we can now self-host...almost
[jackhill/mal.git] / runtest.py
index e90fc68..c5b1f42 100755 (executable)
@@ -1,15 +1,34 @@
 #!/usr/bin/env python
 
+from __future__ import print_function
 import os, sys, re
 import argparse, time
+import signal, atexit
 
-import pty
 from subprocess import Popen, STDOUT, PIPE
 from select import select
 
-# TODO: do we need to support '\n' too
-sep = "\r\n"
-#sep = "\n"
+# Pseudo-TTY and terminal manipulation
+import pty, array, fcntl, termios
+
+IS_PY_3 = sys.version_info[0] == 3
+
+debug_file = None
+log_file = None
+
+def debug(data):
+    if debug_file:
+        debug_file.write(data)
+        debug_file.flush()
+
+def log(data, end='\n'):
+    if log_file:
+        log_file.write(data + end)
+        log_file.flush()
+    print(data, end=end)
+    sys.stdout.flush()
+
+sep = "\n"
 rundir = None
 
 parser = argparse.ArgumentParser(
@@ -22,27 +41,72 @@ parser.add_argument('--test-timeout', default=20, type=int,
         help="default timeout for each individual test action")
 parser.add_argument('--pre-eval', default=None, type=str,
         help="Mal code to evaluate prior to running the test")
-parser.add_argument('--redirect', action='store_true',
-        help="Run implementation in bash and redirect output to /dev/null")
+parser.add_argument('--no-pty', action='store_true',
+        help="Use direct pipes instead of pseudo-tty")
+parser.add_argument('--log-file', type=str,
+        help="Write messages to the named file in addition the screen")
+parser.add_argument('--debug-file', type=str,
+        help="Write all test interaction the named file")
+parser.add_argument('--hard', action='store_true',
+        help="Turn soft tests (soft, deferrable, optional) into hard failures")
 
-parser.add_argument('test_file', type=argparse.FileType('r'),
+# Control whether deferrable and optional tests are executed
+parser.add_argument('--deferrable', dest='deferrable', action='store_true',
+        help="Enable deferrable tests that follow a ';>>> deferrable=True'")
+parser.add_argument('--no-deferrable', dest='deferrable', action='store_false',
+        help="Disable deferrable tests that follow a ';>>> deferrable=True'")
+parser.set_defaults(deferrable=True)
+parser.add_argument('--optional', dest='optional', action='store_true',
+        help="Enable optional tests that follow a ';>>> optional=True'")
+parser.add_argument('--no-optional', dest='optional', action='store_false',
+        help="Disable optional tests that follow a ';>>> optional=True'")
+parser.set_defaults(optional=True)
+
+parser.add_argument('test_file', type=str,
         help="a test file formatted as with mal test data")
 parser.add_argument('mal_cmd', nargs="*",
         help="Mal implementation command line. Use '--' to "
              "specify a Mal command line with dashed options.")
+parser.add_argument('--crlf', dest='crlf', action='store_true',
+        help="Write \\r\\n instead of \\n to the input")
 
