Update update_libcxx.py to handle recent changes (#24392)

`update_libcxx.py` updates
  - Now it handles the new `llvm-libc` directory
  - It excludes unused `ryu_(long_double)constants.h` from libc
- It deletes `libcxx/include/__cxx03` for now because we are not using
it.
- Make `clean_dir` and `copy_tree` handle the case when the local dir
does not exist, which can happen when new directories are added
- Make `copy_tree` handle removing files in `excludes` when they are in
subdirectories (In libc, many `CMakeLists.txt` are in subdirectories)

This also adds `readme.txt` to `llvm-libc` directory.

Author: aheejin

system/lib/llvm-libc/README.txt

@@ -0,0 +1,31 @@
+llvm's libc
+-----------
+
+These files are from llvm-project based on release 20.1.4.
+
+We maintain a local fork of llvm-project that contains any emscripten
+specific patches:
+
+  https://github.com/emscripten-core/llvm-project
+
+The current patch is based on the emscripten-libs-20 branch.
+
+We do not use LLVM's libc directly, but libcxx uses a subset of headers from
+libc. So we only import these directories that libcxx depend on:
+- libc/hdr
+- libc/include/llvm-libc-macros
+- libc/include/llvm-libc-types
+- libc/shared
+- libc/src/__support
+
+Update Instructions
+-------------------
+
+Run `system/lib/update_libcxx.py path/to/llvm-project`
+
+Modifications
+-------------
+
+For a list of changes from upstream see the libc files that are part of:
+
+https://github.com/llvm/llvm-project/compare/llvmorg-20.1.4...emscripten-core:emscripten-libs-20

system/lib/llvm-libc/readme.txt

@@ -15,11 +15,21 @@
 local_src = os.path.join(local_root, 'src')
 local_inc = os.path.join(local_root, 'include')
 
-preserve_files = ('readme.txt', 'symbols')
-excludes = ('CMakeLists.txt',)
+preserve_files = ('readme.txt', '__assertion_handler', '__config_site')
+# ryu_constants.h / ryu_long_double_constants.h from libc are not used
+excludes = ('CMakeLists.txt', 'ryu_constants.h', 'ryu_long_double_constants.h')
 
+libc_copy_dirs = [
+    ('hdr',),
+    ('include', 'llvm-libc-macros'),
+    ('include', 'llvm-libc-types'),
+    ('shared',),
+    ('src', '__support'),
+]
 
 def clean_dir(dirname):
+  if not os.path.exists(dirname):
+    return
   for f in os.listdir(dirname):
     if f in preserve_files:
       continue
@@ -31,12 +41,19 @@ def clean_dir(dirname):
 
 
 def copy_tree(upstream_dir, local_dir):
+  if not os.path.exists(local_dir):
+    os.makedirs(local_dir)
   for f in os.listdir(upstream_dir):
     full = os.path.join(upstream_dir, f)
     if os.path.isdir(full):
       shutil.copytree(full, os.path.join(local_dir, f))
     elif f not in excludes:
       shutil.copy2(full, os.path.join(local_dir, f))
+  for root, dirs, files in os.walk(local_dir):
+    for f in files:
+      if f in excludes:
+        full = os.path.join(root, f)
+        os.remove(full)
 
 
 def main():
@@ -60,6 +77,24 @@ def main():
   shutil.copy2(os.path.join(libcxx_dir, 'CREDITS.TXT'), local_root)
   shutil.copy2(os.path.join(libcxx_dir, 'LICENSE.TXT'), local_root)
 
+  # We don't use frozen c++03 headers for now
+  shutil.rmtree(os.path.join(local_inc, '__cxx03'))
+
+  # libcxx includes headers from LLVM's libc
+  libc_upstream_dir = os.path.join(llvm_dir, 'libc')
+  assert os.path.exists(libc_upstream_dir)
+  libc_local_dir = os.path.join(script_dir, 'llvm-libc')
+
+  for dirname in libc_copy_dirs:
+    local_dir = os.path.join(libc_local_dir, *dirname)
+    clean_dir(local_dir)
+
+  for dirname in libc_copy_dirs:
+    upstream_dir = os.path.join(libc_upstream_dir, *dirname)
+    local_dir = os.path.join(libc_local_dir, *dirname)
+    copy_tree(upstream_dir, local_dir)
+
+  shutil.copy2(os.path.join(libc_upstream_dir, 'LICENSE.TXT'), libc_local_dir)
 
 if __name__ == '__main__':
   main()