Python3 bytecode is stale

Python3 bytecode is stale

[Stefan Agner]

Hi,

Recently I noticed that Python3 programs got rather slow. Running the
test application took 12.5 seconds. We are using a read-only file system
and I immediately suspected the bytecode cache. After mounting the
rootfs rw, from the second execution on running the test application
took below 4 seconds.

By simply executing a Hello World, Python declares all code objects as
being stale and tries to rewrite them:

# python3 -v -c 'print("Hello World")'
import _frozen_importlib # frozen
import _imp # builtin
import sys # builtin
import '_warnings' # <class '_frozen_importlib.BuiltinImporter'>
import '_thread' # <class '_frozen_importlib.BuiltinImporter'>
import '_weakref' # <class '_frozen_importlib.BuiltinImporter'>
import '_frozen_importlib_external' # <class
'_frozen_importlib.FrozenImporter'>
import '_io' # <class '_frozen_importlib.BuiltinImporter'>
import 'marshal' # <class '_frozen_importlib.BuiltinImporter'>
import 'posix' # <class '_frozen_importlib.BuiltinImporter'>
import _thread # previously loaded ('_thread')
import '_thread' # <class '_frozen_importlib.BuiltinImporter'>
import _weakref # previously loaded ('_weakref')
import '_weakref' # <class '_frozen_importlib.BuiltinImporter'>
# installing zipimport hook
import 'zipimport' # <class '_frozen_importlib.BuiltinImporter'>
# installed zipimport hook
# bytecode is stale for 'encodings'
# code object from /usr/lib/python3.5/encodings/__init__.py
# could not create
'/usr/lib/python3.5/encodings/__pycache__/__init__.cpython-35.pyc':
OSError(30, 'Read-only file system')
# wrote
'/usr/lib/python3.5/encodings/__pycache__/__init__.cpython-35.pyc'
# bytecode is stale for 'codecs'
# code object from /usr/lib/python3.5/codecs.py
# could not create
'/usr/lib/python3.5/__pycache__/codecs.cpython-35.pyc': OSError(30,
'Read-only file system')
# wrote '/usr/lib/python3.5/__pycache__/codecs.cpython-35.pyc'
import '_codecs' # <class '_frozen_importlib.BuiltinImporter'>
import 'codecs' # <_frozen_importlib_external.SourceFileLoader object at
0xb6b33130>
# bytecode is stale for 'encodings.aliases'
# code object from /usr/lib/python3.5/encodings/aliases.py
# could not create
'/usr/lib/python3.5/encodings/__pycache__/aliases.cpython-35.pyc':
OSError(30, 'Read-only file system')
# wrote
'/usr/lib/python3.5/encodings/__pycache__/aliases.cpython-35.pyc'
import 'encodings.aliases' #
<_frozen_importlib_external.SourceFileLoader object at 0xb6b3d9d0>
import 'encodings' # <_frozen_importlib_external.SourceFileLoader object
at 0xb6b154b0>
# bytecode is stale for 'encodings.ascii'
# code object from /usr/lib/python3.5/encodings/ascii.py
# could not create
'/usr/lib/python3.5/encodings/__pycache__/ascii.cpython-35.pyc':
OSError(30, 'Read-only file system')
# wrote '/usr/lib/python3.5/encodings/__pycache__/ascii.cpython-35.pyc'
import 'encodings.ascii' # <_frozen_importlib_external.SourceFileLoader
object at 0xb6b3db10>
import '_signal' # <class '_frozen_importlib.BuiltinImporter'>
....


This is on a Armv7 device. OpenEmbedded layers are current master.
Python version is 3.5.6. Is this a known issue? Any idea?

--
Stefan

Re: Python3 bytecode is stale

[Mike Looijmans]

Most likely the date/time stamps on the files are such that the "py" files 
appear to be newer.

A very simple fix would be to simply delete the ".py" files, they aren't 
really needed anyway so you'll save a lot of space in the process.

