Fix unintended lstrip_blocks behavior. Fixes pallets#1138

Introduced in pallets#858. Tests will follow, also results of performance
testing.

Author: Peter Dolak

src/jinja2/lexer.py

@@ -637,6 +637,7 @@ def tokeniter(self, source, name, filename=None, state=None):
         balancing_stack = []
         lstrip_unless_re = self.lstrip_unless_re
         newlines_stripped = 0
+        line_starting = True
 
         while 1:
             # tokenizer loop
@@ -686,11 +687,11 @@ def tokeniter(self, source, name, filename=None, state=None):
                         ):
                             # The start of text between the last newline and the tag.
                             l_pos = text.rfind("\n") + 1
-
-                            # If there's only whitespace between the newline and the
-                            # tag, strip it.
-                            if not lstrip_unless_re.search(text, l_pos):
-                                groups = (text[:l_pos],) + groups[1:]
+                            if l_pos > 0 or line_starting:
+                                # If there's only whitespace between the newline and the
+                                # tag, strip it.
+                                if not lstrip_unless_re.search(text, l_pos):
+                                    groups = (text[:l_pos],) + groups[1:]
 
                     for idx, token in enumerate(tokens):
                         # failure group
@@ -747,6 +748,8 @@ def tokeniter(self, source, name, filename=None, state=None):
                         yield lineno, tokens, data
                     lineno += data.count("\n")
 
+                line_starting = m.group()[-1:] == "\n"
+
                 # fetch new position into new variable so that we can check
                 # if there is a internal parsing error which would result
                 # in an infinite loop