Cherry-pick with test fixes

[nanjekyejoannah]

Cherry-pick with fixes, to make string warning ready.

[ltratt]

A few blank line additions have crept into this commit, which might make merge conflicts a bit more likely later. Probably worth putting things back as they were.

[ltratt]

Please squash.

[ltratt]

Please squash.

[nanjekyejoannah]

Squashed

[ltratt]

Ah, this has failed because CI is set on the migration branch (not regression_fix). Which of those branches would you like to target?

[nanjekyejoannah]

The PR targets regression_fix

[ltratt]

Does that mean we don't want the migration branch?

[nanjekyejoannah]

Does that mean we don't want the migration branch?

I will explain more tommorrow.

This is going to be merged to migration eventually, that was we discussed when we created this branch IIRC.

[ltratt]

OK.

[ltratt]

Is this one OK to merge?

[ltratt]

@nanjekyejoannah Shall we merge this?

[ltratt]

Hmm, this took over an hour in our CI system. I'll try again in case it was a one-off, but I'm wondering if something's making things take unduly long.

[ltratt]

@nanjekyejoannah In CI a Python process is currently taking 37GiB of RAM, so I think something's gone wrong with this PR. I wonder if it's because of this:

if not sys.py3kwarning:
    sys.setrecursionlimit(1 << 30)

at https://github.com/softdevteam/pygrate2/pull/36/files#diff-1d0c2ffbcb01254c4ab02053d85527be50e03ea2908dc6cdba9fc185b8df9b7bR10 ?