 class Runner():
-    def __init__(self, args, redirect=False):
-        print "args: %s" % repr(args)
-        if redirect:
-            print "using redirect"
-            self.p = Popen(args, bufsize=0, stdin=PIPE, stdout=PIPE, stderr=STDOUT)
+    def __init__(self, args, no_pty=False, line_break="\n"):
+        #print "args: %s" % repr(args)
+        self.no_pty = no_pty
+
+        # Cleanup child process on exit
+        atexit.register(self.cleanup)
+
+        self.p = None
+        env = os.environ
+        env['TERM'] = 'dumb'
+        env['INPUTRC'] = '/dev/null'
+        env['PERL_RL'] = 'false'
+        if no_pty:
+            self.p = Popen(args, bufsize=0,
+                           stdin=PIPE, stdout=PIPE, stderr=STDOUT,
+                           preexec_fn=os.setsid,
+                           env=env)
             self.stdin = self.p.stdin
             self.stdout = self.p.stdout
         else:
             # provide tty to get 'interactive' readline to work
             master, slave = pty.openpty()
-            self.p = Popen(args, bufsize=0, stdin=slave, stdout=slave, stderr=STDOUT)
+
+            # Set terminal size large so that readline will not send
+            # ANSI/VT escape codes when the lines are long.
+            buf = array.array('h', [100, 200, 0, 0])
+            fcntl.ioctl(master, termios.TIOCSWINSZ, buf, True)
+
+            self.p = Popen(args, bufsize=0,
+                           stdin=slave, stdout=slave, stderr=STDOUT,
+                           preexec_fn=os.setsid,
+                           env=env)
+            # Now close slave so that we will get an exception from
+            # read when the child exits early
+            # http://stackoverflow.com/questions/11165521
+            os.close(slave)
             self.stdin = os.fdopen(master, 'r+b', 0)
             self.stdout = self.stdin
 
@@ -50,132 +114,231 @@ class Runner():
         self.buf = ""
         self.last_prompt = ""
 
+        self.line_break = line_break
+
     def read_to_prompt(self, prompts, timeout):
         end_time = time.time() + timeout
         while time.time() < end_time:
-            [outs,_,_] = select([self.stdin], [], [], 1)
-            if self.stdin in outs:
-                new_data = self.stdin.read(1)
-                #print "new_data: '%s'" % new_data
-                self.buf += new_data
+            [outs,_,_] = select([self.stdout], [], [], 1)
+            if self.stdout in outs:
+                new_data = self.stdout.read(1)
+                new_data = new_data.decode("utf-8") if IS_PY_3 else new_data
+                #print("new_data: '%s'" % new_data)
+                debug(new_data)
+                # Perform newline cleanup
+                self.buf += new_data.replace("\r", "")
                 for prompt in prompts:
                     regexp = re.compile(prompt)
                     match = regexp.search(self.buf)
                     if match:
                         end = match.end()
-                        buf = self.buf[0:end-len(prompt)]
+                        buf = self.buf[0:match.start()]
                         self.buf = self.buf[end:]
                         self.last_prompt = prompt
                         return buf
         return None
 
-    def write(self, str):
-        self.stdout.write(str)
+    def writeline(self, str):
+        def _to_bytes(s):
+            return bytes(s, "utf-8") if IS_PY_3 else s
+
+        self.stdin.write(_to_bytes(str.replace('\r', '\x16\r') + self.line_break))
 
     def cleanup(self):
+        #print "cleaning up"
         if self.p:
-            self.p.terminate()
+            try:
+                os.killpg(self.p.pid, signal.SIGTERM)
+            except OSError:
+                pass
             self.p = None
 
+class TestReader:
+    def __init__(self, test_file):
+        self.line_num = 0
+        f = open(test_file, newline='') if IS_PY_3 else open(test_file)
+        self.data = f.read().split('\n')
+        self.soft = False
+        self.deferrable = False
+        self.optional = False
+
+    def next(self):
+        self.msg = None
+        self.form = None
+        self.out = ""
+        self.ret = None
+
+        while self.data:
+            self.line_num += 1
+            line = self.data.pop(0)
+            if re.match(r"^\s*$", line):   # blank line
+                continue
+            elif line[0:3] == ";;;":       # ignore comment
+                continue
+            elif line[0:2] == ";;":        # output comment
+                self.msg = line[3:]
+                return True
+            elif line[0:5] == ";>>> ":     # settings/commands
+                settings = {}
+                exec(line[5:], {}, settings)
+                if 'soft' in settings:
+                    self.soft = settings['soft']
+                if 'deferrable' in settings and settings['deferrable']:
+                    self.deferrable = "\nSkipping deferrable and optional tests"
+                    return True
+                if 'optional' in settings and settings['optional']:
+                    self.optional = "\nSkipping optional tests"
+                    return True
+                continue
+            elif line[0:1] == ";":         # unexpected comment
+                raise Exception("Test data error at line %d:\n%s" % (self.line_num, line))
+            self.form = line   # the line is a form to send
+
+            # Now find the output and return value
+            while self.data:
+                line = self.data[0]
+                if line[0:3] == ";=>":
+                    self.ret = line[3:]
+                    self.line_num += 1
+                    self.data.pop(0)
+                    break
+                elif line[0:2] == ";/":
+                    self.out = self.out + line[2:] + sep
+                    self.line_num += 1
+                    self.data.pop(0)
+                else:
+                    self.ret = ""
+                    break
+            if self.ret != None: break
+
+        if self.out[-1:] == sep and not self.ret:
+            # If there is no return value, output should not end in
+            # separator
+            self.out = self.out[0:-1]
+        return self.form
 
