cpython_chunk_10.py

# Auto-generated from cpython_chunk_10.txt

TEXT_DATA = r"""
# The bytecode interpreter
This document describes the workings and implementation of the bytecode
interpreter, the part of python that executes compiled Python code. Its
entry point is in [Python/ceval.c](../Python/ceval.c).
At a high level, the interpreter consists of a loop that iterates over the
bytecode instructions, executing each of them via a switch statement that
has a case implementing each opcode. This switch statement is generated
from the instruction definitions in [Python/bytecodes.c](../Python/bytecodes.c)
which are written in [a DSL](../Tools/cases_generator/interpreter_definition.md)
developed for this purpose.
Recall that the [Python Compiler](compiler.md) produces a [`CodeObject`](code_objects.md),
which contains the bytecode instructions along with static data that is required to execute them,
such as the consts list, variable names,
[exception table](exception_handling.md#format-of-the-exception-table), and so on.
When the interpreter's
[`PyEval_EvalCode()`](https://docs.python.org/3.14/c-api/veryhigh.html#c.PyEval_EvalCode)
function is called to execute a `CodeObject`, it constructs a [`Frame`](frames.md) and calls
[`_PyEval_EvalFrame()`](https://docs.python.org/3.14/c-api/veryhigh.html#c.PyEval_EvalCode)
to execute the code object in this frame. The frame holds the dynamic state of the
`CodeObject`'s execution, including the instruction pointer, the globals and builtins.
It also has a reference to the `CodeObject` itself.
In addition to the frame, `_PyEval_EvalFrame()` also receives a
[`Thread State`](https://docs.python.org/3/c-api/init.html#c.PyThreadState)
object, `tstate`, which includes things like the exception state and the
recursion depth.  The thread state also provides access to the per-interpreter
state (`tstate->interp`), which has a pointer to the per-runtime (that is,
truly global) state (`tstate->interp->runtime`).
Finally, `_PyEval_EvalFrame()` receives an integer argument `throwflag`
which, when nonzero, indicates that the interpreter should just raise the current exception
(this is used in the implementation of
[`gen.throw`](https://docs.python.org/3.14/reference/expressions.html#generator.throw).
By default, [`_PyEval_EvalFrame()`](https://docs.python.org/3.14/c-api/veryhigh.html#c.PyEval_EvalCode)
simply calls [`_PyEval_EvalFrameDefault()`] to execute the frame. However, as per
[`PEP 523`](https://peps.python.org/pep-0523/) this is configurable by setting
`interp->eval_frame`. In the following, we describe the default function,
`_PyEval_EvalFrameDefault()`.
## Instruction decoding
The first task of the interpreter is to decode the bytecode instructions.
Bytecode is stored as an array of 16-bit code units (`_Py_CODEUNIT`).
Each code unit contains an 8-bit `opcode` and an 8-bit argument (`oparg`), both unsigned.
In order to make the bytecode format independent of the machine byte order when stored on disk,
`opcode` is always the first byte and `oparg` is always the second byte.
Macros are used to extract the `opcode` and `oparg` from a code unit
(`_Py_OPCODE(word)` and `_Py_OPARG(word)`).
Some instructions (for example, `NOP` or `POP_TOP`) have no argument -- in this case
we ignore `oparg`.
A simplified version of the interpreter's main loop looks like this:
```c
_Py_CODEUNIT *first_instr = code->co_code_adaptive;
_Py_CODEUNIT *next_instr = first_instr;
while (1) {
_Py_CODEUNIT word = *next_instr++;
unsigned char opcode = _Py_OPCODE(word);
unsigned int oparg = _Py_OPARG(word);
switch (opcode) {
// ... A case for each opcode ...
}
}
```
This loop iterates over the instructions, decoding each into its `opcode`
and `oparg`, and then executes the switch case that implements this `opcode`.
The instruction format supports 256 different opcodes, which is sufficient.
However, it also limits `oparg` to 8-bit values, which is too restrictive.
To overcome this, the `EXTENDED_ARG` opcode allows us to prefix any instruction
with one or more additional data bytes, which combine into a larger oparg.
For example, this sequence of code units:
EXTENDED_ARG  1
EXTENDED_ARG  0
LOAD_CONST    2
would set `opcode` to `LOAD_CONST` and `oparg` to `65538` (that is, `0x1_00_02`).
The compiler should limit itself to at most three `EXTENDED_ARG` prefixes, to allow the
resulting `oparg` to fit in 32 bits, but the interpreter does not check this.
In the following, a `code unit` is always two bytes, while an `instruction` is a
sequence of code units consisting of zero to three `EXTENDED_ARG` opcodes followed by
a primary opcode.
The following loop, to be inserted just above the `switch` statement, will make the above
snippet decode a complete instruction:
```c
while (opcode == EXTENDED_ARG) {
word = *next_instr++;
opcode = _Py_OPCODE(word);
oparg = (oparg << 8) | _Py_OPARG(word);
}
```
For various reasons we'll get to later (mostly efficiency, given that `EXTENDED_ARG`
is rare) the actual code is different.
## Jumps
Note that when the `switch` statement is reached, `next_instr` (the "instruction offset")
already points to the next instruction.
Thus, jump instructions can be implemented by manipulating `next_instr`:
- A jump forward (`JUMP_FORWARD`) sets `next_instr += oparg`.
- A jump backward (`JUMP_BACKWARD`) sets `next_instr -= oparg`.
## Inline cache entries
Some (specialized or specializable) instructions have an associated "inline cache".
The inline cache consists of one or more two-byte entries included in the bytecode
array as additional words following the `opcode`/`oparg` pair.
The size of the inline cache for a particular instruction is fixed by its `opcode`.
Moreover, the inline cache size for all instructions in a
[family of specialized/specializable instructions](#Specialization)
(for example, `LOAD_ATTR`, `LOAD_ATTR_SLOT`, `LOAD_ATTR_MODULE`) must all be
the same.  Cache entries are reserved by the compiler and initialized with zeros.
Although they are represented by code units, cache entries do not conform to the
`opcode` / `oparg` format.
If an instruction has an inline cache, the layout of its cache is described in
the instruction's definition in [`Python/bytecodes.c`](../Python/bytecodes.c).
The structs defined in [`pycore_code.h`](../Include/internal/pycore_code.h)
allow us to access the cache by casting `next_instr` to a pointer to the relevant
`struct`.  The size of such a `struct` must be independent of the machine
architecture, word size and alignment requirements.  For a 32-bit field, the
`struct` should use `_Py_CODEUNIT field[2]`.
The instruction implementation is responsible for advancing `next_instr` past the inline cache.
For example, if an instruction's inline cache is four bytes (that is, two code units) in size,
the code for the instruction must contain `next_instr += 2;`.
This is equivalent to a relative forward jump by that many code units.
(In the interpreter definition DSL, this is coded as `JUMPBY(n)`, where `n` is the number
of code units to jump, typically given as a named constant.)
Serializing non-zero cache entries would present a problem because the serialization
(:mod:`marshal`) format must be independent of the machine byte order.
More information about the use of inline caches can be found in
[PEP 659](https://peps.python.org/pep-0659/#ancillary-data).
## The evaluation stack
Most instructions read or write some data in the form of object references (`PyObject *`).
The CPython bytecode interpreter is a stack machine, meaning that its instructions operate
by pushing data onto and popping it off the stack.
The stack forms part of the frame for the code object. Its maximum depth is calculated
by the compiler and stored in the `co_stacksize` field of the code object, so that the
stack can be pre-allocated as a contiguous array of `PyObject*` pointers, when the frame
is created.
The stack effects of each instruction are also exposed through the
[opcode metadata](../Include/internal/pycore_opcode_metadata.h) through two
functions that report how many stack elements the instructions consumes,
and how many it produces (`_PyOpcode_num_popped` and `_PyOpcode_num_pushed`).