To do this compile-time, you could add something like this in a bbappend:
PACKAGES =+ "${PN}-src"
FILES_${PN}-src = "{installdir}/*.py {installdir}/*/*.py"

That'll push the "py" files into another package and prevent them being 
installed by default. Change "{installdir}" into something more appropriate.
Of just delete them in a do_install_append.


On 25-01-19 14:46, Stefan Agner wrote:
> Hi,
> 
> Recently I noticed that Python3 programs got rather slow. Running the
> test application took 12.5 seconds. We are using a read-only file system
> and I immediately suspected the bytecode cache. After mounting the
> rootfs rw, from the second execution on running the test application
> took below 4 seconds.
> 
> By simply executing a Hello World, Python declares all code objects as
> being stale and tries to rewrite them:
> 
> # python3 -v -c 'print("Hello World")'
> import _frozen_importlib # frozen
> import _imp # builtin
> import sys # builtin
> import '_warnings' # <class '_frozen_importlib.BuiltinImporter'>
> import '_thread' # <class '_frozen_importlib.BuiltinImporter'>
> import '_weakref' # <class '_frozen_importlib.BuiltinImporter'>
> import '_frozen_importlib_external' # <class
> '_frozen_importlib.FrozenImporter'>
> import '_io' # <class '_frozen_importlib.BuiltinImporter'>
> import 'marshal' # <class '_frozen_importlib.BuiltinImporter'>
> import 'posix' # <class '_frozen_importlib.BuiltinImporter'>
> import _thread # previously loaded ('_thread')
> import '_thread' # <class '_frozen_importlib.BuiltinImporter'>
> import _weakref # previously loaded ('_weakref')
> import '_weakref' # <class '_frozen_importlib.BuiltinImporter'>
> # installing zipimport hook
> import 'zipimport' # <class '_frozen_importlib.BuiltinImporter'>
> # installed zipimport hook
> # bytecode is stale for 'encodings'
> # code object from /usr/lib/python3.5/encodings/__init__.py
> # could not create
> '/usr/lib/python3.5/encodings/__pycache__/__init__.cpython-35.pyc':
> OSError(30, 'Read-only file system')
> # wrote
> '/usr/lib/python3.5/encodings/__pycache__/__init__.cpython-35.pyc'
> # bytecode is stale for 'codecs'
> # code object from /usr/lib/python3.5/codecs.py
> # could not create
> '/usr/lib/python3.5/__pycache__/codecs.cpython-35.pyc': OSError(30,
> 'Read-only file system')
> # wrote '/usr/lib/python3.5/__pycache__/codecs.cpython-35.pyc'
> import '_codecs' # <class '_frozen_importlib.BuiltinImporter'>
> import 'codecs' # <_frozen_importlib_external.SourceFileLoader object at
> 0xb6b33130>
> # bytecode is stale for 'encodings.aliases'
> # code object from /usr/lib/python3.5/encodings/aliases.py
> # could not create
> '/usr/lib/python3.5/encodings/__pycache__/aliases.cpython-35.pyc':
> OSError(30, 'Read-only file system')
> # wrote
> '/usr/lib/python3.5/encodings/__pycache__/aliases.cpython-35.pyc'
> import 'encodings.aliases' #
> <_frozen_importlib_external.SourceFileLoader object at 0xb6b3d9d0>
> import 'encodings' # <_frozen_importlib_external.SourceFileLoader object
> at 0xb6b154b0>
> # bytecode is stale for 'encodings.ascii'
> # code object from /usr/lib/python3.5/encodings/ascii.py
> # could not create
> '/usr/lib/python3.5/encodings/__pycache__/ascii.cpython-35.pyc':
> OSError(30, 'Read-only file system')
> # wrote '/usr/lib/python3.5/encodings/__pycache__/ascii.cpython-35.pyc'
> import 'encodings.ascii' # <_frozen_importlib_external.SourceFileLoader
> object at 0xb6b3db10>
> import '_signal' # <class '_frozen_importlib.BuiltinImporter'>
> ....
> 
> 
> This is on a Armv7 device. OpenEmbedded layers are current master.
> Python version is 3.5.6. Is this a known issue? Any idea?
> 
> --
> Stefan
> 

Re: Python3 bytecode is stale

[Stefan Agner]

On 25.01.2019 15:35, Mike Looijmans wrote:
> Most likely the date/time stamps on the files are such that the "py" files 
> appear to be newer.

