[cfe-dev] Support floating point literals with exponent but no point

[Christopher Lamb]

GCC allows literals of the form "1e-16f" which clang currently  
rejects. The following patch is proposed:

Index: Lex/LiteralSupport.cpp
===================================================================
--- Lex/LiteralSupport.cpp      (revision 44312)
+++ Lex/LiteralSupport.cpp      (working copy)
@@ -261,7 +261,7 @@
        s = SkipOctalDigits(s);
        if (s == ThisTokEnd) {
          // Done.
-      } else if (isxdigit(*s)) {
+      } else if (isxdigit(*s) && !(*s == 'e' || *s == 'E')) {
          TokLoc = PP.AdvanceToTokenCharacter(TokLoc, s-begin);
          Diag(TokLoc, diag::err_invalid_octal_digit, std::string(s, s 
+1));
          return;
@@ -290,7 +290,7 @@
      s = SkipDigits(s);
      if (s == ThisTokEnd) {
        // Done.
-    } else if (isxdigit(*s)) {
+    } else if (isxdigit(*s) && !(*s == 'e' || *s == 'E')) {
        Diag(TokLoc, diag::err_invalid_decimal_digit, std::string(s, s 
+1));
        return;
      } else if (*s == '.') {


With testcase:

float  F  = 1e-19f;
double D  = 2e32;
float  F2 = 01e-19f;
double D2 = 02e32;

--
Christopher Lamb