For example, the `BINARY_OP` instruction pops two objects from the stack and pushes the
result back onto the stack.
The stack grows up in memory; the operation `PUSH(x)` is equivalent to `*stack_pointer++ = x`,
whereas `x = POP()` means `x = *--stack_pointer`.
Overflow and underflow checks are active in debug mode, but are otherwise optimized away.
At any point during execution, the stack level is knowable based on the instruction pointer
alone, and some properties of each item on the stack are also known.
In particular, only a few instructions may push a `NULL` onto the stack, and the positions
that may be `NULL` are known.
A few other instructions (`GET_ITER`, `FOR_ITER`) push or pop an object that is known to
be an iterator.
Instruction sequences that do not allow statically knowing the stack depth are deemed illegal;
the bytecode compiler never generates such sequences.
For example, the following sequence is illegal, because it keeps pushing items on the stack:
LOAD_FAST 0
JUMP_BACKWARD 2
> [!NOTE]
> Do not confuse the evaluation stack with the call stack, which is used to implement calling
> and returning from functions.
## Error handling
When the implementation of an opcode raises an exception, it jumps to the
`exception_unwind` label in [Python/ceval.c](../Python/ceval.c).
The exception is then handled as described in the
[`exception handling documentation`](exception_handling.md#handling-exceptions).
## Python-to-Python calls
The `_PyEval_EvalFrameDefault()` function is recursive, because sometimes
the interpreter calls some C function that calls back into the interpreter.
In 3.10 and before, this was the case even when a Python function called
another Python function:
The `CALL` opcode would call the `tp_call` dispatch function of the
callee, which would extract the code object, create a new frame for the call
stack, and then call back into the interpreter.  This approach is very general
but consumes several C stack frames for each nested Python call, thereby
increasing the risk of an (unrecoverable) C stack overflow.
Since 3.11, the `CALL` instruction special-cases function objects to "inline"
the call.  When a call gets inlined, a new frame gets pushed onto the call
stack and the interpreter "jumps" to the start of the callee's bytecode.
When an inlined callee executes a `RETURN_VALUE` instruction, the frame is
popped off the call stack and the interpreter returns to its caller,
by popping a frame off the call stack and "jumping" to the return address.
There is a flag in the frame (`frame->is_entry`) that indicates whether
the frame was inlined (set if it wasn't).
If `RETURN_VALUE` finds this flag set, it performs the usual cleanup and
returns from `_PyEval_EvalFrameDefault()` altogether, to a C caller.
A similar check is performed when an unhandled exception occurs.
## The call stack
Up through 3.10, the call stack was implemented as a singly-linked list of
[frame objects](frames.md). This was expensive because each call would require a
heap allocation for the stack frame.
Since 3.11, frames are no longer fully-fledged objects. Instead, a leaner internal
`_PyInterpreterFrame` structure is used. Most frames are allocated contiguously in a
per-thread stack (see `_PyThreadState_PushFrame` in [Python/pystate.c](../Python/pystate.c)),
which improves memory locality and reduces overhead.
If the current `datastack_chunk` has enough space (`_PyThreadState_HasStackSpace`)
then the lightweight `_PyFrame_PushUnchecked` can be used instead of `_PyThreadState_PushFrame`.
Sometimes an actual `PyFrameObject` is needed, such as when Python code calls
`sys._getframe()` or an extension module calls
[`PyEval_GetFrame()`](https://docs.python.org/3/c-api/reflection.html#c.PyEval_GetFrame).
In this case we allocate a proper `PyFrameObject` and initialize it from the
`_PyInterpreterFrame`.
Things get more complicated when generators are involved, since those do not
follow the push/pop model. This includes async functions, which are based on
the same mechanism.  A generator object has space for a `_PyInterpreterFrame`
structure, including the variable-size part (used for locals and the eval stack).
When a generator (or async) function is first called, a special opcode
`RETURN_GENERATOR` is executed, which is responsible for creating the
generator object.  The generator object's `_PyInterpreterFrame` is initialized
with a copy of the current stack frame.  The current stack frame is then popped
off the frame stack and the generator object is returned.
(Details differ depending on the `is_entry` flag.)
When the generator is resumed, the interpreter pushes its `_PyInterpreterFrame`
onto the frame stack and resumes execution.
See also the [generators](generators.md) section.
<!--
## All sorts of variables
The bytecode compiler determines the scope in which each variable name is defined,
and generates instructions accordingly.  For example, loading a local variable
onto the stack is done using `LOAD_FAST`, while loading a global is done using
`LOAD_GLOBAL`.
The key types of variables are:
- fast locals: used in functions
- (slow or regular) locals: used in classes and at the top level
- globals and builtins: the compiler cannot distinguish between globals and
builtins (though at runtime, the specializing interpreter can)
- cells: used for nonlocal references
(TODO: Write the rest of this section. Alas, the author got distracted and won't have time to continue this for a while.)
-->
<!--
Other topics
------------
(TODO: Each of the following probably deserves its own section.)
- co_consts, co_names, co_varnames, and their ilk
- How calls work (how args are transferred, return, exceptions)
- Eval breaker (interrupts, GIL)
- Tracing
- Setting the current lineno (debugger-induced jumps)
- Specialization, inline caches etc.
-->
## Introducing a new bytecode instruction
It is occasionally necessary to add a new opcode in order to implement
a new feature or change the way that existing features are compiled.
This section describes the changes required to do this.
First, you must choose a name for the bytecode, implement it in
[`Python/bytecodes.c`](../Python/bytecodes.c) and add a documentation
entry in [`Doc/library/dis.rst`](../Doc/library/dis.rst).
Then run `make regen-cases` to assign a number for it (see
[`Include/opcode_ids.h`](../Include/opcode_ids.h)) and regenerate a
number of files with the actual implementation of the bytecode in
[`Python/generated_cases.c.h`](../Python/generated_cases.c.h) and
metadata about it in additional files.
With a new bytecode you must also change what is called the "magic number" for
.pyc files: bump the value of the variable `MAGIC_NUMBER` in
[`Lib/importlib/_bootstrap_external.py`](../Lib/importlib/_bootstrap_external.py).
Changing this number will lead to all .pyc files with the old `MAGIC_NUMBER`
to be recompiled by the interpreter on import.  Whenever `MAGIC_NUMBER` is
changed, the ranges in the `magic_values` array in
[`PC/launcher.c`](../PC/launcher.c) may also need to be updated.  Changes to
[`Lib/importlib/_bootstrap_external.py`](../Lib/importlib/_bootstrap_external.py)
will take effect only after running `make regen-importlib`.
> [!NOTE]
> Running `make regen-importlib` before adding the new bytecode target to
> [`Python/bytecodes.c`](../Python/bytecodes.c)
> (followed by `make regen-cases`) will result in an error. You should only run
> `make regen-importlib` after the new bytecode target has been added.
> [!NOTE]
> On Windows, running the `./build.bat` script will automatically
> regenerate the required files without requiring additional arguments.
Finally, you need to introduce the use of the new bytecode.  Update
[`Python/codegen.c`](../Python/codegen.c) to emit code with this bytecode.
Optimizations in [`Python/flowgraph.c`](../Python/flowgraph.c) may also
need to be updated.  If the new opcode affects a control flow or the block
stack, you may have to update the `frame_setlineno()` function in
[`Objects/frameobject.c`](../Objects/frameobject.c).  It may also be necessary
to update [`Lib/dis.py`](../Lib/dis.py) if the new opcode interprets its
argument in a special way (like `FORMAT_VALUE` or `MAKE_FUNCTION`).
If you make a change here that can affect the output of bytecode that
is already in existence and you do not change the magic number, make
sure to delete your old .py(c|o) files!  Even though you will end up changing
the magic number if you change the bytecode, while you are debugging your work
you may be changing the bytecode output without constantly bumping up the
magic number.  This can leave you with stale .pyc files that will not be
recreated.
