Python3 build error: unicode problem in pip_installed_packages()

[mderickx]

The python 3 build was broken by #23615 because of incorrect unicode usage causing the building of sage failing in the following way

chapoton@icj-laptop:~/sage3$ ./sage -br
cd . && export                                    \
    SAGE_ROOT=/doesnotexist                               \
    SAGE_SRC=/doesnotexist                                \
    SAGE_SRC_ROOT=/doesnotexist                           \
    SAGE_DOC_SRC=/doesnotexist                            \
    SAGE_BUILD_DIR=/doesnotexist                          \
    SAGE_PKGS=/home/chapoton/sage3/build/pkgs                \
    SAGE_CYTHONIZED=/home/chapoton/sage3/src/build/cythonized      \
&& sage-python23 -u setup.py --no-user-cfg build install
************************************************************************
Traceback (most recent call last):
  File "setup.py", line 69, in <module>
    from module_list import ext_modules, library_order, aliases
  File "/home/chapoton/sage3/src/module_list.py", line 166, in <module>
    from sage_setup.optional_extension import OptionalExtension
  File "/home/chapoton/sage3/src/sage_setup/optional_extension.py", line 24, in <module>
    all_packages = list_packages(local=True)
  File "/home/chapoton/sage3/src/sage/misc/package.py", line 226, in list_packages
    installed = installed_packages(exclude_pip)
  File "/home/chapoton/sage3/src/sage/misc/package.py", line 286, in installed_packages
    installed.update(pip_installed_packages())
  File "/home/chapoton/sage3/src/sage/misc/package.py", line 148, in pip_installed_packages
    return {package['name'].lower():package['version'] for package in json.loads(stdout)}
  File "/home/chapoton/sage3/local/lib/python3.6/json/__init__.py", line 354, in loads
    return _default_decoder.decode(s)
  File "/home/chapoton/sage3/local/lib/python3.6/json/decoder.py", line 339, in decode
    obj, end = self.raw_decode(s, idx=_w(s, 0).end())
  File "/home/chapoton/sage3/local/lib/python3.6/json/decoder.py", line 357, in raw_decode
    raise JSONDecodeError("Expecting value", s, err.value) from None
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)
************************************************************************
Error building the Sage library
************************************************************************
Makefile:34 : la recette pour la cible « sage » a échouée
make: *** [sage] Erreur 1

[mderickx]

New commits:

​f3e5282

Trac #23822: fix python3 build error: unicode problem in pip_installed_packages()

[chapoton]

ok, let it be

[jhpalmieri]

Do you need the .decode, or can you just use

stdout = proc.communicate()[0]

?

[mderickx]

In python 2 it is not needed but good praxis. In python 3 it is needed since stdout is a bytestring before decoding and the json library only takes unicode strings and not bytestrings as input.

[jhpalmieri]

When I was looking at #23876, I came across this bug and didn't use .decode, and it seemed to work with Python 3. That is, before making any changes, Sage crashed on startup, complaining about this problem, but when I switched to stdout = proc.communicate()[0], it didn't complain about that. (It crashed elsewhere, as described at #23876.) So I am not convinced that it is needed with Python 3.

Okay, I checked the Python documentation. I agree that stdout should be a binary stream, but according to the ​json documentation, json.loads can take something of type bytes as input.

Edit: this behavior of json is new as of Python 3.6.

[mderickx]

I could have sworn that I only added the decode() because it was failing otherwise. But sage was already on 3.6 when I wrote this. I dont think either solution is clearly better then the other. But the one with a decode already has positive review so I propose to leave it as is.