[UpdateCCTestChecks] Detect function mangled name on separate line

[greened]

Summary

Sometimes functions with large comment blocks in front of them have their declarations output on several lines by c-index-test. Hence the one-line function name/line/mangled pattern will not work to detect them. Break the pattern up into two patterns and keep state after seeing the name/line information until we finally see the mangled name.

[MaskRay]

Can you give an example demonstrating the issue?

[greened]

In D68272#1690823, @MaskRay wrote:

Can you give an example demonstrating the issue?

This test:

/***********************************/
/*                                 */
/* A test.                         */
/*                                 */
/***********************************/
int foo(void) {
  return 1;
}

c-index-test -write-pch test.pch test.c
c-index-test -test-print-mangle test.pch

Produces this:

FunctionDecl=foo:6:5 (Definition) RawComment=[/***********************************/] RawCommentRange=[5:1 - 5:38] BriefComment=[********************************] FullCommentAsHTML=[<p class="para-brief">********************************</p>] FullCommentAsXML=[<Function file="/ptmp/dag/test/test.c" line="6" column="5"><Name>foo</Name><USR>c:@F@foo</USR><Declaration>int foo()</Declaration><Abstract><Para>********************************</Para></Abstract></Function>]
// CHECK:  CommentAST=[
// CHECK:    (CXComment_FullComment
// CHECK:       (CXComment_Paragraph
// CHECK:         (CXComment_Text Text=[********************************])))] [mangled=foo]

Note that mangled=foo appears on a separate line.

llvm/utils/update_cc_test_checks.py

status = subprocess.run([args.c_index_test, '-write-pch', f.name, *clang_args],
                             stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
     if status.returncode:
       sys.stderr.write(status.stdout.decode())
       sys.exit(2)
     output = subprocess.check_output([args.c_index_test,
         '-test-print-mangle', f.name])
     if sys.version_info[0] > 2:
       output = output.decode()
-
-  RE = re.compile(r'^FunctionDecl=(\w+):(\d+):\d+ \(Definition\) \[mangled=([^]]+)\]')
+  DeclRE = re.compile(r'^FunctionDecl=(\w+):(\d+):\d+ \(Definition\)')
+  MangleRE = re.compile(r'.*\[mangled=([^]]+)\]')
+  MatchedDecl = False
   for line in output.splitlines():
-    m = RE.match(line)
-    if not m: continue
-    spell, line, mangled = m.groups()
+    # Get the function source name, line number and mangled name.  Sometimes
+    # c-index-test outputs the mangled name on a separate line (this can happen
+    # with block comments in front of functions).  Keep scanning until we see
+    # the mangled name.
+    decl_m = DeclRE.match(line)
+    mangle_m = MangleRE.match(line)
+
+    if decl_m:
+      MatchedDecl = True
+      spell, lineno = decl_m.groups()
+    if MatchedDecl and mangle_m:
+      mangled = mangle_m.group(1)
+      MatchedDecl = False
+    else:
+      continue
+
     if mangled == '_' + spell:
       # HACK for MacOS (where the mangled name includes an _ for C but the IR won't):
       mangled = spell
     # Note -test-print-mangle does not print file names so if #include is used,
     # the line number may come from an included file.
-    ret[int(line)-1] = (spell, mangled)
+    ret[int(lineno)-1] = (spell, mangled)
   if args.verbose:
     for line, func_name in sorted(ret.items()):
       print('line {}: found function {}'.format(line+1, func_name), file=sys.stderr)
   return ret
 
 
 def config():
   parser = argparse.ArgumentParser(