Running `find . -name '*.py[co]' -exec rm -f '{}' +` should delete all .pyc
files you have, forcing new ones to be created and thus allow you test out your
new bytecode properly.  Run `make regen-importlib` for updating the
bytecode of frozen importlib files.  You have to run `make` again after this
to recompile the generated C files.
## Specialization
Bytecode specialization, which was introduced in
[PEP 659](https://peps.python.org/pep-0659/), speeds up program execution by
rewriting instructions based on runtime information. This is done by replacing
a generic instruction with a faster version that works for the case that this
program encounters. Each specializable instruction is responsible for rewriting
itself, using its [inline caches](#inline-cache-entries) for
bookkeeping.
When an adaptive instruction executes, it may attempt to specialize itself,
depending on the argument and the contents of its cache. This is done
by calling one of the `_Py_Specialize_XXX` functions in
[`Python/specialize.c`](../Python/specialize.c).
The specialized instructions are responsible for checking that the special-case
assumptions still apply, and de-optimizing back to the generic version if not.
## Families of instructions
A *family* of instructions consists of an adaptive instruction along with the
specialized instructions that it can be replaced by.
It has the following fundamental properties:
* It corresponds to a single instruction in the code
generated by the bytecode compiler.
* It has a single adaptive instruction that records an execution count and,
at regular intervals, attempts to specialize itself. If not specializing,
it executes the base implementation.
* It has at least one specialized form of the instruction that is tailored
for a particular value or set of values at runtime.
* All members of the family must have the same number of inline cache entries,
to ensure correct execution.
Individual family members do not need to use all of the entries,
but must skip over any unused entries when executing.
The current implementation also requires the following,
although these are not fundamental and may change:
* All families use one or more inline cache entries,
the first entry is always the counter.
* All instruction names should start with the name of the adaptive
instruction.
* Specialized forms should have names describing their specialization.
## Example family
The `LOAD_GLOBAL` instruction (in [Python/bytecodes.c](../Python/bytecodes.c))
already has an adaptive family that serves as a relatively simple example.
The `LOAD_GLOBAL` instruction performs adaptive specialization,
calling `_Py_Specialize_LoadGlobal()` when the counter reaches zero.
There are two specialized instructions in the family, `LOAD_GLOBAL_MODULE`
which is specialized for global variables in the module, and
`LOAD_GLOBAL_BUILTIN` which is specialized for builtin variables.
## Performance analysis
The benefit of a specialization can be assessed with the following formula:
`Tbase/Tadaptive`.
Where `Tbase` is the mean time to execute the base instruction,
and `Tadaptive` is the mean time to execute the specialized and adaptive forms.
`Tadaptive = (sum(Ti*Ni) + Tmiss*Nmiss)/(sum(Ni)+Nmiss)`
`Ti` is the time to execute the `i`th instruction in the family and `Ni` is
the number of times that instruction is executed.
`Tmiss` is the time to process a miss, including de-optimzation
and the time to execute the base instruction.
The ideal situation is where misses are rare and the specialized
forms are much faster than the base instruction.
`LOAD_GLOBAL` is near ideal, `Nmiss/sum(Ni) ≈ 0`.
In which case we have `Tadaptive ≈ sum(Ti*Ni)`.
Since we can expect the specialized forms `LOAD_GLOBAL_MODULE` and
`LOAD_GLOBAL_BUILTIN` to be much faster than the adaptive base instruction,
we would expect the specialization of `LOAD_GLOBAL` to be profitable.
## Design considerations
While `LOAD_GLOBAL` may be ideal, instructions like `LOAD_ATTR` and
`CALL_FUNCTION` are not. For maximum performance we want to keep `Ti`
low for all specialized instructions and `Nmiss` as low as possible.
Keeping `Nmiss` low means that there should be specializations for almost
all values seen by the base instruction. Keeping `sum(Ti*Ni)` low means
keeping `Ti` low which means minimizing branches and dependent memory
accesses (pointer chasing). These two objectives may be in conflict,
requiring judgement and experimentation to design the family of instructions.
The size of the inline cache should as small as possible,
without impairing performance, to reduce the number of
`EXTENDED_ARG` jumps, and to reduce pressure on the CPU's data cache.
### Gathering data
Before choosing how to specialize an instruction, it is important to gather
some data. What are the patterns of usage of the base instruction?
Data can best be gathered by instrumenting the interpreter. Since a
specialization function and adaptive instruction are going to be required,
instrumentation can most easily be added in the specialization function.
### Choice of specializations
The performance of the specializing adaptive interpreter relies on the
quality of specialization and keeping the overhead of specialization low.
Specialized instructions must be fast. In order to be fast,
specialized instructions should be tailored for a particular
set of values that allows them to:
1. Verify that incoming value is part of that set with low overhead.
2. Perform the operation quickly.
This requires that the set of values is chosen such that membership can be
tested quickly and that membership is sufficient to allow the operation to be
performed quickly.
For example, `LOAD_GLOBAL_MODULE` is specialized for `globals()`
dictionaries that have a keys with the expected version.
This can be tested quickly:
* `globals->keys->dk_version == expected_version`
and the operation can be performed quickly:
* `value = entries[cache->index].me_value;`.
Because it is impossible to measure the performance of an instruction without
also measuring unrelated factors, the assessment of the quality of a
specialization will require some judgement.
As a general rule, specialized instructions should be much faster than the
base instruction.
### Implementation of specialized instructions
In general, specialized instructions should be implemented in two parts:
1. A sequence of guards, each of the form
`DEOPT_IF(guard-condition-is-false, BASE_NAME)`.
2. The operation, which should ideally have no branches and
a minimum number of dependent memory accesses.
In practice, the parts may overlap, as data required for guards
can be re-used in the operation.
If there are branches in the operation, then consider further specialization
to eliminate the branches.
### Maintaining stats
Finally, take care that stats are gathered correctly.
After the last `DEOPT_IF` has passed, a hit should be recorded with
`STAT_INC(BASE_INSTRUCTION, hit)`.
After an optimization has been deferred in the adaptive instruction,
that should be recorded with `STAT_INC(BASE_INSTRUCTION, deferred)`.
Additional resources
--------------------
* Brandt Bucher's talk about the specializing interpreter at PyCon US 2023.
[Slides](https://github.com/brandtbucher/brandtbucher/blob/master/2023/04/21/inside_cpython_311s_new_specializing_adaptive_interpreter.pdf)
[Video](https://www.youtube.com/watch?v=PGZPSWZSkJI&t=1470s)

#!/usr/bin/env python3
import asyncio
import argparse
import json
import os
import platform
import re
import shlex
import shutil
import signal
import subprocess
import sys
import sysconfig
from asyncio import wait_for
from contextlib import asynccontextmanager
from datetime import datetime, timezone
from glob import glob
from os.path import abspath, basename, relpath
from pathlib import Path
from subprocess import CalledProcessError
from tempfile import TemporaryDirectory
SCRIPT_NAME = Path(__file__).name
ANDROID_DIR = Path(__file__).resolve().parent
PYTHON_DIR = ANDROID_DIR.parent
in_source_tree = (
ANDROID_DIR.name == "Android" and (PYTHON_DIR / "pyconfig.h.in").exists()
)
ENV_SCRIPT = ANDROID_DIR / "android-env.sh"
TESTBED_DIR = ANDROID_DIR / "testbed"
CROSS_BUILD_DIR = PYTHON_DIR / "cross-build"
HOSTS = ["aarch64-linux-android", "x86_64-linux-android"]
APP_ID = "org.python.testbed"
DECODE_ARGS = ("UTF-8", "backslashreplace")
try:
android_home = Path(os.environ['ANDROID_HOME'])
except KeyError:
sys.exit("The ANDROID_HOME environment variable is required.")
