[clang] [clang][DependencyScanning] Implementation of `CompilerInstanceWithContext` to Improve By-Name Queries (PR #164345)

[Ben Langmuir via cfe-commits]

================
@@ -704,3 +703,163 @@ bool DependencyScanningAction::runInvocation(
 
   return Result;
 }
+
+const std::string CompilerInstanceWithContext::FakeFileBuffer =
+    std::string(MAX_NUM_NAMES, ' ');
+
+llvm::Error CompilerInstanceWithContext::initialize() {
+  std::tie(OverlayFS, CommandLine) = initVFSForByNameScanning(
+      Worker.BaseFS, CommandLine, CWD, "ScanningByName");
+
+  DiagPrinterWithOS =
+      std::make_unique<TextDiagnosticsPrinterWithOutput>(CommandLine);
+  DiagEngineWithCmdAndOpts = std::make_unique<DignosticsEngineWithDiagOpts>(
+      CommandLine, OverlayFS, DiagPrinterWithOS->DiagPrinter);
+
+  std::tie(Driver, Compilation) = buildCompilation(
+      CommandLine, *DiagEngineWithCmdAndOpts->DiagEngine, OverlayFS, Alloc);
+
+  if (!Compilation) {
+    return llvm::make_error<llvm::StringError>("Failed to build compilation",
+                                               llvm::inconvertibleErrorCode());
+  }
+
+  assert(Compilation->getJobs().size() &&
+         "Must have a job list of non-zero size");
+  const driver::Command &Command = *(Compilation->getJobs().begin());
+  const auto &CommandArgs = Command.getArguments();
+  size_t ArgSize = CommandArgs.size();
+  assert(ArgSize >= 1 && "Cannot have a command with 0 args");
+  const char *FirstArg = CommandArgs[0];
+  if (StringRef(FirstArg) != "-cc1")
+    return llvm::make_error<llvm::StringError>(
+        "Incorrect compilation command, missing cc1",
+        llvm::inconvertibleErrorCode());
+  OriginalInvocation = std::make_unique<CompilerInvocation>();
+
+  if (!CompilerInvocation::CreateFromArgs(
----------------
benlangmuir wrote:

It seems like we have a fair bit of duplication here with the code in `createCompilerInvocation`, `forEachDriverJob`, and `scanDependencies`.  Have you considered whether we can refactor some of this? It doesn't necessarily need to happen in this PR, but it would be helpful to understand where you think it should end up.

https://github.com/llvm/llvm-project/pull/164345