clang-format: Fix another assertion discovered by the fuzzer.

In the process, fix an old todo that I don't really know how to write
tests for. The problem is that Clang's lexer creates very strange token
sequences for these. However, the new approach seems generally better
and easier to read so I am submitting it nonetheless.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@236589 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Format/Format.cpp b/lib/Format/Format.cpp
index 9aad4f7d0e2cc11965ff687c191fcce201065313..8348a535a43e8e9d699b4c36381cbb332b7514e5 100644 (file)
--- a/lib/Format/Format.cpp
+++ b/lib/Format/Format.cpp
@@ -1000,18 +1000,25 @@ private:
     // Consume and record whitespace until we find a significant token.
     unsigned WhitespaceLength = TrailingWhitespace;
     while (FormatTok->Tok.is(tok::unknown)) {
+      StringRef Text = FormatTok->TokenText;
+      auto EscapesNewline = [&](int pos) {
+        // A '\r' here is just part of '\r\n'. Skip it.
+        if (pos >= 0 && Text[pos] == '\r')
+          --pos;
+        // See whether there is an odd number of '\' before this.
+        unsigned count = 0;
+        for (; pos >= 0; --pos, ++count)
+          if (Text[count] != '\\')
+            break;
+        return count & 1;
+      };
       // FIXME: This miscounts tok:unknown tokens that are not just
       // whitespace, e.g. a '`' character.
-      for (int i = 0, e = FormatTok->TokenText.size(); i != e; ++i) {
-        switch (FormatTok->TokenText[i]) {
+      for (int i = 0, e = Text.size(); i != e; ++i) {
+        switch (Text[i]) {
         case '\n':
           ++FormatTok->NewlinesBefore;
-          // FIXME: This is technically incorrect, as it could also
-          // be a literal backslash at the end of the line.
-          if (i == 0 || (FormatTok->TokenText[i - 1] != '\\' &&
-                         (FormatTok->TokenText[i - 1] != '\r' || i == 1 ||
-                          FormatTok->TokenText[i - 2] != '\\')))
-            FormatTok->HasUnescapedNewline = true;
+          FormatTok->HasUnescapedNewline = !EscapesNewline(i - 1);
           FormatTok->LastNewlineOffset = WhitespaceLength + i + 1;
           Column = 0;
           break;
@@ -1030,8 +1037,7 @@ private:
           Column += Style.TabWidth - Column % Style.TabWidth;
           break;
         case '\\':
-          if (i + 1 == e || (FormatTok->TokenText[i + 1] != '\r' &&
-                             FormatTok->TokenText[i + 1] != '\n'))
+          if (i + 1 == e || (Text[i + 1] != '\r' && Text[i + 1] != '\n'))
             FormatTok->Type = TT_ImplicitStringLiteral;
           break;
         default:
@@ -1056,6 +1062,7 @@ private:
            FormatTok->TokenText[1] == '\n') {
       ++FormatTok->NewlinesBefore;
       WhitespaceLength += 2;
+      FormatTok->LastNewlineOffset = 2;
       Column = 0;
       FormatTok->TokenText = FormatTok->TokenText.substr(2);
     }

diff --git a/unittests/Format/FormatTest.cpp b/unittests/Format/FormatTest.cpp
index 31befad678337efde802548ee935f792280e408e..ea15312eb5c72d5f3a191843c4a1adbbfc2e1fd1 100644 (file)
--- a/unittests/Format/FormatTest.cpp
+++ b/unittests/Format/FormatTest.cpp
@@ -2951,15 +2951,13 @@ TEST_F(FormatTest, FormatUnbalancedStructuralElements) {
             format("#define A } }\nint i;", getLLVMStyleWithColumns(11)));
 }
 
-TEST_F(FormatTest, EscapedNewlineAtStartOfToken) {
+TEST_F(FormatTest, EscapedNewlines) {
   EXPECT_EQ(
       "#define A \\\n  int i;  \\\n  int j;",
       format("#define A \\\nint i;\\\n  int j;", getLLVMStyleWithColumns(11)));
   EXPECT_EQ("template <class T> f();", format("\\\ntemplate <class T> f();"));
-}
-
-TEST_F(FormatTest, NoEscapedNewlineHandlingInBlockComments) {
   EXPECT_EQ("/* \\  \\  \\\n*/", format("\\\n/* \\  \\  \\\n*/"));
+  EXPECT_EQ("<a\n\\\\\n>", format("<a\n\\\\\n>"));
 }
 
 TEST_F(FormatTest, DontCrashOnBlockComments) {