adb = Path(
f"{android_home}/platform-tools/adb"
+ (".exe" if os.name == "nt" else "")
)
gradlew = Path(
f"{TESTBED_DIR}/gradlew"
+ (".bat" if os.name == "nt" else "")
)
# Whether we've seen any output from Python yet.
python_started = False
# Buffer for verbose output which will be displayed only if a test fails and
# there has been no output from Python.
hidden_output = []
def log_verbose(context, line, stream=sys.stdout):
if context.verbose:
stream.write(line)
else:
hidden_output.append((stream, line))
def delete_glob(pattern):
# Path.glob doesn't accept non-relative patterns.
for path in glob(str(pattern)):
path = Path(path)
print(f"Deleting {path} ...")
if path.is_dir() and not path.is_symlink():
shutil.rmtree(path)
else:
path.unlink()
def subdir(*parts, create=False):
path = CROSS_BUILD_DIR.joinpath(*parts)
if not path.exists():
if not create:
sys.exit(
f"{path} does not exist. Create it by running the appropriate "
f"`configure` subcommand of {SCRIPT_NAME}.")
else:
path.mkdir(parents=True)
return path
def run(command, *, host=None, env=None, log=True, **kwargs):
kwargs.setdefault("check", True)
if env is None:
env = os.environ.copy()
if host:
host_env = android_env(host)
print_env(host_env)
env.update(host_env)
if log:
print(">", join_command(command))
return subprocess.run(command, env=env, **kwargs)
# Format a command so it can be copied into a shell. Like shlex.join, but also
# accepts arguments which are Paths, or a single string/Path outside of a list.
def join_command(args):
if isinstance(args, (str, Path)):
return str(args)
else:
return shlex.join(map(str, args))
# Format the environment so it can be pasted into a shell.
def print_env(env):
for key, value in sorted(env.items()):
print(f"export {key}={shlex.quote(value)}")
def android_env(host):
if host:
prefix = subdir(host) / "prefix"
else:
prefix = ANDROID_DIR / "prefix"
sysconfig_files = prefix.glob("lib/python*/_sysconfigdata__android_*.py")
sysconfig_filename = next(sysconfig_files).name
host = re.fullmatch(r"_sysconfigdata__android_(.+).py", sysconfig_filename)[1]
env_output = subprocess.run(
f"set -eu; "
f"HOST={host}; "
f"PREFIX={prefix}; "
f". {ENV_SCRIPT}; "
f"export",
check=True, shell=True, capture_output=True, encoding='utf-8',
).stdout
env = {}
for line in env_output.splitlines():
# We don't require every line to match, as there may be some other
# output from installing the NDK.
if match := re.search(
"^(declare -x |export )?(\\w+)=['\"]?(.*?)['\"]?$", line
):
key, value = match[2], match[3]
if os.environ.get(key) != value:
env[key] = value
if not env:
raise ValueError(f"Found no variables in {ENV_SCRIPT.name} output:\n"
+ env_output)
return env
def build_python_path():
"""The path to the build Python binary."""
build_dir = subdir("build")
binary = build_dir / "python"
if not binary.is_file():
binary = binary.with_suffix(".exe")
if not binary.is_file():
raise FileNotFoundError("Unable to find `python(.exe)` in "
f"{build_dir}")
return binary
def configure_build_python(context):
if context.clean:
clean("build")
os.chdir(subdir("build", create=True))
command = [relpath(PYTHON_DIR / "configure")]
if context.args:
command.extend(context.args)
run(command)
def make_build_python(context):
os.chdir(subdir("build"))
run(["make", "-j", str(os.cpu_count())])
# To create new builds of these dependencies, usually all that's necessary is to
# push a tag to the cpython-android-source-deps repository, and GitHub Actions
# will do the rest.
#
# If you're a member of the Python core team, and you'd like to be able to push
# these tags yourself, please contact Malcolm Smith or Russell Keith-Magee.
def unpack_deps(host, prefix_dir):
os.chdir(prefix_dir)
deps_url = "https://github.com/beeware/cpython-android-source-deps/releases/download"
for name_ver in ["bzip2-1.0.8-3", "libffi-3.4.4-3", "openssl-3.0.18-0",
"sqlite-3.50.4-0", "xz-5.4.6-1", "zstd-1.5.7-1"]:
filename = f"{name_ver}-{host}.tar.gz"
download(f"{deps_url}/{name_ver}/{filename}")
shutil.unpack_archive(filename)
os.remove(filename)
def download(url, target_dir="."):
out_path = f"{target_dir}/{basename(url)}"
run(["curl", "-Lf", "--retry", "5", "--retry-all-errors", "-o", out_path, url])
return out_path
def configure_host_python(context):
if context.clean:
clean(context.host)
host_dir = subdir(context.host, create=True)
prefix_dir = host_dir / "prefix"
if not prefix_dir.exists():
prefix_dir.mkdir()
unpack_deps(context.host, prefix_dir)
os.chdir(host_dir)
command = [
# Basic cross-compiling configuration
relpath(PYTHON_DIR / "configure"),
f"--host={context.host}",
f"--build={sysconfig.get_config_var('BUILD_GNU_TYPE')}",
f"--with-build-python={build_python_path()}",
"--without-ensurepip",
# Android always uses a shared libpython.
"--enable-shared",
"--without-static-libpython",
# Dependent libraries. The others are found using pkg-config: see
# android-env.sh.
f"--with-openssl={prefix_dir}",
]
if context.args:
command.extend(context.args)
run(command, host=context.host)
def make_host_python(context):
# The CFLAGS and LDFLAGS set in android-env include the prefix dir, so
# delete any previous Python installation to prevent it being used during
# the build.
host_dir = subdir(context.host)
prefix_dir = host_dir / "prefix"
for pattern in ("include/python*", "lib/libpython*", "lib/python*"):
delete_glob(f"{prefix_dir}/{pattern}")
# The Android environment variables were already captured in the Makefile by
# `configure`, and passing them again when running `make` may cause some
# flags to be duplicated. So we don't use the `host` argument here.
os.chdir(host_dir)
run(["make", "-j", str(os.cpu_count())])
# The `make install` output is very verbose and rarely useful, so
# suppress it by default.
run(
["make", "install", f"prefix={prefix_dir}"],
capture_output=not context.verbose,
)
def build_all(context):
steps = [configure_build_python, make_build_python, configure_host_python,
make_host_python]
for step in steps:
step(context)
def clean(host):
delete_glob(CROSS_BUILD_DIR / host)
def clean_all(context):
for host in HOSTS + ["build"]:
clean(host)
def setup_ci():
if "GITHUB_ACTIONS" in os.environ:
# Enable emulator hardware acceleration
# (https://github.blog/changelog/2024-04-02-github-actions-hardware-accelerated-android-virtualization-now-available/).
if platform.system() == "Linux":
run(
["sudo", "tee", "/etc/udev/rules.d/99-kvm4all.rules"],
input='KERNEL=="kvm", GROUP="kvm", MODE="0666", OPTIONS+="static_node=kvm"\n',
text=True,
)
run(["sudo", "udevadm", "control", "--reload-rules"])
run(["sudo", "udevadm", "trigger", "--name-match=kvm"])
# Free up disk space by deleting unused versions of the NDK
# (https://github.com/freakboy3742/pyspamsum/pull/108).
for line in ENV_SCRIPT.read_text().splitlines():
if match := re.fullmatch(r"ndk_version=(.+)", line):
ndk_version = match[1]
break
else:
raise ValueError(f"Failed to find NDK version in {ENV_SCRIPT.name}")
for item in (android_home / "ndk").iterdir():
if item.name[0].isdigit() and item.name != ndk_version:
delete_glob(item)
def setup_sdk():
sdkmanager = android_home / (
"cmdline-tools/latest/bin/sdkmanager"
+ (".bat" if os.name == "nt" else "")
)
