Source maps (Refs #1252, Reopens #1267)

emcc

@@ -49,7 +49,7 @@ emcc can be influenced by a few environment variables:
 
 import os, sys, shutil, tempfile, subprocess, shlex, time, re, logging
 from subprocess import PIPE, STDOUT
-from tools import shared
+from tools import shared, jsrun
 from tools.shared import Compression, execute, suffix, unsuffixed, unsuffixed_basename
 from tools.response_file import read_response_file
 
@@ -203,10 +203,18 @@ Options that are modified or new in %s include:
                             -g2  Preserve function names
                             -g3  Preserve variable names
                             -g4  Preserve LLVM debug info (if -g was
-                                 used when compiling the C/C++ sources)
-                                 and show line number debug comments.
-                                 This is the highest level of debuggability.
-                                 (default in -O0)
+                                 used when compiling the C/C++ sources),
+                                 show line number debug comments, and
+                                 generate source maps. This is the highest
+                                 level of debuggability. Note that this
+                                 may make -O1 and above significantly
+                                 slower because JS optimization will be
+                                 limited to 1 core.  (default in -O0)
+
+  -g2                      Like -g1, but we generate source maps as well,
+                           and we preserve comments even with -O1 and above.
+                           Note that this may be considerably slower because
+                           JS optimization is limited to a single core.
 
   --typed-arrays <mode>    0: No typed arrays
                            1: Parallel typed arrays
@@ -731,6 +739,14 @@ try:
 
   settings_changes = []
 
+  def validate_arg_level(level_string, max_level, err_msg):
+    try:
+      level = int(level_string)
+      assert 0 <= level <= max_level
+    except:
+      raise Exception(err_msg)
+    return level
+
   for i in range(len(newargs)):
     newargs[i] = newargs[i].strip() # On Windows Vista (and possibly others), excessive spaces in the command line leak into the items in this array, so trim e.g. 'foo.cpp ' -> 'foo.cpp'
     if newargs[i].startswith('-O'):
@@ -739,11 +755,7 @@ try:
       if requested_level == 's':
         requested_level = 2
         settings_changes.append('INLINING_LIMIT=50')
-      try:
-        opt_level = int(requested_level)
-        assert 0 <= opt_level <= 3
-      except:
-        raise Exception('Invalid optimization level: ' + newargs[i])
+      opt_level = validate_arg_level(requested_level, 3, 'Invalid optimization level: ' + newargs[i])
       newargs[i] = ''
     elif newargs[i].startswith('--llvm-opts'):
       check_bad_eq(newargs[i])
@@ -787,12 +799,8 @@ try:
       newargs[i+1] = ''
     elif newargs[i].startswith('-g'):
       requested_level = newargs[i][2:] or '3'
-      try:
-        debug_level = int(requested_level)
-        assert 0 <= debug_level <= 4
-      except:
-        raise Exception('Invalid debug level: ' + newargs[i])
-      newargs[i] = '-g' # discard level for clang args
+      debug_level = validate_arg_level(requested_level, 4, 'Invalid debug level: ' + newargs[i])
+      newargs[i] = '-g' # we'll need this to get LLVM debug info
     elif newargs[i] == '--bind':
       bind = True
       newargs[i] = ''
@@ -884,6 +892,11 @@ try:
   if opt_level == 0: debug_level = 4
   if closure is None and opt_level == 3: closure = True
 
+  # TODO: support source maps with js_transform
+  if js_transform and debug_level >= 4:
+    logging.warning('disabling source maps because a js transform is being done')
+    debug_level = 3
+
   if DEBUG: start_time = time.time() # done after parsing arguments, which might affect debug state
 
   if closure:
@@ -1408,6 +1421,7 @@ try:
   # Optimize, if asked to
   if not LEAVE_INPUTS_RAW:
     link_opts = [] if debug_level >= 4 else ['-strip-debug'] # remove LLVM debug if we are not asked for it
+
     if llvm_opts > 0:
       if not os.environ.get('EMCC_OPTIMIZE_NORMALLY'):
         shared.Building.llvm_opt(in_temp(target_basename + '.bc'), llvm_opts)
@@ -1495,9 +1509,11 @@ try:
     final += '.tr.js'
     posix = True if not shared.WINDOWS else False
     logging.debug('applying transform: %s' % js_transform)
-    execute(shlex.split(js_transform, posix=posix) + [os.path.abspath(final)])
+    subprocess.check_call(shlex.split(js_transform, posix=posix) + [os.path.abspath(final)])
     if DEBUG: save_intermediate('transformed')
 
