diff options
Diffstat (limited to 'sources/shiboken6/libshiboken/embed')
4 files changed, 546 insertions, 0 deletions
diff --git a/sources/shiboken6/libshiboken/embed/embedding_generator.py b/sources/shiboken6/libshiboken/embed/embedding_generator.py new file mode 100644 index 000000000..51c46ce54 --- /dev/null +++ b/sources/shiboken6/libshiboken/embed/embedding_generator.py @@ -0,0 +1,221 @@ +# Copyright (C) 2022 The Qt Company Ltd. +# SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only +from __future__ import annotations + +""" +embedding_generator.py + +This file takes the content of the two supported directories and inserts +it into a zip file. The zip file is then converted into a C++ source +file that can easily be unpacked again with Python (see signature.cpp, +constant 'PySide_PythonCode'). + +Note that this _is_ a zipfile, but since it is embedded into the shiboken +binary, we cannot use the zipimport module from Python. +But a similar solution is possible that allows for normal imports. + +See signature_bootstrap.py for details. +""" + +import sys +import os +import subprocess +import textwrap +import tempfile +import argparse +import marshal +import traceback +from pathlib import Path + +# work_dir is set to the source for testing, only. +# It can be overridden in the command line. +work_dir = Path(__file__).parent.resolve() +embed_dir = work_dir +cur_dir = Path.cwd() +source_dir = work_dir.parents[2] +assert source_dir.name == "sources" +build_script_dir = work_dir.parents[3] +assert (build_script_dir / "build_scripts").exists() + +sys.path.insert(0, os.fspath(build_script_dir)) + +from build_scripts import utils + + +def runpy(cmd, **kw): + subprocess.call([sys.executable, '-E'] + cmd.split(), **kw) + + +def create_zipfile(use_pyc, quiet): + """ + Collect all Python files, compile them, create a zip file + and make a chunked base64 encoded file from it. + """ + zip_name = "signature.zip" + inc_name = "signature_inc.h" + flag = '-b' + os.chdir(work_dir) + + # Remove all left-over py[co] and other files first, in case we use '--reuse-build'. + # Note that we could improve that with the PyZipfile function to use .pyc files + # in different folders, but that makes only sense when COIN allows us to have + # multiple Python versions in parallel. + for root, dirs, files in os.walk(work_dir): + for name in files: + fpath = Path(root) / name + ew = name.endswith + if ew(".pyc") or ew(".pyo") or ew(".zip") or ew(".inc"): + os.remove(fpath) + # We copy every Python file into this dir, but only for the right version. + # For testing in the source dir, we need to filter. + ignore = [] + utils.copydir(source_dir / "shiboken6" / "shibokenmodule" / "files.dir" / "shibokensupport", + work_dir / "shibokensupport", + ignore=ignore, file_filter_function=lambda name, n2: name.endswith(".py")) + if embed_dir != work_dir: + utils.copyfile(embed_dir / "signature_bootstrap.py", work_dir) + + if not use_pyc: + pass # We cannot compile, unless we have folders per Python version + else: + files = ' '.join(fn for fn in os.listdir('.')) + runpy(f'-m compileall -q {flag} {files}') + files = ' '.join(fn for fn in os.listdir('.') if not fn == zip_name) + runpy(f'-m zipfile -c {zip_name} {files}') + tmp = tempfile.TemporaryFile(mode="w+") + runpy(f'-m base64 {zip_name}', stdout=tmp) + # now generate the include file + tmp.seek(0) + with open(inc_name, "w") as inc: + _embed_file(tmp, inc) + tmp.close() + + # also generate a simple embeddable .pyc file for signature_bootstrap.pyc + boot_name = "signature_bootstrap.py" if not use_pyc else "signature_bootstrap.pyc" + with open(boot_name, "rb") as ldr, open("signature_bootstrap_inc.h", "w") as inc: + _embed_bytefile(ldr, inc, not use_pyc) + os.chdir(cur_dir) + if quiet: + return + + # have a look at our populated folder unless quiet option + def tree(directory): + print(f'+ {directory}') + for path in