# Gradle will fail if it needs to install an SDK package whose license
# hasn't been accepted, so pre-accept all licenses.
if not all((android_home / "licenses" / path).exists() for path in [
"android-sdk-arm-dbt-license", "android-sdk-license"
]):
run(
[sdkmanager, "--licenses"],
text=True,
capture_output=True,
input="y\n" * 100,
)
# Gradle may install this automatically, but we can't rely on that because
# we need to run adb within the logcat task.
if not adb.exists():
run([sdkmanager, "platform-tools"])
# To avoid distributing compiled artifacts without corresponding source code,
# the Gradle wrapper is not included in the CPython repository. Instead, we
# extract it from the Gradle GitHub repository.
def setup_testbed():
paths = ["gradlew", "gradlew.bat", "gradle/wrapper/gradle-wrapper.jar"]
if all((TESTBED_DIR / path).exists() for path in paths):
return
# The wrapper version isn't important, as any version of the wrapper can
# download any version of Gradle. The Gradle version actually used for the
# build is specified in testbed/gradle/wrapper/gradle-wrapper.properties.
version = "8.9.0"
for path in paths:
out_path = TESTBED_DIR / path
out_path.parent.mkdir(exist_ok=True)
download(
f"https://raw.githubusercontent.com/gradle/gradle/v{version}/{path}",
out_path.parent,
)
os.chmod(out_path, 0o755)
# run_testbed will build the app automatically, but it's useful to have this as
# a separate command to allow running the app outside of this script.
def build_testbed(context):
setup_sdk()
setup_testbed()
run(
[gradlew, "--console", "plain", "packageDebug", "packageDebugAndroidTest"],
cwd=TESTBED_DIR,
)
# Work around a bug involving sys.exit and TaskGroups
# (https://github.com/python/cpython/issues/101515).
def exit(*args):
raise MySystemExit(*args)
class MySystemExit(Exception):
pass
# The `test` subcommand runs all subprocesses through this context manager so
# that no matter what happens, they can always be cancelled from another task,
# and they will always be cleaned up on exit.
@asynccontextmanager
async def async_process(*args, **kwargs):
process = await asyncio.create_subprocess_exec(*args, **kwargs)
try:
yield process
finally:
if process.returncode is None:
# Allow a reasonably long time for Gradle to clean itself up,
# because we don't want stale emulators left behind.
timeout = 10
process.terminate()
try:
await wait_for(process.wait(), timeout)
except TimeoutError:
print(
f"Command {args} did not terminate after {timeout} seconds "
f" - sending SIGKILL"
)
process.kill()
# Even after killing the process we must still wait for it,
# otherwise we'll get the warning "Exception ignored in __del__".
await wait_for(process.wait(), timeout=1)
async def async_check_output(*args, **kwargs):
async with async_process(
*args, stdout=subprocess.PIPE, stderr=subprocess.PIPE, **kwargs
) as process:
stdout, stderr = await process.communicate()
if process.returncode == 0:
return stdout.decode(*DECODE_ARGS)
else:
raise CalledProcessError(
process.returncode, args,
stdout.decode(*DECODE_ARGS), stderr.decode(*DECODE_ARGS)
)
# Return a list of the serial numbers of connected devices. Emulators will have
# serials of the form "emulator-5678".
async def list_devices():
serials = []
header_found = False
lines = (await async_check_output(adb, "devices")).splitlines()
for line in lines:
# Ignore blank lines, and all lines before the header.
line = line.strip()
if line == "List of devices attached":
header_found = True
elif header_found and line:
try:
serial, status = line.split()
except ValueError:
raise ValueError(f"failed to parse {line!r}")
if status == "device":
serials.append(serial)
if not header_found:
raise ValueError(f"failed to parse {lines}")
return serials
async def find_device(context, initial_devices):
if context.managed:
print("Waiting for managed device - this may take several minutes")
while True:
new_devices = set(await list_devices()).difference(initial_devices)
if len(new_devices) == 0:
await asyncio.sleep(1)
elif len(new_devices) == 1:
serial = new_devices.pop()
print(f"Serial: {serial}")
return serial
else:
exit(f"Found more than one new device: {new_devices}")
else:
return context.connected
# An older version of this script in #121595 filtered the logs by UID instead.
# But logcat can't filter by UID until API level 31. If we ever switch back to
# filtering by UID, we'll also have to filter by time so we only show messages
# produced after the initial call to `stop_app`.
#
# We're more likely to miss the PID because it's shorter-lived, so there's a
# workaround in PythonSuite.kt to stop it being *too* short-lived.
async def find_pid(serial):
print("Waiting for app to start - this may take several minutes")
shown_error = False
while True:
try:
# `pidof` requires API level 24 or higher. The level 23 emulator
# includes it, but it doesn't work (it returns all processes).
pid = (await async_check_output(
adb, "-s", serial, "shell", "pidof", "-s", APP_ID
)).strip()
except CalledProcessError as e:
# If the app isn't running yet, pidof gives no output. So if there
# is output, there must have been some other error. However, this
# sometimes happens transiently, especially when running a managed
# emulator for the first time, so don't make it fatal.
if (e.stdout or e.stderr) and not shown_error:
print_called_process_error(e)
print("This may be transient, so continuing to wait")
shown_error = True
else:
# Some older devices (e.g. Nexus 4) return zero even when no process
# was found, so check whether we actually got any output.
if pid:
print(f"PID: {pid}")
return pid
# Loop fairly rapidly to avoid missing a short-lived process.
await asyncio.sleep(0.2)
async def logcat_task(context, initial_devices):
# Gradle may need to do some large downloads of libraries and emulator
# images. This will happen during find_device in --managed mode, or find_pid
# in --connected mode.
startup_timeout = 600
serial = await wait_for(find_device(context, initial_devices), startup_timeout)
pid = await wait_for(find_pid(serial), startup_timeout)
# `--pid` requires API level 24 or higher.
args = [adb, "-s", serial, "logcat", "--pid", pid,  "--format", "tag"]
logcat_started = False
async with async_process(
*args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
) as process:
while line := (await process.stdout.readline()).decode(*DECODE_ARGS):
if match := re.fullmatch(r"([A-Z])/(.*)", line, re.DOTALL):
logcat_started = True
level, message = match.groups()
else:
# If the regex doesn't match, this is either a logcat startup
# error, or the second or subsequent line of a multi-line
# message. Python won't produce multi-line messages, but other
# components might.
level, message = None, line
# Exclude high-volume messages which are rarely useful.
if context.verbose < 2 and "from python test_syslog" in message:
continue
# Put high-level messages on stderr so they're highlighted in the
# buildbot logs. This will include Python's own stderr.
stream = (
sys.stderr
if level in ["W", "E", "F"]  # WARNING, ERROR, FATAL (aka ASSERT)
else sys.stdout
)
# To simplify automated processing of the output, e.g. a buildbot
# posting a failure notice on a GitHub PR, we strip the level and
# tag indicators from Python's stdout and stderr.
for prefix in ["python.stdout: ", "python.stderr: "]:
if message.startswith(prefix):
global python_started
python_started = True
stream.write(message.removeprefix(prefix))
break
else:
# Non-Python messages add a lot of noise, but they may
# sometimes help explain a failure.
log_verbose(context, line, stream)
# If the device disconnects while logcat is running, which always
# happens in --managed mode, some versions of adb return non-zero.
# Distinguish this from a logcat startup error by checking whether we've
# received any logcat messages yet.
status = await wait_for(process.wait(), timeout=1)
if status != 0 and not logcat_started:
raise CalledProcessError(status, args)
def stop_app(serial):
run([adb, "-s", serial, "shell", "am", "force-stop", APP_ID], log=False)
async def gradle_task(context):
env = os.environ.copy()
if context.managed:
task_prefix = context.managed
else:
task_prefix = "connected"
env["ANDROID_SERIAL"] = context.connected
if context.ci_mode:
