New issue 3105: Help in understanding code unmarshaling https://bitbucket.org/pypy/pypy/issues/3105/help-in-understanding-code-unmarshaling
Rocky Bernstein: I am trying to get PyPy decompilation working on PyPy 7.1.1. Actually, though the problem is in unmarshaling code objects. I have tried to create a simple example which shows the problem, but alas, I am unable to do this. So the below is a little bit long in description, and I apologize for that. . Lets first start out with this PyPy 3.6 byte compiled file fragment that has been has rendered. You can find the full file [here](https://github.com/rocky/python-uncompyle6/blob/pypy3.6/test/bytecode_pypy3.6/11_classbug.pyc). ``` $ uncompyle6 python-uncompyle6/test/bytecode_pypy3.6/11_classbug.pyc 00000000: a000 0d0a 8aa0 3257 8b00 0000 6300 0000 ......2W....c... 00000010: 0000 0000 0000 0000 0004 0000 0040 0000 .............@.. 00000020: 0073 1400 0000 4700 6400 6401 8400 6401 .s....G.d.d...d. 00000030: 6500 8303 5a01 6402 5300 2903 6301 0000 e...Z.d.S.).c... 00000040: 0000 0000 0001 0000 0003 0000 0000 0000 ................ 00000050: 0073 1c00 0000 6500 5a01 6401 5a02 8700 .s....e.Z.d.Z... 00000060: 6601 6402 6403 8408 5a03 8700 0400 5a04 f.d.d...Z.....Z. 00000070: 5300 2904 4ef4 1200 0000 5f54 656d 706c S.).N....._Templ 00000080: 6174 654d 6574 6163 6c61 7373 6304 0000 ateMetaclassc... 00000090: 0000 0000 0004 0000 0005 0000 0003 0000 ................ 000000a0: 0073 1800 0000 7400 7401 7c00 8302 c902 .s....t.t.|..... 000000b0: 7c01 7c02 7c03 ca03 0100 6400 5300 2901 |.|.|.....d.S.). 000000c0: 4e29 03f4 0500 0000 7375 7065 7272 0000 N)......superr.. 000000d0: 0000 f408 0000 005f 5f69 6e69 745f 5f29 .......__init__) 000000e0: 04f4 0300 0000 636c 73f4 0400 0000 6e61 ......cls.....na 000000f0: 6d65 f505 0000 0062 6173 6573 f503 0000 me.....bases.... 00000100: 0064 6374 2901 f409 0000 005f 5f63 6c61 .dct)......__cla 00000110: 7373 5f5f 2900 7520 0000 0073 696d 706c ss__).u ...simpl 00000120: 655f 736f 7572 6365 2f64 6566 2f31 315f e_source/def/11_ 00000130: 636c 6173 7362 7567 2e70 7972 0200 0000 classbug.pyr.... ^^ ^^ 00000140: 0200 0000 7302 0000 0000 01f5 1b00 0000 ....s........... ``` I have underlined the important part of the code I will focus on below. Here is a fragment of a debug session of the cross-python disassembly called [ ](http://github.com/rocky/xdis) To compare against, here is CPython 3.6 with corresponding but different bytecode, which can e found [ here](http://github.com/rocky/xdis). Below, inside a debugger ere the reference to `co_code` after `co_filename` is to object 3, and this is correct. ``` 00000100: 696d 706c 655f 736f 7572 6365 2f64 6566 imple_source/def 00000110: 2f31 315f 636c 6173 7362 7567 2e70 7972 /11_classbug.pyr ^^ 00000120: 0300 0000 0200 0000 7302 0000 0000 017a ........s......z ^^ 00000130: 1b5f 5465 6d70 6c61 7465 4d65 7461 636c ._TemplateMetacl 00000140: 6173 732e 5f5f 696e 6974 5f5f 2905 da08 ass.__init__)... 00000150: 5f5f 6e61 6d65 5f5f da0a 5f5f 6d6f 6475 __name__..__modu 00000160: 6c65 5f5f da0c 5f5f 7175 616c 6e61 6d65 le__..__qualname 00000170: 5f5f 7203 0000 00da 0d5f 5f63 6c61 7373 __r......__class 00000180: 6365 6c6c 5f5f 7208 0000 0072 0800 0000 cell__r....r.... 00000190: 2901 7207 0000 0072 0900 0000 7201 0000 ).r....r....r... 000001a0: 0001 0000 0073 0200 0000 0801 7201 0000 .....s......r... 000001b0: 004e 2902 da04 7479 7065 7201 0000 0072 .N)...typer....r 000001c0: 0800 0000 7208 0000 0072 0800 0000 7209 ....r....r....r. 000001d0: 0000 00da 083c 6d6f 6475 6c65 3e01 0000 ........ 000001e0: (python-xdis/xdis/unmarshal.py:763 @20): t_object_reference -- 763 o = internObjects[refnum - 1] (trepan3k) refnum 3 (trepan3k) internObjects ['_TemplateMetaclass', 'super', '__init__', 'cls', 'name', 'bases', '__class__', (), 'simple_source/def/11_classbug.py'] (trepan3k) internObjects[refnum - 1] '__init__' (trepan3k) hex(fp.tell()-4) '0x120' ``` What's going on and how? Thanks! What's going on and how can I fix? Thanks _______________________________________________ pypy-issue mailing list pypy-issue@python.org https://mail.python.org/mailman/listinfo/pypy-issue