Indeed, this seems to be the issue. We are using OSTree, which removes
mtimes. Ricardo (now on CC) also pointed me to the relevant issue in
OSTree:
https://github.com/ostreedev/ostree/issues/1469

> 
> A very simple fix would be to simply delete the ".py" files, they aren't 
> really needed anyway so you'll save a lot of space in the process.
> 
> To do this compile-time, you could add something like this in a bbappend:
> PACKAGES =+ "${PN}-src"
> FILES_${PN}-src = "{installdir}/*.py {installdir}/*/*.py"
> 
> That'll push the "py" files into another package and prevent them being 
> installed by default. Change "{installdir}" into something more appropriate.
> Of just delete them in a do_install_append.

Hm, interesting idea, will give this a try.

Wouldn't be something like this be a nice OpenEmbedded addition in
general? Optimizing embedded rootfs size by deleting py files seems to
be a common use case.

In a simple test deleting all py files on my rootfs did not go well (the
application does not start anymore). I need to investigate further.

--
Stefan

> 
> 
> On 25-01-19 14:46, Stefan Agner wrote:
>> Hi,
>>
>> Recently I noticed that Python3 programs got rather slow. Running the
>> test application took 12.5 seconds. We are using a read-only file system
>> and I immediately suspected the bytecode cache. After mounting the
>> rootfs rw, from the second execution on running the test application
>> took below 4 seconds.
>>
>> By simply executing a Hello World, Python declares all code objects as
>> being stale and tries to rewrite them:
>>
>> # python3 -v -c 'print("Hello World")'
>> import _frozen_importlib # frozen
>> import _imp # builtin
>> import sys # builtin
>> import '_warnings' # <class '_frozen_importlib.BuiltinImporter'>
>> import '_thread' # <class '_frozen_importlib.BuiltinImporter'>
>> import '_weakref' # <class '_frozen_importlib.BuiltinImporter'>
>> import '_frozen_importlib_external' # <class
>> '_frozen_importlib.FrozenImporter'>
>> import '_io' # <class '_frozen_importlib.BuiltinImporter'>
>> import 'marshal' # <class '_frozen_importlib.BuiltinImporter'>
>> import 'posix' # <class '_frozen_importlib.BuiltinImporter'>
>> import _thread # previously loaded ('_thread')
>> import '_thread' # <class '_frozen_importlib.BuiltinImporter'>
>> import _weakref # previously loaded ('_weakref')
>> import '_weakref' # <class '_frozen_importlib.BuiltinImporter'>
>> # installing zipimport hook
>> import 'zipimport' # <class '_frozen_importlib.BuiltinImporter'>
>> # installed zipimport hook
>> # bytecode is stale for 'encodings'
>> # code object from /usr/lib/python3.5/encodings/__init__.py
>> # could not create
>> '/usr/lib/python3.5/encodings/__pycache__/__init__.cpython-35.pyc':
>> OSError(30, 'Read-only file system')
>> # wrote
>> '/usr/lib/python3.5/encodings/__pycache__/__init__.cpython-35.pyc'
>> # bytecode is stale for 'codecs'
>> # code object from /usr/lib/python3.5/codecs.py
>> # could not create
>> '/usr/lib/python3.5/__pycache__/codecs.cpython-35.pyc': OSError(30,
>> 'Read-only file system')
>> # wrote '/usr/lib/python3.5/__pycache__/codecs.cpython-35.pyc'
>> import '_codecs' # <class '_frozen_importlib.BuiltinImporter'>
>> import 'codecs' # <_frozen_importlib_external.SourceFileLoader object at
>> 0xb6b33130>
>> # bytecode is stale for 'encodings.aliases'
>> # code object from /usr/lib/python3.5/encodings/aliases.py
>> # could not create
>> '/usr/lib/python3.5/encodings/__pycache__/aliases.cpython-35.pyc':
>> OSError(30, 'Read-only file system')
>> # wrote
>> '/usr/lib/python3.5/encodings/__pycache__/aliases.cpython-35.pyc'
>> import 'encodings.aliases' #
>> <_frozen_importlib_external.SourceFileLoader object at 0xb6b3d9d0>
>> import 'encodings' # <_frozen_importlib_external.SourceFileLoader object
>> at 0xb6b154b0>
>> # bytecode is stale for 'encodings.ascii'
>> # code object from /usr/lib/python3.5/encodings/ascii.py
>> # could not create
>> '/usr/lib/python3.5/encodings/__pycache__/ascii.cpython-35.pyc':
>> OSError(30, 'Read-only file system')
>> # wrote '/usr/lib/python3.5/encodings/__pycache__/ascii.cpython-35.pyc'
>> import 'encodings.ascii' # <_frozen_importlib_external.SourceFileLoader
>> object at 0xb6b3db10>
>> import '_signal' # <class '_frozen_importlib.BuiltinImporter'>
>> ....
>>
>>
>> This is on a Armv7 device. OpenEmbedded layers are current master.
>> Python version is 3.5.6. Is this a known issue? Any idea?
>>
>> --
>> Stefan
>>