context.args[0:0] = [
# See _add_ci_python_opts in libregrtest/main.py.
"-W", "error", "-bb", "-E",
# Randomization is disabled because order-dependent failures are
# much less likely to pass on a rerun in single-process mode.
"-m", "test",
f"--{context.ci_mode}-ci", "--single-process", "--no-randomize"
]
if not any(arg in context.args for arg in ["-c", "-m"]):
context.args[0:0] = ["-m", "test"]
args = [
gradlew, "--console", "plain", f"{task_prefix}DebugAndroidTest",
] + [
f"-P{name}={value}"
for name, value in [
("python.sitePackages", context.site_packages),
("python.cwd", context.cwd),
(
"android.testInstrumentationRunnerArguments.pythonArgs",
json.dumps(context.args),
),
]
if value
]
if context.verbose >= 2:
args.append("--info")
log_verbose(context, f"> {join_command(args)}\n")
try:
async with async_process(
*args, cwd=TESTBED_DIR, env=env,
stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
) as process:
while line := (await process.stdout.readline()).decode(*DECODE_ARGS):
# Gradle may take several minutes to install SDK packages, so
# it's worth showing those messages even in non-verbose mode.
if line.startswith('Preparing "Install'):
sys.stdout.write(line)
else:
log_verbose(context, line)
status = await wait_for(process.wait(), timeout=1)
if status == 0:
exit(0)
else:
raise CalledProcessError(status, args)
finally:
# Gradle does not stop the tests when interrupted.
if context.connected:
stop_app(context.connected)
async def run_testbed(context):
setup_ci()
setup_sdk()
setup_testbed()
if context.managed:
# In this mode, Gradle will create a device with an unpredictable name.
# So we save a list of the running devices before starting Gradle, and
# find_device then waits for a new device to appear.
initial_devices = await list_devices()
else:
# In case the previous shutdown was unclean, make sure the app isn't
# running, otherwise we might show logs from a previous run. This is
# unnecessary in --managed mode, because Gradle creates a new emulator
# every time.
stop_app(context.connected)
initial_devices = None
try:
async with asyncio.TaskGroup() as tg:
tg.create_task(logcat_task(context, initial_devices))
tg.create_task(gradle_task(context))
except* MySystemExit as e:
raise SystemExit(*e.exceptions[0].args) from None
except* CalledProcessError as e:
# If Python produced no output, then the user probably wants to see the
# verbose output to explain why the test failed.
if not python_started:
for stream, line in hidden_output:
stream.write(line)
# Extract it from the ExceptionGroup so it can be handled by `main`.
raise e.exceptions[0]
def package_version(prefix_dir):
patchlevel_glob = f"{prefix_dir}/include/python*/patchlevel.h"
patchlevel_paths = glob(patchlevel_glob)
if len(patchlevel_paths) != 1:
sys.exit(f"{patchlevel_glob} matched {len(patchlevel_paths)} paths.")
for line in open(patchlevel_paths[0]):
if match := re.fullmatch(r'\s*#define\s+PY_VERSION\s+"(.+)"\s*', line):
version = match[1]
break
else:
sys.exit(f"Failed to find Python version in {patchlevel_paths[0]}.")
# If not building against a tagged commit, add a timestamp to the version.
# Follow the PyPA version number rules, as this will make it easier to
# process with other tools.
if version.endswith("+"):
version += datetime.now(timezone.utc).strftime("%Y%m%d.%H%M%S")
return version
def package(context):
prefix_dir = subdir(context.host, "prefix")
version = package_version(prefix_dir)
with TemporaryDirectory(prefix=SCRIPT_NAME) as temp_dir:
temp_dir = Path(temp_dir)
# Include all tracked files from the Android directory.
for line in run(
["git", "ls-files"],
cwd=ANDROID_DIR, capture_output=True, text=True, log=False,
).stdout.splitlines():
src = ANDROID_DIR / line
dst = temp_dir / line
dst.parent.mkdir(parents=True, exist_ok=True)
shutil.copy2(src, dst, follow_symlinks=False)
# Include anything from the prefix directory which could be useful
# either for embedding Python in an app, or building third-party
# packages against it.
for rel_dir, patterns in [
("include", ["openssl*", "python*", "sqlite*"]),
("lib", ["engines-3", "libcrypto*.so", "libpython*", "libsqlite*",
"libssl*.so", "ossl-modules", "python*"]),
("lib/pkgconfig", ["*crypto*", "*ssl*", "*python*", "*sqlite*"]),
]:
for pattern in patterns:
for src in glob(f"{prefix_dir}/{rel_dir}/{pattern}"):
dst = temp_dir / relpath(src, prefix_dir.parent)
dst.parent.mkdir(parents=True, exist_ok=True)
if Path(src).is_dir():
shutil.copytree(
src, dst, symlinks=True,
ignore=lambda *args: ["__pycache__"]
)
else:
shutil.copy2(src, dst, follow_symlinks=False)
# Strip debug information.
if not context.debug:
so_files = glob(f"{temp_dir}/**/*.so", recursive=True)
run([android_env(context.host)["STRIP"], *so_files], log=False)
dist_dir = subdir(context.host, "dist", create=True)
package_path = shutil.make_archive(
f"{dist_dir}/python-{version}-{context.host}", "gztar", temp_dir
)
print(f"Wrote {package_path}")
return package_path
def ci(context):
for step in [
configure_build_python,
make_build_python,
configure_host_python,
make_host_python,
package,
]:
caption = (
step.__name__.replace("_", " ")
.capitalize()
.replace("python", "Python")
)
print(f"::group::{caption}")
result = step(context)
if step is package:
package_path = result
print("::endgroup::")
if (
"GITHUB_ACTIONS" in os.environ
and (platform.system(), platform.machine()) != ("Linux", "x86_64")
):
print(
"Skipping tests: GitHub Actions does not support the Android "
"emulator on this platform."
)
else:
with TemporaryDirectory(prefix=SCRIPT_NAME) as temp_dir:
print("::group::Tests")
# Prove the package is self-contained by using it to run the tests.
shutil.unpack_archive(package_path, temp_dir)
launcher_args = [
"--managed", "maxVersion", "-v", f"--{context.ci_mode}-ci"
]
run(
["./android.py", "test", *launcher_args],
cwd=temp_dir
)
print("::endgroup::")
def env(context):
print_env(android_env(getattr(context, "host", None)))
# Handle SIGTERM the same way as SIGINT. This ensures that if we're terminated
# by the buildbot worker, we'll make an attempt to clean up our subprocesses.
def install_signal_handler():
def signal_handler(*args):
os.kill(os.getpid(), signal.SIGINT)
signal.signal(signal.SIGTERM, signal_handler)
def parse_args():
parser = argparse.ArgumentParser()
subcommands = parser.add_subparsers(dest="subcommand", required=True)
def add_parser(*args, **kwargs):
parser = subcommands.add_parser(*args, **kwargs)
parser.add_argument(
"-v", "--verbose", action="count", default=0,
help="Show verbose output. Use twice to be even more verbose.")
return parser
# Subcommands
build = add_parser(
"build", help="Run configure-build, make-build, configure-host and "
"make-host")
configure_build = add_parser(
"configure-build", help="Run `configure` for the build Python")
add_parser(
"make-build", help="Run `make` for the build Python")
configure_host = add_parser(
"configure-host", help="Run `configure` for Android")
make_host = add_parser(
"make-host", help="Run `make` for Android")
add_parser("clean", help="Delete all build directories")
add_parser("build-testbed", help="Build the testbed app")
test = add_parser("test", help="Run the testbed app")
package = add_parser("package", help="Make a release package")
ci = add_parser("ci", help="Run build, package and test")
env = add_parser("env", help="Print environment variables")
# Common arguments
for subcommand in [build, configure_build, configure_host, ci]:
subcommand.add_argument(
"--clean", action="store_true", default=False, dest="clean",
help="Delete the relevant build directories first")
host_commands = [build, configure_host, make_host, package, ci]
if in_source_tree:
host_commands.append(env)
for subcommand in host_commands:
subcommand.add_argument(
"host", metavar="HOST", choices=HOSTS,
help="Host triplet: choices=[%(choices)s]")
for subcommand in [build, configure_build, configure_host, ci]:
subcommand.add_argument("args", nargs="*",
help="Extra arguments to pass to `configure`")
# Test arguments
device_group = test.add_mutually_exclusive_group(required=True)
device_group.add_argument(
"--connected", metavar="SERIAL", help="Run on a connected device. "
"Connect it yourself, then get its serial from `adb devices`.")
