[CRIU] [PATCH 2/7] pycriu: add python package, v3

[Pavel Emelyanov]

On 12/02/2014 05:05 PM, Ruslan Kuprieiev wrote:
> On 12/02/2014 03:04 PM, Pavel Emelyanov wrote:
>>> diff --git a/pycriu/.gitignore b/pycriu/.gitignore
>>> new file mode 100644
>>> index 0000000..8d503da
>>> --- /dev/null
>>> +++ b/pycriu/.gitignore
>>> @@ -0,0 +1,3 @@
>>> +*_pb2.py
>>> +*.pyc
>>> +rpc.py
>>> diff --git a/pycriu/Makefile b/pycriu/Makefile
>>> new file mode 100644
>>> index 0000000..135fe1c
>>> --- /dev/null
>>> +++ b/pycriu/Makefile
>>> @@ -0,0 +1,16 @@
>>> +all: images rpc.py
>>> +
>>> +.PHONY: all images clean
>>> +
>>> +images:
>>> +	$(Q) $(MAKE) -C images all
>>> +
>>> +# rpc_pb2.py doesn't depend on any other file, so
>>> +# it is safe to rename it, dropping ugly _pb2 suffix.
>>> +rpc.py:
>>> +	$(Q) protoc -I=$(SRC_DIR)/protobuf/ --python_out=./ $(SRC_DIR)/protobuf/$(@:.py=.proto)
>>> +	$(Q) mv $(@:.py=_pb2.py) $@
>>
>> Why do we need rpc.py?
> 
> Because it is handy to have all pythonic criu stuff in one
> package called pycriu.

Hm... OK, will we then have a patch for p.haul that would get
rpc.py from CRIU rather than generating one itself?

>>
>>> +clean:
>>> +	$(Q) $(MAKE) -C images clean
>>> +	$(Q) $(RM) rpc.py *.pyc
>>> diff --git a/pycriu/__init__.py b/pycriu/__init__.py
>>> new file mode 100644
>>> index 0000000..5205973
>>> --- /dev/null
>>> +++ b/pycriu/__init__.py
>>> @@ -0,0 +1,2 @@
>>> +import rpc
>>> +import images
>>> diff --git a/pycriu/images/.gitignore b/pycriu/images/.gitignore
>>> new file mode 100644
>>> index 0000000..234bfe9
>>> --- /dev/null
>>> +++ b/pycriu/images/.gitignore
>>> @@ -0,0 +1,4 @@
>>> +*.pyc
>>> +*_pb2.py
>>> +magic.py
>>> +pb.py
>>> diff --git a/pycriu/images/Makefile b/pycriu/images/Makefile
>>> new file mode 100644
>>> index 0000000..6aa115e
>>> --- /dev/null
>>> +++ b/pycriu/images/Makefile
>>> @@ -0,0 +1,26 @@
>>> +all: pb.py protobuf magic.py
>>> +
>>> +.PHONY: all protobuf clean pb.py
>>> +
>>> +proto := $(filter-out $(SRC_DIR)/protobuf/rpc.proto, $(wildcard $(SRC_DIR)/protobuf/*.proto))
>>> +proto-py-modules := $(foreach m,$(proto),$(subst -,_,$(notdir $(m:.proto=_pb2))))
>>> +
>>> +# We don't need rpc_pb2.py here, as it is not related to the
>>> +# images.
>>> +# Unfortunately, we can't drop ugly _pb2 suffixes here, because
>>> +# some _pb2 files depend on others _pb2 files.
>>> +protobuf:
>>> +	$(Q) protoc -I=$(SRC_DIR)/protobuf --python_out=./ $(proto)
>>> +
>>> +magic.py: $(SRC_DIR)/scripts/magic-gen.py $(SRC_DIR)/include/magic.h
>>> +	$(E) "  GEN  " $@
>>> +	$(Q) python $^ $@
>>> +
>>> +pb.py: protobuf
>>> +	$(Q) echo "# Autogenerated. Do not edit!" > $@
>>> +	$(Q) for m in $(proto-py-modules); do \
>>> +		echo "from $$m import *" >> $@ ;\
>>> +	done
>>> +
>>> +clean:
>>> +	$(Q) $(RM) ./*_pb2.py ./*.pyc magic.py pb.py
>>> diff --git a/pycriu/images/__init__.py b/pycriu/images/__init__.py
>>> new file mode 100644
>>> index 0000000..379943b
>>> --- /dev/null
>>> +++ b/pycriu/images/__init__.py
>>> @@ -0,0 +1,3 @@
>>> +from magic import *
>>> +from images import *
>>> +from pb import *
>>> diff --git a/pycriu/images/images.py b/pycriu/images/images.py
>>> new file mode 100644
>>> index 0000000..afcaa87
>>> --- /dev/null
>>> +++ b/pycriu/images/images.py
>>> @@ -0,0 +1,208 @@
>>> +#!/bin/env python
>>> +import io
>>> +import google
>>> +import struct
>>> +import os
>>> +import sys
>>> +
>>> +from magic import *
>>> +from pb import *
>>> +
>>> +class _cr_desc:
>>> +	def __init__(self, magic, head, body='', single=False):
>>
>> I'm not sure the single bit is needed. But what's needed is some
>> sign that image may have raw data inside. For example:
>>
>> # ../crit convert --in dump/static/pipe00/6596/1/pipes-data.img
>> Traceback (most recent call last):
>>    ...
>>    File "/usr/lib/python2.7/site-packages/google/protobuf/internal/decoder.py", line 685, in _RaiseInvalidWireType
>>      raise _DecodeError('Tag had invalid wire type.')
>> google.protobuf.message.DecodeError: Tag had invalid wire type.
>>
>> That's because crit tries todecode raw pipe data as message.
>>
> 
> Not sure about single bit, but will try to optimize it.

But you don't use one anyway :)

> Yes, you are right about raw data. Will fix.

Thanks :)