def pyi_modgraph():
# Explicitly set the log level since the plugin `pytest-catchlog` (un-)
# sets the root logger's level to NOTSET for the setup phase, which will
# lead to TRACE messages been written out.
import PyInstaller.log as logging
logging.logger.setLevel(logging.DEBUG)
initialize_modgraph()
def pyi_modgraph():
# Explicitly set the log level since the plugin `pytest-catchlog` (un-)
# sets the root logger's level to NOTSET for the setup phase, which will
# lead to TRACE messages been written out.
import PyInstaller.log as logging
logging.logger.setLevel(logging.DEBUG)
return initialize_modgraph()
def fresh_pyi_modgraph(monkeypatch):
"""
Get a fresh PyiModuleGraph
"""
def fake_base_modules(self):
# speed up set up
self._base_modules = ()
logging.logger.setLevel(logging.DEBUG)
# ensure we get a fresh PyiModuleGraph
monkeypatch.setattr(analysis, "_cached_module_graph_", None)
# speed up setup
monkeypatch.setattr(analysis.PyiModuleGraph, "_analyze_base_modules",
fake_base_modules)
return analysis.initialize_modgraph()
def assemble(self):
"""
This method is the MAIN method for finding all necessary files to be bundled.
"""
from PyInstaller.config import CONF
for m in self.excludes:
logger.debug("Excluding module '%s'" % m)
self.graph = initialize_modgraph(excludes=self.excludes,
user_hook_dirs=self.hookspath)
# TODO Find a better place where to put 'base_library.zip' and when to created it.
# For Python 3 it is necessary to create file 'base_library.zip'
# containing core Python modules. In Python 3 some built-in modules
# are written in pure Python. base_library.zip is a way how to have
# those modules as "built-in".
libzip_filename = os.path.join(CONF['workpath'], 'base_library.zip')
create_py3_base_library(libzip_filename, graph=self.graph)
# Bundle base_library.zip as data file.
# Data format of TOC item: ('relative_path_in_dist_dir', 'absolute_path_on_disk', 'DATA')
self.datas.append(
(os.path.basename(libzip_filename), libzip_filename, 'DATA'))
# Expand sys.path of module graph.
# The attribute is the set of paths to use for imports: sys.path,
# plus our loader, plus other paths from e.g. --path option).
self.graph.path = self.pathex + self.graph.path
self.graph.set_setuptools_nspackages()
logger.info("running Analysis %s", self.tocbasename)
# Get paths to Python and, in Windows, the manifest.
python = compat.python_executable
if not is_win:
# Linux/MacOS: get a real, non-link path to the running Python executable.
while os.path.islink(python):
python = os.path.join(os.path.dirname(python),
os.readlink(python))
depmanifest = None
else:
# Windows: Create a manifest to embed into built .exe, containing the same
# dependencies as python.exe.
depmanifest = winmanifest.Manifest(
type_="win32",
name=CONF['specnm'],
processorArchitecture=winmanifest.processor_architecture(),
version=(1, 0, 0, 0))
depmanifest.filename = os.path.join(
CONF['workpath'], CONF['specnm'] + ".exe.manifest")
# We record "binaries" separately from the modulegraph, as there
# is no way to record those dependencies in the graph. These include
# the python executable and any binaries added by hooks later.
# "binaries" are not the same as "extensions" which are .so or .dylib
# that are found and recorded as extension nodes in the graph.
# Reset seen variable before running bindepend. We use bindepend only for
# the python executable.
bindepend.seen.clear()
# Add binary and assembly dependencies of Python.exe.
# This also ensures that its assembly depencies under Windows get added to the
# built .exe's manifest. Python 2.7 extension modules have no assembly
# dependencies, and rely on the app-global dependencies set by the .exe.
self.binaries.extend(
bindepend.Dependencies([('', python, '')],
manifest=depmanifest,
redirects=self.binding_redirects)[1:])
if is_win:
depmanifest.writeprettyxml()
### Module graph.
#
# Construct the module graph of import relationships between modules
# required by this user's application. For each entry point (top-level
# user-defined Python script), all imports originating from this entry
# point are recursively parsed into a subgraph of the module graph. This
# subgraph is then connected to this graph's root node, ensuring
# imported module nodes will be reachable from the root node -- which is
# is (arbitrarily) chosen to be the first entry point's node.
# List to hold graph nodes of scripts and runtime hooks in use order.
priority_scripts = []