device_group.add_argument(
"--managed", metavar="NAME", help="Run on a Gradle-managed device. "
"These are defined in `managedDevices` in testbed/app/build.gradle.kts.")
test.add_argument(
"--site-packages", metavar="DIR", type=abspath,
help="Directory to copy as the app's site-packages.")
test.add_argument(
"--cwd", metavar="DIR", type=abspath,
help="Directory to copy as the app's working directory.")
test.add_argument(
"args", nargs="*", help=f"Python command-line arguments. "
f"Separate them from {SCRIPT_NAME}'s own arguments with `--`. "
f"If neither -c nor -m are included, `-m test` will be prepended, "
f"which will run Python's own test suite.")
# Package arguments.
for subcommand in [package, ci]:
subcommand.add_argument(
"-g", action="store_true", default=False, dest="debug",
help="Include debug information in package")
# CI arguments
for subcommand in [test, ci]:
group = subcommand.add_mutually_exclusive_group(required=subcommand is ci)
group.add_argument(
"--fast-ci", action="store_const", dest="ci_mode", const="fast",
help="Add test arguments for GitHub Actions")
group.add_argument(
"--slow-ci", action="store_const", dest="ci_mode", const="slow",
help="Add test arguments for buildbots")
return parser.parse_args()
def main():
install_signal_handler()
# Under the buildbot, stdout is not a TTY, but we must still flush after
# every line to make sure our output appears in the correct order relative
# to the output of our subprocesses.
for stream in [sys.stdout, sys.stderr]:
stream.reconfigure(line_buffering=True)
context = parse_args()
dispatch = {
"configure-build": configure_build_python,
"make-build": make_build_python,
"configure-host": configure_host_python,
"make-host": make_host_python,
"build": build_all,
"clean": clean_all,
"build-testbed": build_testbed,
"test": run_testbed,
"package": package,
"ci": ci,
"env": env,
}
try:
result = dispatch[context.subcommand](context)
if asyncio.iscoroutine(result):
asyncio.run(result)
except CalledProcessError as e:
print_called_process_error(e)
sys.exit(1)
def print_called_process_error(e):
for stream_name in ["stdout", "stderr"]:
content = getattr(e, stream_name)
if isinstance(content, bytes):
content = content.decode(*DECODE_ARGS)
stream = getattr(sys, stream_name)
if content:
stream.write(content)
if not content.endswith("\n"):
stream.write("\n")
# shlex uses single quotes, so we surround the command with double quotes.
print(
f'Command "{join_command(e.cmd)}" returned exit status {e.returncode}'
)
if __name__ == "__main__":
main()

# Python for Android
If you obtained this README as part of a release package, then the only
applicable sections are "Prerequisites", "Testing", and "Using in your own app".
If you obtained this README as part of the CPython source tree, then you can
also follow the other sections to compile Python for Android yourself.
However, most app developers should not need to do any of these things manually.
Instead, use one of the tools listed
[here](https://docs.python.org/3/using/android.html), which will provide a much
easier experience.
## Prerequisites
If you already have an Android SDK installed, export the `ANDROID_HOME`
environment variable to point at its location. Otherwise, here's how to install
it:
* Download the "Command line tools" from <https://developer.android.com/studio>.
* Create a directory `android-sdk/cmdline-tools`, and unzip the command line
tools package into it.
* Rename `android-sdk/cmdline-tools/cmdline-tools` to
`android-sdk/cmdline-tools/latest`.
* `export ANDROID_HOME=/path/to/android-sdk`
The `android.py` script will automatically use the SDK's `sdkmanager` to install
any packages it needs.
The script also requires the following commands to be on the `PATH`:
* `curl`
* `java` (or set the `JAVA_HOME` environment variable)
## Building
Python can be built for Android on any POSIX platform supported by the Android
development tools, which currently means Linux or macOS.
First we'll make a "build" Python (for your development machine), then use it to
help produce a "host" Python for Android. So make sure you have all the usual
tools and libraries needed to build Python for your development machine.
The easiest way to do a build is to use the `android.py` script. You can either
have it perform the entire build process from start to finish in one step, or
you can do it in discrete steps that mirror running `configure` and `make` for
each of the two builds of Python you end up producing.
The discrete steps for building via `android.py` are:
```sh
./android.py configure-build
./android.py make-build
./android.py configure-host HOST
./android.py make-host HOST
```
`HOST` identifies which architecture to build. To see the possible values, run
`./android.py configure-host --help`.
To do all steps in a single command, run:
```sh
./android.py build HOST
```
In the end you should have a build Python in `cross-build/build`, and a host
Python in `cross-build/HOST`.
You can use `--` as a separator for any of the `configure`-related commands –
including `build` itself – to pass arguments to the underlying `configure`
call. For example, if you want a pydebug build that also caches the results from
`configure`, you can do:
```sh
./android.py build HOST -- -C --with-pydebug
```
## Packaging
After building an architecture as described in the section above, you can
package it for release with this command:
```sh
./android.py package HOST
```
`HOST` is defined in the section above.
This will generate a tarball in `cross-build/HOST/dist`, whose structure is
similar to the `Android` directory of the CPython source tree.
## Testing
The Python test suite can be run on Linux, macOS, or Windows.
On Linux, the emulator needs access to the KVM virtualization interface. This may
require adding your user to a group, or changing your udev rules. On GitHub
Actions, the test script will do this automatically using the commands shown
[here](https://github.blog/changelog/2024-04-02-github-actions-hardware-accelerated-android-virtualization-now-available/).
The test suite can usually be run on a device with 2 GB of RAM, but this is
borderline, so you may need to increase it to 4 GB. As of Android
Studio Koala, 2 GB is the default for all emulators, although the user interface
may indicate otherwise. Locate the emulator's directory under `~/.android/avd`,
and find `hw.ramSize` in both config.ini and hardware-qemu.ini. Either set these
manually to the same value, or use the Android Studio Device Manager, which will
update both files.
You can run the test suite either:
* Within the CPython repository, after doing a build as described above. On
Windows, you won't be able to do the build on the same machine, so you'll have
to copy the `cross-build/HOST/prefix` directory from somewhere else.
* Or by taking a release package built using the `package` command, extracting
it wherever you want, and using its own copy of `android.py`.
The test script supports the following modes:
* In `--connected` mode, it runs on a device or emulator you have already
connected to the build machine. List the available devices with
`$ANDROID_HOME/platform-tools/adb devices -l`, then pass a device ID to the
script like this:
```sh
./android.py test --connected emulator-5554
```
* In `--managed` mode, it uses a temporary headless emulator defined in the
`managedDevices` section of testbed/app/build.gradle.kts. This mode is slower,
but more reproducible.
We currently define two devices: `minVersion` and `maxVersion`, corresponding
to our minimum and maximum supported Android versions. For example:
```sh
./android.py test --managed maxVersion
```
By default, the only messages the script will show are Python's own stdout and
stderr. Add the `-v` option to also show Gradle output, and non-Python logcat
messages.
Any other arguments on the `android.py test` command line will be passed through
to `python -m test` – use `--` to separate them from android.py's own options.
See the [Python Developer's
Guide](https://devguide.python.org/testing/run-write-tests/) for common options
– most of them will work on Android, except for those that involve subprocesses,
such as `-j`.
