fix: ignore hh_client output during extension init

src/LSPHackTypeChecker.ts

@@ -46,6 +46,8 @@
       }),
     );
 
+    // `hh_client --lsp` fails to spawn `hh_server` if not already running,
+    // so do an initial `hh_client` call to start the server if needed.
     await vscode.window.withProgress(
       {
         location: vscode.ProgressLocation.Window,
@@ -115,7 +117,7 @@
     if (
       this.config.enableCoverageCheck &&
       languageClient.initializeResult &&
       (<any>languageClient.initializeResult.capabilities).typeCoverageProvider
     ) {
       await new HackCoverageChecker(languageClient).start(context);
     }

src/proxy.ts

@@ -19,11 +19,14 @@
   }
 }
 
+/** Run `hh_client` to start `hh_server` if needed. */
 export async function start(
   config: HackConfig,
   log: vscode.LogOutputChannel,
 ): Promise<string> {
-  return run(config, log, []);
+  // Ignore errors to avoid exceeding `maxBuffer` in case of many preexisting errors.
+  // We only run this command to ensure `hh_server` starts and don't care about the output.
+  return run(config, log, ["--max-errors", "0"]);
 }
 
 async function run(
@@ -48,7 +51,7 @@
       command,
       args,
       { maxBuffer: 1024 * 1024 },
       (err: any, stdout, stderr) => {
         if (!stdout) {
           // all hh_client --check output goes to stderr by default
           stdout = stderr;