Sometimes, it seems the file analyzed can contain non-ASCII characters, giving errors such as:

codec can't decode byte 0xc3 in position 5138: ordinal not in range(128)

It seems there is no reason to not support UTF8. This change has been applied for a few weeks on Nexedi ERP5 production without problems.

/cc @tomo

Can we release a 0.128.2& if merging this? That way production will converge back to no fixup. I do not know if this applies to the Python 3 version not using backport LZMA, so we might also want to apply there.