Fix issue with windows newlines not being normalised (fixes #27) (#30)

This PR fixes #27 

It also contains a small additonal fix for #29, where `_` and `-` are
also not allowed before a check prefix.

filecheck/finput.py

@@ -125,6 +125,10 @@ class FInput:
     range: InputRange = field(default_factory=lambda: InputRange(0, sys.maxsize))
     ranges: list[InputRange] = field(default_factory=list)
 
+    @staticmethod
+    def canonicalize_line_ends(text: str) -> str:
+        return text.replace("\r\n", "\n")
+
     @staticmethod
     def from_opts(opts: Options) -> FInput:
         """
@@ -135,7 +139,7 @@ def from_opts(opts: Options) -> FInput:
             f = sys.stdin
         else:
             f = open(opts.input_file, "r")
-        return FInput(opts.input_file, f.read())
+        return FInput(opts.input_file, FInput.canonicalize_line_ends(f.read()))
 
     def advance_by(self, dist: int):
         """

filecheck/parser.py

@@ -19,7 +19,7 @@
 def pattern_for_opts(opts: Options) -> tuple[re.Pattern[str], re.Pattern[str]]:
     prefixes = f"({'|'.join(map(re.escape, opts.check_prefixes))})"
     return re.compile(
-        r"(^|[^a-zA-Z])"
+        r"(^|[^a-zA-Z-_])"
         + prefixes
         + r"(-(DAG|COUNT-\d+|NOT|EMPTY|NEXT|SAME|LABEL))?(\{LITERAL})?:\s?([^\n]*)\n?"
     ), re.compile(f"({'|'.join(map(re.escape, opts.comment_prefixes))}).*{prefixes}")

tests/filecheck/windows_line_endings.test

@@ -0,0 +1,5 @@
+// RUN: echo -e "1\r\n2\r\n\r\n3" | filecheck %s
+CHECK: 1
+CHECK-NEXT: 2
+CHECK-EMPTY:
+CHECK-NEXT: 3