sorted(directory.rglob('*')): + depth = len(path.relative_to(directory).parts) + spacer = ' ' * depth + print(f'{spacer}+ {path.name}') + + print("++++ Current contents of") + tree(work_dir) + print("++++") + + +def _embed_file(fin, fout): + """ + Format a text file for embedding in a C++ source file. + """ + # MSVC has a 64k string limitation. In C, it would be easy to create an + # array of 64 byte strings and use them as one big array. In C++ this does + # not work, since C++ insists in having the terminating nullbyte. + # Therefore, we split the string after an arbitrary number of lines + # (chunked file). + limit = 50 + text = fin.readlines() + print(textwrap.dedent(""" + // This is a ZIP archive of all Python files in the directory + // "shiboken6/shibokenmodule/files.dir/shibokensupport/signature" + // There is also a toplevel file "signature_bootstrap.py[c]" that will be + // directly executed from C++ as a bootstrap loader. + """).strip(), file=fout) + block, blocks = 0, len(text) // limit + 1 + for idx, line in enumerate(text): + if idx % limit == 0: + if block: + fout.write(')"\n') + comma = "," if block else "" + block += 1 + fout.write(f'\n{comma} // Block {block} of {blocks}\nR"(') + else: + fout.write('\n') + fout.write(line.strip()) + fout.write(')"\n\n/* Sentinel */, ""\n') + + +def _embed_bytefile(fin, fout, is_text): + """ + Format a binary file for embedding in a C++ source file. + This version works directly with a single .pyc file. + """ + fname = fin.name + remark = ("No .pyc file because '--LIMITED-API=yes'" if is_text else + "The .pyc header is stripped away") + print(textwrap.dedent(f""" + /* + * This is the file "{fname}" as a simple byte array. + * It can be directly embedded without any further processing. + * {remark}. + */ + """), file=fout) + headsize = ( 0 if is_text else + 16 if sys.version_info >= (3, 7) else 12 if sys.version_info >= (3, 3) else 8) + binstr = fin.read()[headsize:] + if is_text: + try: + compile(binstr, fin.name, "exec") + except SyntaxError as e: + print(e) + traceback.print_exc(file=sys.stdout) + print(textwrap.dedent(f""" + ************************************************************************* + *** + *** Could not compile the boot loader '{fname}'! + *** + ************************************************************************* + """)) + raise SystemError + else: + try: + marshal.loads(binstr) + except ValueError as e: + print(e) + traceback.print_exc(file=sys.stdout) + version = sys.version_info[:3] + print(textwrap.dedent(f""" + ************************************************************************* + *** + *** This Python version {version} seems to have a new .pyc header size. + *** Please correct the 'headsize' constant ({headsize}). + *** + ************************************************************************* + """)) + raise SystemError + + print(file=fout) + use_ord = sys.version_info[0] == 2 + for i in range(0, len(binstr), 16): + for c in bytes(binstr[i : i + 16]): + ord_c = ord(c) if use_ord else c + print(f"{ord_c:#4},", file=fout, end="") + print(file=fout) + print("/* End Of File */", file=fout) + + +def str2bool(v): + if v.lower() in ('yes', 'true', 't', 'y', '1'): + return True + elif v.lower() in ('no', 'false', 'f', 'n', '0'): + return False + else: + raise argparse.ArgumentTypeError('Boolean value expected.') + + +if __name__ == "__main__": + parser = argparse.ArgumentParser() + parser.add_argument('--cmake-dir', nargs="?") + parser.add_argument('--use-pyc', type=str2bool) + parser.add_argument('--quiet', action='store_true') + args = parser.parse_args() + if args.cmake_dir: + work_dir = Path(args.cmake_dir).resolve() + create_zipfile(args.use_pyc, args.quiet) diff --git a/sources/shiboken6/libshiboken/embed/module_collector.py b/sources/shiboken6/libshiboken/embed/module_collector.py new file mode 100644 index 000000000..8f7be6437 --- /dev/null +++ b/sources/shiboken6/libshiboken/embed/module_collector.py @@ -0,0 +1,72 @@ +# Copyright (C) 2022 The Qt Company Ltd. +# SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only +from __future__ import annotations + +""" +module_collector.py + +Collect a number of modules listed on the command line. + +The purpose of this script is to generate the scripts needed for +a complete isolation of the signature extension. + +Usage: + +Run this script in one of the used python versions. +It will create an executable archive of the files on the command line. +""" + +import sys +import os +import argparse +import pickle +from textwrap import dedent +from pathlib import path + + +def source_archive(module, modname): + fname = Path(module.__file__).stem + ".py" + with open(fname) as source: + text = source.read() + encoded = text.replace("'''", "(triple_single)") + # modname = module.__name__ + # Do not use: Some modules rename themselves! + version = ".".join(map(str, sys.version_info[:3])) + shortname = fname.stem + preamble = dedent(fr""" + # BEGIN SOURCE ARCHIVE Python {version} module {modname} + + sources = {{}} if "sources" not in globals() else sources + sources["{modname}"] = '''\ + {encoded}'''.replace("(triple_single)", "'''") + + # END SOURCE ARCHIVE Python {version} module {modname} + """) + return preamble + + +def read_all(modules): + collected = "" + for modname in modules: + mod = __import__(modname) + collected += source_archive(mod, modname) + return collected + + +def license_header(): + license = Path(__file__).parent / "qt_python_license.txt" + with license.open() as f: + return f.read() + + +if __name__ == "__main__": + parser = argparse.ArgumentParser() + parser.add_argument('modules', nargs="+") + args = parser.parse_args() + print("modules:", args.modules) + ret = license_header() + read_all(args.modules) + ma_mi = "_".join(map(str, sys.version_info[:2])) + outpath = Path(__file__).parents[2] / Path("shibokenmodule", + "files.dir", "shibokensupport", f"python_minilib_{ma_mi}.py") + with outpath.open("w") as f: + f.write(ret) diff --git a/sources/shiboken6/libshiboken/embed/qt_python_license.txt b/sources/shiboken6/libshiboken/embed/qt_python_license.txt new file mode 100644 index 000000000..e5fdfdf4d --- /dev/null +++ b/sources/shiboken6/libshiboken/embed/qt_python_license.txt @@ -0,0 +1,48 @@ +# Copyright (C) 2022 The Qt Company Ltd. +# SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +## +## PSF LICENSE AGREEMENT FOR PYTHON 3.7.0 +## +## 1. This LICENSE AGREEMENT is between the Python Software Foundation ("PSF"), and +## the Individual or Organization ("Licensee") accessing and otherwise using Python +## 3.7.0 software in source or binary form and its associated documentation. +## +## 2. Subject to the terms and conditions of this License Agreement, PSF hereby +## grants Licensee a nonexclusive, royalty-free, world-wide license to reproduce, +## analyze, test, perform and/or display publicly, prepare derivative works, +## distribute, and otherwise use Python 3.7.0 alone or in any derivative +## version, provided, however, that PSF's License Agreement and PSF's notice of +## copyright, i.e., "Copyright © 2001-2018 Python Software Foundation; All Rights +## Reserved" are retained in Python 3.7.0 alone or in any derivative version +## prepared by Licensee. +## +## 3. In the event Licensee prepares a derivative work that is based on or +## incorporates Python 3.7.0 or any part thereof, and wants to make the +## derivative work available to others as provided herein, then Licensee hereby +## agrees to include in any such work a brief summary of the changes made to Python +## 3.7.0. +## +## 4. PSF is making Python 3.7.0 available to Licensee on an "AS IS" basis. +## PSF MAKES NO REPRESENTATIONS OR WARRANTIES, EXPRESS OR IMPLIED. BY WAY OF +## EXAMPLE, BUT NOT LIMITATION, PSF MAKES NO AND DISCLAIMS ANY REPRESENTATION OR +## WARRANTY OF MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE OR THAT THE +## USE OF PYTHON 3.7.0 WILL NOT INFRINGE ANY THIRD PARTY RIGHTS. +## +## 5. PSF SHALL NOT BE LIABLE TO LICENSEE OR ANY OTHER USERS OF PYTHON 3.7.0 +## FOR ANY INCIDENTAL, SPECIAL, OR CONSEQUENTIAL DAMAGES OR LOSS AS A RESULT OF +## MODIFYING, DISTRIBUTING, OR OTHERWISE USING PYTHON 3.7.0, OR ANY DERIVATIVE +## THEREOF, EVEN IF ADVISED OF THE POSSIBILITY THEREOF. +## +## 6. This License Agreement will automatically terminate upon a material breach of +## its terms and conditions. +## +## 7. Nothing in this License Agreement shall be deemed to create any relationship +## of agency, partnership, or joint venture between PSF and Licensee. This License +## Agreement does not grant permission to use PSF trademarks or trade name in a +## trademark sense to endorse or promote products or services of Licensee, or any +## third party. +## +## 8. By copying, installing or otherwise using Python 3.7.0, Licensee agrees +## to be bound by the terms and conditions of this License Agreement. +## diff --git a/sources/shiboken6/libshiboken/embed/signature_bootstrap.py b/sources/shiboken6/libshiboken/embed/signature_bootstrap.py new file mode 100644 index 000000000..b0ba77107 --- /dev/null +++ b/sources/shiboken6/libshiboken/embed/signature_bootstrap.py @@ -0,0 +1,205 @@ +# Copyright (C) 2022 The Qt Company Ltd. +# SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only +from __future__ import annotations + +""" +signature_bootstrap.py +---------------------- + +This file was originally directly embedded into the C source. +After it grew more and more, I now prefer to have it as Python file. + +Meanwhile, there is also no more a stub loader necessary: +Because we meanwhile have embedding support, we could also load this file +directly from a .pyc file. + +This file replaces the hard to read Python stub in 'signature.cpp', and we +could distinguish better between bootstrap related functions and loader +functions. +It is embedded into 'signature.cpp' as "embed/signature_bootstrap.inc". + +# PYSIDE-1436: Python 3.10 had a problem with EmbeddableZipImporter because the +imports were in the functions. Moved them outside into the globals. +""" + +recursion_trap = 0 + +import base64 +import importlib +import io +import os +import sys +import traceback +import zipfile + +from contextlib import contextmanager +from importlib.machinery import ModuleSpec +from pathlib import Path + + +def bootstrap(): + + global recursion_trap + if recursion_trap: + # we are probably called from outside, already + print("Recursion occurred in Bootstrap. Did you start by hand? Then it's ok.") + print("But you should trigger start by '_init_pyside_extension()', only!") + recursion_trap += 1 + + @contextmanager + def ensure_shibokensupport(target, support_path): + # Make sure that we always have the shibokensupport containing package first. + # Also remove any prior loaded module of this name, just in case. + # PYSIDE-1621: support_path can also be a finder instance. + target.insert(0, support_path) + + sbks = "shibokensupport" + if sbks in sys.modules: + del sys.modules[sbks] + prefix = sbks + "." + for key in list(key for key in sys.modules if key.startswith(prefix)): + del sys.modules[key] + try: + import shibokensupport + yield + except Exception as e: + f = sys.stderr + print("Problem importing shibokensupport:", file=f) + print(f"{e.__class__.__name__}: {e}", file=f) + traceback.print_exc() + print("sys.path:", file=f) + for p in sys.path: + print(" " + p, file=f) + f.flush() + sys.exit(-1) + target.remove(support_path) + + # Here we decide if we re-incarnate the embedded files or use embedding. + incarnated = find_incarnated_files() + if incarnated: + target, support_path = sys.path, os.fspath(incarnated) + else: + target, support_path = prepare_zipfile() + with ensure_shibokensupport(target, support_path): + from shibokensupport.signature import loader + return loader + +# Newer functionality: +# This function checks if the support directory exist and returns it. +# If does not exist, we try to create it and return it. +# Otherwise, we return None. + +def find_incarnated_files(): + import shiboken6 as root + files_dir = Path(root.