trim only ascii whitespace for branch (#2521)

* trim only ascii whitespace for branch

* rebuild

3 files changed, 80 insertions, 8 deletions

diff --git a/__test__/input-helper.test.ts b/__test__/input-helper.test.ts
index 27b023c..afe30fb 100644
--- a/__test__/input-helper.test.ts
+++ b/__test__/input-helper.test.ts
@@ -24,9 +24,20 @@ const mockGithubContext: any = {
   payload: {}
 }
 
+// Replicate @actions/core getInput behavior: it trims whitespace by default
+// (String.prototype.trim(), which strips characters such as a leading U+FEFF BOM)
+// unless trimWhitespace is explicitly set to false.
+const getInputImpl = (name: string, options?: {trimWhitespace?: boolean}) => {
+  const val = inputs[name] ?? ''
+  if (options && options.trimWhitespace === false) {
+    return val
+  }
+  return typeof val === 'string' ? val.trim() : val
+}
+
 // Mock @actions/core before loading input-helper
 jest.unstable_mockModule('@actions/core', () => ({
-  getInput: jest.fn((name: string) => inputs[name]),
+  getInput: jest.fn(getInputImpl),
   getBooleanInput: jest.fn((name: string) => inputs[name]),
   getMultilineInput: jest.fn((name: string) =>
     inputs[name] ? String(inputs[name]).split('\n').filter(Boolean) : []
@@ -76,9 +87,7 @@ describe('input-helper tests', () => {
     inputs = {}
     jest.clearAllMocks()
     // Re-apply default mocks
-    ;(core.getInput as jest.Mock<any>).mockImplementation(
-      (name: string) => inputs[name]
-    )
+    ;(core.getInput as jest.Mock<any>).mockImplementation(getInputImpl as any)
     mockDirectoryExistsSync.mockImplementation(
       (p: string) => p === gitHubWorkspace
     )
@@ -176,6 +185,36 @@ describe('input-helper tests', () => {
     expect(settings.commit).toBeFalsy()
   })
 
