Make line numbers signify argument numbers in buildspec

4 files changed, 23 insertions, 12 deletions

diff --git a/build2/b.cxx b/build2/b.cxx
index 1f0464f..17a6645 100644
--- a/build2/b.cxx
+++ b/build2/b.cxx
@@ -203,16 +203,11 @@ main (int argc, char* argv[])
         }
 
         // Merge all the individual buildspec arguments into a single string.
-        // Instead, we could also parse them individually (and merge the
-        // result). The benefit of doing it this way is potentially better
-        // diagnostics (i.e., we could have used <buildspec-1>, <buildspec-2>
-        // to give the idea about which argument is invalid).
-        //
-        // Or we could separate arguments with newlines so that a line number
-        // signifies the argument number.
+        // We wse newlines to separate arguments so that line numbers in
+        // diagnostics signify argument numbers. Clever, huh?
         //
         if (argn != 0)
-          args += ' ';
+          args += '\n';
 
         args += s;
 
diff --git a/build2/lexer b/build2/lexer
index aac1c40..76308e4 100644
--- a/build2/lexer
+++ b/build2/lexer
@@ -121,7 +121,8 @@ namespace build2
       lexer_mode mode;
 
       char sep_pair;
-      bool sep_space; // Are whitespaces separators (see skip_spaces())?
+      bool sep_space;   // Are whitespaces separators (see skip_spaces())?
+      bool sep_newline; // Is newline special (see skip_spaces())?
       bool quotes;    // Recognize quoted fragments.
 
       const char* escapes; // Effective escape sequences to recognize.
diff --git a/build2/lexer.cxx b/build2/lexer.cxx
index 5989548..d3afdb0 100644
--- a/build2/lexer.cxx
+++ b/build2/lexer.cxx
@@ -26,6 +26,7 @@ namespace build2
     const char* s1 (nullptr);
     const char* s2 (nullptr);
     bool s (true);
+    bool n (true);
     bool q (true);
 
     if (!esc)
@@ -70,8 +71,11 @@ namespace build2
         //
         // 2. Recognizes comma.
         //
+        // 3. Treat newline as an ordinary space.
+        //
         s1 = " $(){}[],\t\n";
         s2 = "           ";
+        n = false;
         break;
       }
     case lexer_mode::single_quoted:
@@ -87,7 +91,7 @@ namespace build2
     default: assert (false); // Unhandled custom mode.
     }
 
-    state_.push (state {m, ps, s, q, *esc, s1, s2});
+    state_.push (state {m, ps, s, n, q, *esc, s1, s2});
   }
 
   token lexer::
@@ -585,9 +589,11 @@ namespace build2
     bool r (sep_);
     sep_ = false;
 
+    const state& s (state_.top ());
+
     // In some special modes we don't skip spaces.
     //
-    if (!state_.top ().sep_space)
+    if (!s.sep_space)
       return r;
 
     xchar c (peek ());
@@ -605,6 +611,14 @@ namespace build2
         }
       case '\n':
         {
+          // In some modes we treat newlines as ordinary spaces.
+          //
+          if (!s.sep_newline)
+          {
+            r = true;
+            break;
+          }
+
           // Skip empty lines.
           //
           if (start)
diff --git a/build2/test/script/lexer.cxx b/build2/test/script/lexer.cxx
index c7f9193..fab5cc2 100644
--- a/build2/test/script/lexer.cxx
+++ b/build2/test/script/lexer.cxx
@@ -22,6 +22,7 @@ namespace build2
         const char* s1 (nullptr);
         const char* s2 (nullptr);
         bool s (true);
+        bool n (true);
         bool q (true);
 
         if (!esc)
@@ -135,7 +136,7 @@ namespace build2
         }
 
         assert (ps == '\0');
-        state_.push (state {m, ps, s, q, *esc, s1, s2});
+        state_.push (state {m, ps, s, n, q, *esc, s1, s2});
       }
 
       token lexer::