Re: Python3 bytecode is stale

[Mike Looijmans]

On 25-01-19 18:26, Stefan Agner wrote:
> On 25.01.2019 15:35, Mike Looijmans wrote:
>> Most likely the date/time stamps on the files are such that the "py" files
>> appear to be newer.
> 
> Indeed, this seems to be the issue. We are using OSTree, which removes
> mtimes. Ricardo (now on CC) also pointed me to the relevant issue in
> OSTree:
> https://github.com/ostreedev/ostree/issues/1469
> 
>>
>> A very simple fix would be to simply delete the ".py" files, they aren't
>> really needed anyway so you'll save a lot of space in the process.
>>
>> To do this compile-time, you could add something like this in a bbappend:
>> PACKAGES =+ "${PN}-src"
>> FILES_${PN}-src = "{installdir}/*.py {installdir}/*/*.py"
>>
>> That'll push the "py" files into another package and prevent them being
>> installed by default. Change "{installdir}" into something more appropriate.
>> Of just delete them in a do_install_append.
> 
> Hm, interesting idea, will give this a try.
> 
> Wouldn't be something like this be a nice OpenEmbedded addition in
> general? Optimizing embedded rootfs size by deleting py files seems to
> be a common use case.

Here's a bbappend example for python:

https://github.com/OpenPLi/openpli-oe-core/blob/develop/meta-openpli/recipes-devtools/python/python_2.7.13.bbappend#L20

> In a simple test deleting all py files on my rootfs did not go well (the
> application does not start anymore). I need to investigate further.

The "main" module (the one started from commandline) needs to be in .py 
format still, otherwise you cannot "run" it. As a workaround, you can 
run it using "python -m", or make sure that that one .py file gets 
installed.

It's actually even possible to put the pyc files into a ZIP file and 
load directly from that at runtime. This reduces the number of files. On 
a compressed filesystem, this won't make a difference of course, and is 
likely to increase the total size.

--
M.

