.exe and we end up returning the former. Now we end up returning the latter. Not sure what to do about this.

[nikic]

I don't see a good way forward here. fs::can_execute has platform specific code. I don't want to replicate its logic here, and making VFS platform-dependent also seems dubious. The cleanest solution would be to make fs::can_execute itself operate on VFS, but I think this doesn't really work with the current layering. There is also a related problem that findProgramByName() isn't VFS-based either, so there remains a partial dependency on the real FS.

clang/lib/Driver/Driver.cpp

 void Driver::generatePrefixedToolNames(
     StringRef Tool, const ToolChain &TC,
     SmallVectorImpl<std::string> &Names) const {
   // FIXME: Needs a better variable than TargetTriple
   Names.emplace_back((TargetTriple + "-" + Tool).str());
   Names.emplace_back(Tool);
 }
 
-static bool ScanDirForExecutable(SmallString<128> &Dir, StringRef Name) {
+static bool canExecute(llvm::vfs::FileSystem &VFS, StringRef Path) {
+  auto Status = VFS.status(Path);
+  return Status &&
+         (Status->getPermissions() & llvm::sys::fs::perms::all_exe) != 0;
+}
+
+static bool ScanDirForExecutable(llvm::vfs::FileSystem &VFS,
+                                 SmallString<128> &Dir, StringRef Name) {
   llvm::sys::path::append(Dir, Name);
-  if (llvm::sys::fs::can_execute(Twine(Dir)))
+  if (canExecute(VFS, Dir))
     return true;
   llvm::sys::path::remove_filename(Dir);
   return false;
 }
 
 std::string Driver::GetProgramPath(StringRef Name, const ToolChain &TC) const {
   SmallVector<std::string, 2> TargetSpecificExecutables;
   generatePrefixedToolNames(Name, TC, TargetSpecificExecutables);
 
   // Respect a limited subset of the '-Bprefix' functionality in GCC by
   // attempting to use this prefix when looking for program paths.
   for (const auto &PrefixDir : PrefixDirs) {
-    if (llvm::sys::fs::is_directory(PrefixDir)) {
+    auto Status = getVFS().status(PrefixDir);
+    if (Status && Status->isDirectory()) {
       SmallString<128> P(PrefixDir);
-      if (ScanDirForExecutable(P, Name))
+      if (ScanDirForExecutable(getVFS(), P, Name))
         return std::string(P.str());
     } else {
       SmallString<128> P((PrefixDir + Name).str());
-      if (llvm::sys::fs::can_execute(Twine(P)))
+      if (canExecute(getVFS(), P))
         return std::string(P.str());
     }
   }
 
   const ToolChain::path_list &List = TC.getProgramPaths();
   for (const auto &TargetSpecificExecutable : TargetSpecificExecutables) {
     // For each possible name of the tool look for it in
     // program paths first, then the path.
     // Higher priority names will be first, meaning that
     // a higher priority name in the path will be found
     // instead of a lower priority name in the program path.
     // E.g. <triple>-gcc on the path will be found instead
     // of gcc in the program path
     for (const auto &Path : List) {
       SmallString<128> P(Path);
-      if (ScanDirForExecutable(P, TargetSpecificExecutable))
+      if (ScanDirForExecutable(getVFS(), P, TargetSpecificExecutable))
         return std::string(P.str());
     }
 
     // Fall back to the path
     if (llvm::ErrorOr<std::string> P =
             llvm::sys::findProgramByName(TargetSpecificExecutable))
       return *P;
   }