+  it('does not reclassify a ref as sha when a BOM is prefixed', async () => {
+    // A fork branch named "<U+FEFF>" + 40 hex chars. core.getInput trims the
+    // BOM by default, which previously collapsed this into a bare SHA and
+    // bypassed the unsafe fork PR checkout guard.
+    inputs.ref = '\uFEFF522d932fae5296da51fdf431934425ecf891c6a2'
+    const settings: IGitSourceSettings = await inputHelper.getInputs()
+    expect(settings.commit).toBeFalsy()
+    expect(settings.ref).toBe('522d932fae5296da51fdf431934425ecf891c6a2')
+  })
+
+  it('does not reclassify a sha-256 ref as sha when a BOM is prefixed', async () => {
+    inputs.ref =
+      '\uFEFF1111111111222222222233333333334444444444555555555566666666667777'
+    const settings: IGitSourceSettings = await inputHelper.getInputs()
+    expect(settings.commit).toBeFalsy()
+    expect(settings.ref).toBe(
+      '1111111111222222222233333333334444444444555555555566666666667777'
+    )
+  })
+
+  it('treats a sha surrounded by ascii whitespace as a commit', async () => {
+    // ASCII whitespace can only come from the workflow author's YAML (git ref
+    // names cannot contain it), so trimming it and treating the value as a
+    // commit is safe.
+    inputs.ref = '  1111111111222222222233333333334444444444  '
+    const settings: IGitSourceSettings = await inputHelper.getInputs()
+    expect(settings.ref).toBeFalsy()
+    expect(settings.commit).toBe('1111111111222222222233333333334444444444')
+  })
+
   it('sets workflow organization ID', async () => {
     const settings: IGitSourceSettings = await inputHelper.getInputs()
     expect(settings.workflowOrganizationId).toBe(123456)
diff --git a/dist/index.js b/dist/index.js
index cc25215..b319687 100644
--- a/dist/index.js
+++ b/dist/index.js
@@ -42100,6 +42100,22 @@ async function getInputs() {
         `${github_context.repo.owner}/${github_context.repo.repo}`.toUpperCase();
     // Source branch, source version
     result.ref = getInput('ref');
+    // core.getInput()'s default trim strips a range of Unicode characters such as a
+    // leading BOM (U+FEFF) or NBSP (U+00A0). Those are valid in a git ref name, so
+    // a fork branch named "<BOM>" + 40 hex chars would trim down to a bare SHA and
+    // be silently reclassified as a commit, bypassing the unsafe fork PR checkout
+    // guard.
+    //
+    // The trim below strips only the ASCII whitespace characters which are all forbidden
+    // in a git branch name.
+    //   \t  U+0009  horizontal tab   - ASCII control, forbidden in ref names
+    //   \n  U+000A  line feed        - ASCII control, forbidden in ref names
+    //   \v  U+000B  vertical tab     - ASCII control, forbidden in ref names
+    //   \f  U+000C  form feed        - ASCII control, forbidden in ref names
+    //   \r  U+000D  carriage return  - ASCII control, forbidden in ref names
+    //   ' ' U+0020  space            - forbidden in ref names
+    const asciiTrimmedRef = getInput('ref', { trimWhitespace: false })
+        .replace(/^[\t\n\v\f\r ]+|[\t\n\v\f\r ]+$/g, '');
     if (!result.ref) {
         if (isWorkflowRepository) {
             result.ref = github_context.ref;
@@ -42112,8 +42128,8 @@ async function getInputs() {
         }
     }
     // SHA?
-    else if (result.ref.match(/^(?:[0-9a-fA-F]{40}|[0-9a-fA-F]{64})$/)) {
-        result.commit = result.ref;
+    else if (asciiTrimmedRef.match(/^(?:[0-9a-fA-F]{40}|[0-9a-fA-F]{64})$/)) {
+        result.commit = asciiTrimmedRef;
         result.ref = '';
     }
     core_debug(`ref = '${result.ref}'`);
diff --git a/src/input-helper.ts b/src/input-helper.ts
index 87b2800..9a98b86 100644
--- a/src/input-helper.ts
+++ b/src/input-helper.ts
@@ -59,6 +59,23 @@ export async function getInputs(): Promise<IGitSourceSettings> {
 
   // Source branch, source version
   result.ref = core.getInput('ref')
+  // core.getInput()'s default trim strips a range of Unicode characters such as a
+  // leading BOM (U+FEFF) or NBSP (U+00A0). Those are valid in a git ref name, so
+  // a fork branch named "<BOM>" + 40 hex chars would trim down to a bare SHA and
+  // be silently reclassified as a commit, bypassing the unsafe fork PR checkout
+  // guard.
+  //
+  // The trim below strips only the ASCII whitespace characters which are all forbidden
+  // in a git branch name.
+  //   \t  U+0009  horizontal tab   - ASCII control, forbidden in ref names
+  //   \n  U+000A  line feed        - ASCII control, forbidden in ref names
+  //   \v  U+000B  vertical tab     - ASCII control, forbidden in ref names
+  //   \f  U+000C  form feed        - ASCII control, forbidden in ref names
+  //   \r  U+000D  carriage return  - ASCII control, forbidden in ref names
+  //   ' ' U+0020  space            - forbidden in ref names
+  const asciiTrimmedRef = core
+    .getInput('ref', {trimWhitespace: false})
+    .replace(/^[\t\n\v\f\r ]+|[\t\n\v\f\r ]+$/g, '')
   if (!result.ref) {
     if (isWorkflowRepository) {
       result.ref = github.context.ref
@@ -72,8 +89,8 @@ export async function getInputs(): Promise<IGitSourceSettings> {
     }
   }
   // SHA?
-  else if (result.ref.match(/^(?:[0-9a-fA-F]{40}|[0-9a-fA-F]{64})$/)) {
-    result.commit = result.ref
+  else if (asciiTrimmedRef.match(/^(?:[0-9a-fA-F]{40}|[0-9a-fA-F]{64})$/)) {
+    result.commit = asciiTrimmedRef
     result.ref = ''
   }
   core.debug(`ref = '${result.ref}'`)