> 
> --
> Stefan
> 
>>
>>
>> On 25-01-19 14:46, Stefan Agner wrote:
>>> Hi,
>>>
>>> Recently I noticed that Python3 programs got rather slow. Running the
>>> test application took 12.5 seconds. We are using a read-only file system
>>> and I immediately suspected the bytecode cache. After mounting the
>>> rootfs rw, from the second execution on running the test application
>>> took below 4 seconds.
>>>
>>> By simply executing a Hello World, Python declares all code objects as
>>> being stale and tries to rewrite them:
>>>
>>> # python3 -v -c 'print("Hello World")'
>>> import _frozen_importlib # frozen
>>> import _imp # builtin
>>> import sys # builtin
>>> import '_warnings' # <class '_frozen_importlib.BuiltinImporter'>
>>> import '_thread' # <class '_frozen_importlib.BuiltinImporter'>
>>> import '_weakref' # <class '_frozen_importlib.BuiltinImporter'>
>>> import '_frozen_importlib_external' # <class
>>> '_frozen_importlib.FrozenImporter'>
>>> import '_io' # <class '_frozen_importlib.BuiltinImporter'>
>>> import 'marshal' # <class '_frozen_importlib.BuiltinImporter'>
>>> import 'posix' # <class '_frozen_importlib.BuiltinImporter'>
>>> import _thread # previously loaded ('_thread')
>>> import '_thread' # <class '_frozen_importlib.BuiltinImporter'>
>>> import _weakref # previously loaded ('_weakref')
>>> import '_weakref' # <class '_frozen_importlib.BuiltinImporter'>
>>> # installing zipimport hook
>>> import 'zipimport' # <class '_frozen_importlib.BuiltinImporter'>
>>> # installed zipimport hook
>>> # bytecode is stale for 'encodings'
>>> # code object from /usr/lib/python3.5/encodings/__init__.py
>>> # could not create
>>> '/usr/lib/python3.5/encodings/__pycache__/__init__.cpython-35.pyc':
>>> OSError(30, 'Read-only file system')
>>> # wrote
>>> '/usr/lib/python3.5/encodings/__pycache__/__init__.cpython-35.pyc'
>>> # bytecode is stale for 'codecs'
>>> # code object from /usr/lib/python3.5/codecs.py
>>> # could not create
>>> '/usr/lib/python3.5/__pycache__/codecs.cpython-35.pyc': OSError(30,
>>> 'Read-only file system')
>>> # wrote '/usr/lib/python3.5/__pycache__/codecs.cpython-35.pyc'
>>> import '_codecs' # <class '_frozen_importlib.BuiltinImporter'>
>>> import 'codecs' # <_frozen_importlib_external.SourceFileLoader object at
>>> 0xb6b33130>
>>> # bytecode is stale for 'encodings.aliases'
>>> # code object from /usr/lib/python3.5/encodings/aliases.py
>>> # could not create
>>> '/usr/lib/python3.5/encodings/__pycache__/aliases.cpython-35.pyc':
>>> OSError(30, 'Read-only file system')
>>> # wrote
>>> '/usr/lib/python3.5/encodings/__pycache__/aliases.cpython-35.pyc'
>>> import 'encodings.aliases' #
>>> <_frozen_importlib_external.SourceFileLoader object at 0xb6b3d9d0>
>>> import 'encodings' # <_frozen_importlib_external.SourceFileLoader object
>>> at 0xb6b154b0>
>>> # bytecode is stale for 'encodings.ascii'
>>> # code object from /usr/lib/python3.5/encodings/ascii.py
>>> # could not create
>>> '/usr/lib/python3.5/encodings/__pycache__/ascii.cpython-35.pyc':
>>> OSError(30, 'Read-only file system')
>>> # wrote '/usr/lib/python3.5/encodings/__pycache__/ascii.cpython-35.pyc'
>>> import 'encodings.ascii' # <_frozen_importlib_external.SourceFileLoader
>>> object at 0xb6b3db10>
>>> import '_signal' # <class '_frozen_importlib.BuiltinImporter'>
>>> ....
>>>
>>>
>>> This is on a Armv7 device. OpenEmbedded layers are current master.
>>> Python version is 3.5.6. Is this a known issue? Any idea?
>>>
>>> --
>>> Stefan
>>>


-- 
Mike Looijmans