 args = parser.parse_args(sys.argv[1:])
-test_data = args.test_file.read().split('\n')
+# Workaround argparse issue with two '--' on command line
+if sys.argv.count('--') > 0:
+    args.mal_cmd = sys.argv[sys.argv.index('--')+1:]
 
 if args.rundir: os.chdir(args.rundir)
 
-r = Runner(args.mal_cmd, redirect=args.redirect)
-
-
-test_idx = 0
-def read_test(data):
-    global test_idx
-    form, output, ret = None, "", None
-    while data:
-        test_idx += 1
-        line = data.pop(0)
-        if re.match(r"^\s*$", line):   # blank line
-            continue
-        elif line[0:3] == ";;;":       # ignore comment
-            continue
-        elif line[0:2] == ";;":        # output comment
-            print line[3:]
-            continue
-        elif line[0:2] == ";":         # unexpected comment
-            print "Test data error at line %d:\n%s" % (test_idx, line)
-            return None, None, None, test_idx
-        form = line   # the line is a form to send
-
-        # Now find the output and return value
-        while data:
-            line = data[0]
-            if line[0:3] == ";=>":
-                ret = line[3:].replace('\\r', '\r').replace('\\n', '\n')
-                test_idx += 1
-                data.pop(0)
-                break
-            elif line[0:2] == "; ":
-                output = output + line[2:] + sep
-                test_idx += 1
-                data.pop(0)
-            else:
-                ret = "*"
-                break
-        if ret: break
+if args.log_file:   log_file   = open(args.log_file, "a")
+if args.debug_file: debug_file = open(args.debug_file, "a")
+
+r = Runner(args.mal_cmd, no_pty=args.no_pty, line_break="\r\n" if args.crlf else "\n")
+t = TestReader(args.test_file)
 
-    return form, output, ret, test_idx
 
-def assert_prompt(timeout):
+def assert_prompt(runner, prompts, timeout):
     # Wait for the initial prompt
-    header = r.read_to_prompt(['user> ', 'mal-user> '], timeout=timeout)
+    header = runner.read_to_prompt(prompts, timeout=timeout)
     if not header == None:
         if header:
-            print "Started with:\n%s" % header
+            log("Started with:\n%s" % header)
     else:
-        print "Did not get 'user> ' or 'mal-user> ' prompt"
-        print "    Got      : %s" % repr(r.buf)
-        r.cleanup()
+        log("Did not receive one of following prompt(s): %s" % repr(prompts))
+        log("    Got      : %s" % repr(r.buf))
         sys.exit(1)
 
 
 # Wait for the initial prompt
-assert_prompt(args.start_timeout)
+try:
+    assert_prompt(r, ['[^\s()<>]+> '], args.start_timeout)
+except:
+    _, exc, _ = sys.exc_info()
+    log("\nException: %s" % repr(exc))
+    log("Output before exception:\n%s" % r.buf)
+    sys.exit(1)
 
 # Send the pre-eval code if any
 if args.pre_eval:
     sys.stdout.write("RUNNING pre-eval: %s" % args.pre_eval)
-    p.write(args.pre_eval)
-    assert_prompt(args.test_timeout)
+    r.writeline(args.pre_eval)
+    assert_prompt(r, ['[^\s()<>]+> '], args.test_timeout)
 