# Assume that if the script does not exist, Modulegraph will raise error.
# Save the graph nodes of each in sequence.
for script in self.inputs:
logger.info("Analyzing %s", script)
priority_scripts.append(self.graph.add_script(script))
# Analyze the script's hidden imports (named on the command line)
self.graph.add_hiddenimports(self.hiddenimports)
### Post-graph hooks.
self.graph.process_post_graph_hooks(self)
# Update 'binaries' TOC and 'datas' TOC.
deps_proc = DependencyProcessor(self.graph,
self.graph._additional_files_cache)
self.binaries.extend(deps_proc.make_binaries_toc())
self.datas.extend(deps_proc.make_datas_toc())
self.zipped_data.extend(deps_proc.make_zipped_data_toc())
# Note: zipped eggs are collected below
### Look for dlls that are imported by Python 'ctypes' module.
# First get code objects of all modules that import 'ctypes'.
logger.info('Looking for ctypes DLLs')
# dict like: {'module1': code_obj, 'module2': code_obj}
ctypes_code_objs = self.graph.get_code_using("ctypes")
for name, co in ctypes_code_objs.items():
# Get dlls that might be needed by ctypes.
logger.debug('Scanning %s for shared libraries or dlls', name)
try:
ctypes_binaries = scan_code_for_ctypes(co)
self.binaries.extend(set(ctypes_binaries))
except Exception as ex:
raise RuntimeError(f"Failed to scan the module '{name}'. "
f"This is a bug. Please report it.") from ex
self.datas.extend((dest, source, "DATA")
for (dest, source) in format_binaries_and_datas(
self.graph.metadata_required()))
# Analyze run-time hooks.
# Run-time hooks has to be executed before user scripts. Add them
# to the beginning of 'priority_scripts'.
priority_scripts = self.graph.analyze_runtime_hooks(
self.custom_runtime_hooks) + priority_scripts
# 'priority_scripts' is now a list of the graph nodes of custom runtime
# hooks, then regular runtime hooks, then the PyI loader scripts.
# Further on, we will make sure they end up at the front of self.scripts
### Extract the nodes of the graph as TOCs for further processing.
# Initialize the scripts list with priority scripts in the proper order.
self.scripts = self.graph.nodes_to_toc(priority_scripts)
# Extend the binaries list with all the Extensions modulegraph has found.
self.binaries = self.graph.make_binaries_toc(self.binaries)
# Fill the "pure" list with pure Python modules.
assert len(self.pure) == 0
self.pure = self.graph.make_pure_toc()
# And get references to module code objects constructed by ModuleGraph
# to avoid writing .pyc/pyo files to hdd.
self.pure._code_cache = self.graph.get_code_objects()
# Add remaining binary dependencies - analyze Python C-extensions and what
# DLLs they depend on.
logger.info('Looking for dynamic libraries')
self.binaries.extend(
bindepend.Dependencies(self.binaries,
redirects=self.binding_redirects))
### Include zipped Python eggs.
logger.info('Looking for eggs')
self.zipfiles.extend(deps_proc.make_zipfiles_toc())
# Verify that Python dynamic library can be found.
# Without dynamic Python library PyInstaller cannot continue.
self._check_python_library(self.binaries)
if is_win:
# Remove duplicate redirects
self.binding_redirects[:] = list(set(self.binding_redirects))
logger.info("Found binding redirects: \n%s",
self.binding_redirects)
# Filter binaries to adjust path of extensions that come from
# python's lib-dynload directory. Prefix them with lib-dynload
# so that we'll collect them into subdirectory instead of
# directly into _MEIPASS
for idx, tpl in enumerate(self.binaries):
name, path, typecode = tpl
if typecode == 'EXTENSION' \
and not os.path.dirname(os.path.normpath(name)) \
and os.path.basename(os.path.dirname(path)) == 'lib-dynload':
name = os.path.join('lib-dynload', name)
self.binaries[idx] = (name, path, typecode)
# Place Python source in data files for the noarchive case.
if self.noarchive:
# Create a new TOC of ``(dest path for .pyc, source for .py, type)``.
new_toc = TOC()
for name, path, typecode in self.pure:
assert typecode == 'PYMODULE'
# Transform a python module name into a file name.
name = name.replace('.', os.sep)
# Special case: modules have an implied filename to add.
if os.path.splitext(os.path.basename(path))[0] == '__init__':
name += os.sep + '__init__'
# Append the extension for the compiled result.
# In python 3.5 (PEP-488) .pyo files were replaced by
# .opt-1.pyc and .opt-2.pyc. However, it seems that for
# bytecode-only module distribution, we always need to
# use the .pyc extension.
name += '.pyc'
new_toc.append((name, path, typecode))
# Put the result of byte-compiling this TOC in datas. Mark all entries as data.
for name, path, typecode in compile_py_files(
new_toc, CONF['workpath']):
self.datas.append((name, path, 'DATA'))
# Store no source in the archive.
self.pure = TOC()
# Write warnings about missing modules.
self._write_warnings()
# Write debug information about hte graph
self._write_graph_debug()
def pyi_modgraph():
return initialize_modgraph()
def pyi_modgraph():
return initialize_modgraph()
