UnicodeDecodeError when adding easyconfig to archive (if archspec is available)

[boegel]

I ran into this crash when installing MAFFT-7.453-iccifort-2019.5.281-with-extensions.eb (when using eb on top of Python 2):

ERROR: Traceback (most recent call last):
  File "/lib/python2.7/site-packages/easybuild_framework-4.2.0.dev0-py2.7.egg/easybuild/main.py", line 114, in build_and_install_software
    (ec_res['success'], app_log, err) = build_and_install_one(ec, init_env)
  File "/lib/python2.7/site-packages/easybuild_framework-4.2.0.dev0-py2.7.egg/easybuild/framework/easyblock.py", line 3239, in build_and_install_one
    repo.add_easyconfig(spec, app.name, det_full_ec_version(app.cfg), buildstats, currentbuildstats)
  File "/lib/python2.7/site-packages/easybuild_framework-4.2.0.dev0-py2.7.egg/easybuild/tools/repository/filerepo.py", line 120, in add_easyconfig
    txt += statstxt
UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 668: ordinal not in range(128)

The culprit is that archspec returns a unicode-type value as CPU arch name, which gets embedded in statstxt, making it also a unicode-type value (rather than a regular str).

When statstxt is added to txt (which is just a str value), it triggers a decoding using the ascii of the txt value, which fails because the description value of MAFFT-7.453-iccifort-2019.5.281-with-extensions.eb contains non-ASCII characters (the ~ in "for alignment of <∼200 sequences"`).

It's easy to fix this, by making sure we cast the unicode value produced by archspec to a regular str value, but it does raise the question why archspec uses unicode values at all, there should be no need for that...