__file__).resolve().parent / "files.dir" + handle_embedding_switch(files_dir) + if files_dir.exists(): + sys.path.insert(0, os.fspath(files_dir)) + # Note: To avoid recursion problems, we need to preload the loader. + # But that has the side-effect that we need to delay the feature + # initialization until all function pointers are set. + # See `post_init_func` in signature_globals.cpp . + import shibokensupport.signature.loader + del sys.path[0] + return files_dir + return None + + +def handle_embedding_switch(files_dir): + """ + This handles the optional environment variable `SBK_EMBED` + if not set : do nothing + if set to 0, false, no : de-virtualize the Python files + if set to 1, true, yes : virtualize again (delete "files.dir") + """ + env_name = "SBK_EMBED" + env_var = os.environ.get(env_name) + if not env_var: + return + if env_var.lower() in ("1", "t", "true", "y", "yes"): + import shutil + shutil.rmtree(files_dir, ignore_errors=True) + elif env_var.lower() in ("0", "f", "false", "n", "no"): + reincarnate_files(files_dir) + + +def reincarnate_files(files_dir): + target, zip = prepare_zipfile() + names = (_ for _ in zip.zfile.namelist() if _.endswith(".py")) + try: + # First check mkdir to get an error when we cannot write. + files_dir.mkdir(exist_ok=True) + except os.error as e: + print(f"SBK_EMBED=False: Warning: Cannot write into {files_dir}") + return None + try: + # Then check for a real error when unpacking the zip file. + zip.zfile.extractall(path=files_dir, members=names) + return files_dir + except Exception as e: + print(f"{e.__class__.__name__}: {e}", file=sys.stderr) + traceback.print_exc() + raise + +# New functionality: Loading from a zip archive. +# There exists the zip importer, but as it is written, only real zip files are +# supported. Before I will start an own implementation, it is easiest to use +# a temporary zip file. +# PYSIDE-1621: make zip file access totally virtual + +def prepare_zipfile(): + """ + Old approach: + + Write the zip file to a real file and return its name. + It will be implicitly opened as such when we add the name to sys.path . + + New approach (Python 3, only): + + Use EmbeddableZipImporter and pass the zipfile structure directly. + The sys.path way does not work, instead we need to use sys.meta_path . + See https://docs.python.org/3/library/sys.html#sys.meta_path + """ + + # 'zipstring_sequence' comes from signature.cpp + zipbytes = base64.b64decode(''.join(zipstring_sequence)) + vzip = zipfile.ZipFile(io.BytesIO(zipbytes)) + return sys.meta_path, EmbeddableZipImporter(vzip) + + +class EmbeddableZipImporter(object): + + def __init__(self, zip_file): + def p2m(filename): + if filename.endswith("/__init__.py"): + return filename[:-12].replace("/", ".") + if filename.endswith(".py"): + return filename[:-3].replace("/", ".") + return None + + self.zfile = zip_file + self._mod2path = {p2m(_.filename) : _.filename for _ in zip_file.filelist} + + def find_spec(self, fullname, path, target=None): + path = self._mod2path.get(fullname) + return ModuleSpec(fullname, self) if path else None + + def create_module(self, spec): + return None + + def exec_module(self, module): + fullname = module.__spec__.name + filename = self._mod2path[fullname] + with self.zfile.open(filename, "r") as f: # "rb" not for zipfile + codeob = compile(f.read(), filename, "exec") + exec(codeob, module.__dict__) + module.__file__ = filename + module.__loader__ = self + if filename.endswith("/__init__.py"): + module.__path__ = [] + module.__package__ = fullname + else: + module.__package__ = fullname.rpartition('.')[0] + sys.modules[fullname] = module + +# eof |