+test_cnt = 0
+pass_cnt = 0
 fail_cnt = 0
+soft_fail_cnt = 0
+failures = []
 
-while test_data:
-    form, out, ret, line_num = read_test(test_data)
-    if form == None:
+while t.next():
+    if args.deferrable == False and t.deferrable:
+        log(t.deferrable)
         break
-    sys.stdout.write("TEST: %s -> [%s,%s]" % (form, repr(out), repr(ret)))
-    sys.stdout.flush()
-    expected = "%s%s%s%s" % (form, sep, out, ret)
 
-    r.write(form + "\n")
+    if args.optional == False and t.optional:
+        log(t.optional)
+        break
+
+    if t.msg != None:
+        log(t.msg)
+        continue
+
+    if t.form == None: continue
+
+    log("TEST: %s -> [%s,%s]" % (repr(t.form), repr(t.out), t.ret), end='')
+
+    # The repeated form is to get around an occasional OS X issue
+    # where the form is repeated.
+    # https://github.com/kanaka/mal/issues/30
+    expects = [".*%s%s%s" % (sep, t.out, re.escape(t.ret)),
+               ".*%s.*%s%s%s" % (sep, sep, t.out, re.escape(t.ret))]
+
+    r.writeline(t.form)
     try:
-        res = r.read_to_prompt(['\r\nuser> ', '\nuser> ',
-                                '\r\nmal-user> ', '\nmal-user> '],
+        test_cnt += 1
+        res = r.read_to_prompt(['\r\n[^\s()<>]+> ', '\n[^\s()<>]+> '],
                                 timeout=args.test_timeout)
         #print "%s,%s,%s" % (idx, repr(p.before), repr(p.after))
-        if ret == "*" or res == expected:
-            print " -> SUCCESS"
+        if (t.ret == "" and t.out == ""):
+            log(" -> SUCCESS (result ignored)")
+            pass_cnt += 1
+        elif (re.search(expects[0], res, re.S) or
+                re.search(expects[1], res, re.S)):
+            log(" -> SUCCESS")
+            pass_cnt += 1
         else:
-            print " -> FAIL (line %d):" % line_num
-            print "    Expected : %s" % repr(expected)
-            print "    Got      : %s" % repr(res)
-            fail_cnt += 1
+            if t.soft and not args.hard:
+                log(" -> SOFT FAIL (line %d):" % t.line_num)
+                soft_fail_cnt += 1
+                fail_type = "SOFT "
+            else:
+                log(" -> FAIL (line %d):" % t.line_num)
+                fail_cnt += 1
+                fail_type = ""
+            log("    Expected : %s" % repr(expects[0]))
+            log("    Got      : %s" % repr(res))
+            failed_test = """%sFAILED TEST (line %d): %s -> [%s,%s]:
+    Expected : %s
+    Got      : %s""" % (fail_type, t.line_num, t.form, repr(t.out),
+                        t.ret, repr(expects[0]), repr(res))
+            failures.append(failed_test)
     except:
-        print "Got Exception"
-        r.cleanup()
+        _, exc, _ = sys.exc_info()
+        log("\nException: %s" % repr(exc))
+        log("Output before exception:\n%s" % r.buf)
         sys.exit(1)
-r.cleanup()
+
+if len(failures) > 0:
+    log("\nFAILURES:")
+    for f in failures:
+        log(f)
+
+results = """
+TEST RESULTS (for %s):
+  %3d: soft failing tests
+  %3d: failing tests
+  %3d: passing tests
+  %3d: total tests
+""" % (args.test_file, soft_fail_cnt, fail_cnt,
+        pass_cnt, test_cnt)
+log(results)
+
+debug("\n") # add some separate to debug log
 
 if fail_cnt > 0:
-    print "FAILURES: %d" % fail_cnt
-    sys.exit(2)
+    sys.exit(1)
 sys.exit(0)