+  js_transform_tempfiles = [final]
+
   # It is useful to run several js optimizer passes together, to save on unneeded unparsing/reparsing
   js_optimizer_queue = []
   def flush_js_optimizer_queue():
@@ -1507,15 +1523,17 @@ try:
         if shared.Settings.ASM_JS:
           js_optimizer_queue = ['asm'] + js_optimizer_queue
         logging.debug('applying js optimization passes: %s', js_optimizer_queue)
-        final = shared.Building.js_optimizer(final, js_optimizer_queue, jcache)
+        final = shared.Building.js_optimizer(final, js_optimizer_queue, jcache, debug_level >= 4)
+        js_transform_tempfiles.append(final)
         if DEBUG: save_intermediate('js_opts')
       else:
         for name in js_optimizer_queue:
           passes = [name]
           if shared.Settings.ASM_JS:
             passes = ['asm'] + passes
           logging.debug('applying js optimization pass: %s', passes)
-          final = shared.Building.js_optimizer(final, passes, jcache)
+          final = shared.Building.js_optimizer(final, passes, jcache, debug_level >= 4)
+          js_transform_tempfiles.append(final)
           save_intermediate(name)
       js_optimizer_queue = []
 
@@ -1524,7 +1542,8 @@ try:
 
     if DEBUG == '2':
       # Clean up the syntax a bit
-      final = shared.Building.js_optimizer(final, [], jcache)
+      final = shared.Building.js_optimizer(final, [], jcache, debug_level >= 4)
+      js_transform_tempfiles.append(final)
       if DEBUG: save_intermediate('pretty')
 
     def get_eliminate():
@@ -1542,6 +1561,8 @@ try:
     flush_js_optimizer_queue()
 
     logging.debug('running closure')
+    # no need to add this to js_transform_tempfiles, because closure and
+    # debug_level > 0 are never simultaneously true
     final = shared.Building.closure_compiler(final)
     if DEBUG: save_intermediate('closure')
 
@@ -1589,19 +1610,33 @@ try:
       src = re.sub('/\* memory initializer \*/ allocate\(([\d,\.concat\(\)\[\]\\n ]+)"i8", ALLOC_NONE, Runtime\.GLOBAL_BASE\)', repl, src, count=1)
       open(final + '.mem.js', 'w').write(src)
       final += '.mem.js'
+      js_transform_tempfiles[-1] = final # simple text substitution preserves comment line number mappings
       if DEBUG:
         if os.path.exists(memfile):
           save_intermediate('meminit')
           logging.debug('wrote memory initialization to %s' % memfile)
         else:
           logging.debug('did not see memory initialization')
 
+  def generate_source_map(map_file_base_name, offset=0):
+    jsrun.run_js(shared.path_from_root('tools', 'source-maps', 'sourcemapper.js'),
+      shared.NODE_JS, js_transform_tempfiles +
+        ['--sourceRoot', os.getcwd(),
+         '--mapFileBaseName', map_file_base_name,
+         '--offset', str(offset)])
+
   # If we were asked to also generate HTML, do that
   if final_suffix == 'html':
     logging.debug('generating HTML')
     shell = open(shell_path).read()
     html = open(target, 'w')
     if not Compression.on:
+      if debug_level >= 4:
+        match = re.match('.*?<script[^>]*>{{{ SCRIPT_CODE }}}</script>', shell,
+            re.DOTALL)
+        if match is None:
+          raise RuntimeError('Could not find script insertion point')
+        generate_source_map(target, match.group().count('\n'))
       html.write(shell.replace('{{{ SCRIPT_CODE }}}', open(final).read()))
     else:
       # Compress the main code
@@ -1672,6 +1707,8 @@ try:
         from tools.split import split_javascript_file
         split_javascript_file(final, unsuffixed(target), split_js_file)
     else:
+        if debug_level >= 4: generate_source_map(target)
+
         # copy final JS to output
         shutil.move(final, target)
 

src/jsifier.js

@@ -695,7 +695,9 @@ function JSify(data, functionsOnly, givenFunctions) {
               }
             }
             i++;
-            return JS + (Debugging.on ? Debugging.getComment(line.lineNum) : '');
+            // invoke instructions span two lines, and the debug info is located
+            // on the second line, hence the +1
+            return JS + (Debugging.on ? Debugging.getComment(line.lineNum + (line.intertype === 'invoke' ? 1 : 0)) : '');
           })
                                   .join('\n')
                                   .split('\n') // some lines include line breaks

src/shell.html

@@ -87,10 +87,6 @@
       };
       Module.setStatus('Downloading...');
     </script>      
-    <script type='text/javascript'>
-
-      {{{ SCRIPT_CODE }}}
-
-    </script>
+    <script type='text/javascript'>{{{ SCRIPT_CODE }}}</script>
   </body>
 </html>