Every time you run `android.py test`, changes in pure-Python files in the
repository's `Lib` directory will be picked up immediately. Changes in C files,
and architecture-specific files such as sysconfigdata, will not take effect
until you re-run `android.py make-host` or `build`.
The testbed app can also be used to test third-party packages. For more details,
run `android.py test --help`, paying attention to the options `--site-packages`,
`--cwd`, `-c` and `-m`.
## Using in your own app
See https://docs.python.org/3/using/android.html.

#include <android/log.h>
#include <errno.h>
#include <jni.h>
#include <pthread.h>
#include <Python.h>
#include <signal.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
static void throw_runtime_exception(JNIEnv *env, const char *message) {
(*env)->ThrowNew(
env,
(*env)->FindClass(env, "java/lang/RuntimeException"),
message);
}
static void throw_errno(JNIEnv *env, const char *error_prefix) {
char error_message[1024];
snprintf(error_message, sizeof(error_message),
"%s: %s", error_prefix, strerror(errno));
throw_runtime_exception(env, error_message);
}
// --- Stdio redirection ------------------------------------------------------
// Most apps won't need this, because the Python-level sys.stdout and sys.stderr
// are redirected to the Android logcat by Python itself. However, in the
// testbed it's useful to redirect the native streams as well, to debug problems
// in the Python startup or redirection process.
//
// Based on
// https://github.com/beeware/briefcase-android-gradle-template/blob/v0.3.11/%7B%7B%20cookiecutter.safe_formal_name%20%7D%7D/app/src/main/cpp/native-lib.cpp
typedef struct {
FILE *file;
int fd;
android_LogPriority priority;
char *tag;
int pipe[2];
} StreamInfo;
// The FILE member can't be initialized here because stdout and stderr are not
// compile-time constants. Instead, it's initialized immediately before the
// redirection.
static StreamInfo STREAMS[] = {
{NULL, STDOUT_FILENO, ANDROID_LOG_INFO, "native.stdout", {-1, -1}},
{NULL, STDERR_FILENO, ANDROID_LOG_WARN, "native.stderr", {-1, -1}},
{NULL, -1, ANDROID_LOG_UNKNOWN, NULL, {-1, -1}},
};
// The maximum length of a log message in bytes, including the level marker and
// tag, is defined as LOGGER_ENTRY_MAX_PAYLOAD in
// platform/system/logging/liblog/include/log/log.h. As of API level 30, messages
// longer than this will be be truncated by logcat. This limit has already been
// reduced at least once in the history of Android (from 4076 to 4068 between API
// level 23 and 26), so leave some headroom.
static const int MAX_BYTES_PER_WRITE = 4000;
static void *redirection_thread(void *arg) {
StreamInfo *si = (StreamInfo*)arg;
ssize_t read_size;
char buf[MAX_BYTES_PER_WRITE];
while ((read_size = read(si->pipe[0], buf, sizeof buf - 1)) > 0) {
buf[read_size] = '\0'; /* add null-terminator */
__android_log_write(si->priority, si->tag, buf);
}
return 0;
}
static char *redirect_stream(StreamInfo *si) {
/* make the FILE unbuffered, to ensure messages are never lost */
if (setvbuf(si->file, 0, _IONBF, 0)) {
return "setvbuf";
}
/* create the pipe and redirect the file descriptor */
if (pipe(si->pipe)) {
return "pipe";
}
if (dup2(si->pipe[1], si->fd) == -1) {
return "dup2";
}
/* start the logging thread */
pthread_t thr;
if ((errno = pthread_create(&thr, 0, redirection_thread, si))) {
return "pthread_create";
}
if ((errno = pthread_detach(thr))) {
return "pthread_detach";
}
return 0;
}
JNIEXPORT void JNICALL Java_org_python_testbed_PythonTestRunner_redirectStdioToLogcat(
JNIEnv *env, jobject obj
) {
STREAMS[0].file = stdout;
STREAMS[1].file = stderr;
for (StreamInfo *si = STREAMS; si->file; si++) {
char *error_prefix;
if ((error_prefix = redirect_stream(si))) {
throw_errno(env, error_prefix);
return;
}
}
}
// --- Python initialization ---------------------------------------------------
static char *init_signals() {
// Some tests use SIGUSR1, but that's blocked by default in an Android app in
// order to make it available to `sigwait` in the Signal Catcher thread.
// (https://cs.android.com/android/platform/superproject/+/android14-qpr3-release:art/runtime/signal_catcher.cc).
// That thread's functionality is only useful for debugging the JVM, so disabling
// it should not weaken the tests.
//
// There's no safe way of stopping the thread completely (#123982), but simply
// unblocking SIGUSR1 is enough to fix most tests.
//
// However, in tests that generate multiple different signals in quick
// succession, it's possible for SIGUSR1 to arrive while the main thread is busy
// running the C-level handler for a different signal. In that case, the SIGUSR1
// may be sent to the Signal Catcher thread instead, which will generate a log
// message containing the text "reacting to signal".
//
// Such tests may need to be changed in one of the following ways:
//   * Use a signal other than SIGUSR1 (e.g. test_stress_delivery_simultaneous in
//     test_signal.py).
//   * Send the signal to a specific thread rather than the whole process (e.g.
//     test_signals in test_threadsignals.py.
sigset_t set;
if (sigemptyset(&set)) {
return "sigemptyset";
}
if (sigaddset(&set, SIGUSR1)) {
return "sigaddset";
}
if ((errno = pthread_sigmask(SIG_UNBLOCK, &set, NULL))) {
return "pthread_sigmask";
}
return NULL;
}
static void throw_status(JNIEnv *env, PyStatus status) {
throw_runtime_exception(env, status.err_msg ? status.err_msg : "");
}
JNIEXPORT int JNICALL Java_org_python_testbed_PythonTestRunner_runPython(
JNIEnv *env, jobject obj, jstring home, jarray args
) {
const char *home_utf8 = (*env)->GetStringUTFChars(env, home, NULL);
char cwd[PATH_MAX];
snprintf(cwd, sizeof(cwd), "%s/%s", home_utf8, "cwd");
if (chdir(cwd)) {
throw_errno(env, "chdir");
return 1;
}
char *error_prefix;
if ((error_prefix = init_signals())) {
throw_errno(env, error_prefix);
return 1;
}
PyConfig config;
PyStatus status;
PyConfig_InitPythonConfig(&config);
jsize argc = (*env)->GetArrayLength(env, args);
const char *argv[argc + 1];
for (int i = 0; i < argc; i++) {
jobject arg = (*env)->GetObjectArrayElement(env, args, i);
argv[i] = (*env)->GetStringUTFChars(env, arg, NULL);
}
argv[argc] = NULL;
// PyConfig_SetBytesArgv "must be called before other methods, since the
// preinitialization configuration depends on command line arguments"
if (PyStatus_Exception(status = PyConfig_SetBytesArgv(&config, argc, (char**)argv))) {
throw_status(env, status);
return 1;
}
status = PyConfig_SetBytesString(&config, &config.home, home_utf8);
if (PyStatus_Exception(status)) {
throw_status(env, status);
return 1;
}
status = Py_InitializeFromConfig(&config);
if (PyStatus_Exception(status)) {
throw_status(env, status);
return 1;
}
return Py_RunMain();
}

cmake_minimum_required(VERSION 3.4.1)
project(testbed)
# Resolve variables from the command line.
string(
REPLACE {{triplet}} ${CMAKE_LIBRARY_ARCHITECTURE}
PYTHON_PREFIX_DIR ${PYTHON_PREFIX_DIR}
)
include_directories(${PYTHON_PREFIX_DIR}/include/python${PYTHON_VERSION})
link_directories(${PYTHON_PREFIX_DIR}/lib)
link_libraries(log python${PYTHON_VERSION})
add_library(main_activity SHARED main_activity.c)


"""