Overlay databases: use --overlay-file-list

This commit changes overlay database creation to use the
--overlay-file-list flag. It also implements Git-based file change
detection to generate the list of files to extract for the overlay
database.

Author: cklin

lib/codeql.js

@@ -24,7 +24,11 @@ import {
 import { isAnalyzingDefaultBranch } from "./git-utils";
 import { Language } from "./languages";
 import { Logger } from "./logging";
-import { OverlayDatabaseMode } from "./overlay-database-utils";
+import {
+  OverlayDatabaseMode,
+  writeBaseDatabaseOidsFile,
+  writeOverlayChangesFile,
+} from "./overlay-database-utils";
 import * as setupCodeql from "./setup-codeql";
 import { ZstdAvailability } from "./tar";
 import { ToolsDownloadStatusReport } from "./tools-download";
@@ -610,7 +614,12 @@ export async function getCodeQLForCmd(
         : "--overwrite";
 
       if (overlayDatabaseMode === OverlayDatabaseMode.Overlay) {
-        extraArgs.push("--overlay");
+        const overlayChangesFile = await writeOverlayChangesFile(
+          config,
+          sourceRoot,
+          logger,
+        );
+        extraArgs.push(`--overlay-file-list=${overlayChangesFile}`);
       } else if (overlayDatabaseMode === OverlayDatabaseMode.OverlayBase) {
         extraArgs.push("--overlay-base");
       }
@@ -636,6 +645,10 @@ export async function getCodeQLForCmd(
         ],
         { stdin: externalRepositoryToken },
       );
+
+      if (overlayDatabaseMode === OverlayDatabaseMode.OverlayBase) {
+        await writeBaseDatabaseOidsFile(config, sourceRoot);
+      }
     },
     async runAutobuild(config: Config, language: Language) {
       applyAutobuildAzurePipelinesTimeoutFix();

lib/codeql.js.map

@@ -10,7 +10,7 @@ import {
 import { ConfigurationError, getRequiredEnvParam } from "./util";
 
 async function runGitCommand(
-  checkoutPath: string | undefined,
+  workingDirectory: string | undefined,
   args: string[],
   customErrorMessage: string,
 ): Promise<string> {
@@ -28,7 +28,7 @@ async function runGitCommand(
           stderr += data.toString();
         },
       },
-      cwd: checkoutPath,
+      cwd: workingDirectory,
     }).exec();
     return stdout;
   } catch (error) {
@@ -322,6 +322,44 @@ export const getGitRoot = async function (
   }
 };
 
+/**
+ * Returns the Git OIDs at `HEAD` of all tracked files that are the given base
+ * path. Untracked files and files not under the given base path are ignored.
+ *
+ * @param basePath A path into the Git repository.
+ * @returns a map from file paths (relative to `basePath`) to Git OIDs.
+ * @throws {Error} if "git ls-tree" produces unexpected output.
+ */
+export const getFileOidsUnderPath = async function (
+  basePath: string,
+): Promise<{ [key: string]: string }> {
+  // Without the --full-name flag, the path is relative to the current working
+  // directory of the git command, which is basePath.
+  // Without the --full-tree flag, the file listing is confined to the current
+  // working directory of the git command, which is basePath.
+  const stdout = await runGitCommand(
+    basePath,
+    ["ls-tree", "--format=%(objectname)_%(path)", "-r", "HEAD"],
+    "Cannot list file OIDs in HEAD.",
+  );
+
+  const fileOidMap: { [key: string]: string } = {};
+  const regex = /^([0-9a-f]{40})_(.+)$/;
+  for (const line of stdout.split("\n")) {
+    if (line) {
+      const match = line.match(regex);
+      if (match) {
+        const oid = match[1];
+        const path = decodeGitFilePath(match[2]);
+        fileOidMap[path] = oid;
+      } else {
+        throw new Error(`Unexpected "git ls-tree" output: ${line}`);
+      }
+    }
+  }
+  return fileOidMap;
+};
+
 function getRefFromEnv(): string {
   // To workaround a limitation of Actions dynamic workflows not setting
   // the GITHUB_REF in some cases, we accept also the ref within the