Validate EM_JS/EM_ASM strings in side modules at build time (emscripten-core#20258)

Author: sbc100

ChangeLog.md

@@ -20,6 +20,8 @@ See docs/process.md for more on how version tagging works.
 
 3.1.47 (in development)
 -----------------------
+- `EM_JS` and `EM_ASM` that are present in side module now have their syntax
+  validated at build time. (#20258)
 
 3.1.46 - 09/15/23
 -----------------

emscripten.py

@@ -21,6 +21,7 @@
 import sys
 
 from tools import building
+from tools import config
 from tools import diagnostics
 from tools import js_manipulation
 from tools import shared
@@ -328,7 +329,27 @@ def emscript(in_wasm, out_wasm, outfile_js, memfile, js_syms):
         if len(signature.params) != len(c_sig):
           diagnostics.warning('em-js-i64', 'using 64-bit arguments in EM_JS function without WASM_BIGINT is not yet fully supported: `%s` (%s, %s)', em_js_func, c_sig, signature.params)
 
+  asm_consts = create_asm_consts(metadata)
+  em_js_funcs = create_em_js(metadata)
+
   if settings.SIDE_MODULE:
+    # When building side modules, valid the EM_ASM and EM_JS string by running
+    # them through node.  Without this step, syntax errors are not surfaced
+    # until runtime.
+    # We use subprocess directly here rather than shared.check_call since
+    # check_call doesn't support the `intput` argument.
+    if asm_consts:
+      validate = '\n'.join([f'var tmp = {f};' for _, f in asm_consts])
+      proc = subprocess.run(config.NODE_JS + ['--check', '-'], input=validate.encode('utf-8'))
+      if proc.returncode:
+        exit_with_error(f'EM_ASM function validation failed (node returned {proc.returncode})')
+
+    if em_js_funcs:
+      validate = '\n'.join(em_js_funcs)
+      proc = subprocess.run(config.NODE_JS + ['--check', '-'], input=validate.encode('utf-8'))
+      if proc.returncode:
+        exit_with_error(f'EM_JS function validation failed (node returned {proc.returncode})')
+
     logger.debug('emscript: skipping remaining js glue generation')
     return
 
@@ -388,8 +409,6 @@ def emscript(in_wasm, out_wasm, outfile_js, memfile, js_syms):
     # In regular runtime, atinits etc. exist in the preamble part
     pre = apply_static_code_hooks(forwarded_json, pre)
 
-  asm_consts = create_asm_consts(metadata)
-  em_js_funcs = create_em_js(metadata)
   asm_const_pairs = ['%s: %s' % (key, value) for key, value in asm_consts]
   extra_code = ''
   if asm_const_pairs or settings.MAIN_MODULE:

test/other/test_em_asm_invalid.c

@@ -0,0 +1,7 @@
+#include <emscripten/em_asm.h>
+
+int main() {
+  EM_ASM({
+    * this is not valid js *
+  });
+}

test/other/test_em_js_invalid.c

@@ -0,0 +1,9 @@
+#include <emscripten/em_js.h>
+
+EM_JS(int, foo, (), {
+  * this is not valid js *
+});
+
+int main() {
+  foo();
+}

test/test_other.py

@@ -11063,6 +11063,12 @@ def test_em_asm_strict_c(self):
     err = self.expect_fail([EMCC, '-std=c11', 'src.c'])
     self.assertIn('EM_ASM does not work in -std=c* modes, use -std=gnu* modes instead', err)
 
+  def test_em_asm_invalid(self):
+    # Test that invalid EM_ASM in side modules since is detected at build time.
+    err = self.expect_fail([EMCC, '-sSIDE_MODULE', test_file('other/test_em_asm_invalid.c')])
+    self.assertContained("SyntaxError: Unexpected token '*'", err)
+    self.assertContained('emcc: error: EM_ASM function validation failed', err)
+
   def test_boost_graph(self):
     self.do_runf(test_file('test_boost_graph.cpp'), emcc_args=['-std=c++14', '-sUSE_BOOST_HEADERS'])
 
@@ -12051,6 +12057,13 @@ def test_runtime_keepalive(self):
     self.set_setting('EXIT_RUNTIME')
     self.do_other_test('test_runtime_keepalive.cpp')
 
+  @crossplatform
+  def test_em_js_invalid(self):
+    # Test that invalid EM_JS in side modules since is detected at build time.
+    err = self.expect_fail([EMCC, '-sSIDE_MODULE', test_file('other/test_em_js_invalid.c')])
+    self.assertContained("SyntaxError: Unexpected token '*'", err)
+    self.assertContained('emcc: error: EM_JS function validation failed', err)
+
   @crossplatform
   def test_em_js_side_module(self):
     self.build(test_file('other/test_em_js_side.c'), js_outfile=False, emcc_args=['-sSIDE_MODULE'], output_basename='side')