Signextend and friends

[cburgdorf]

What was wrong?

There were a few cases where we needed to call adjust_numeric_size to "clean" a 256 bit stack value that is holding a numeric value <= 128 bits (See #524)

How was it fixed?

Changed move_expression to call adjust_numeric_size when loading integer values from memory or storage.

[cburgdorf]

@g-r-a-n-t I can't really tell if that is all that is needed in this regard. I know that Solidity performs way more of these cleanup operations at a bunch of places (e.g. inside checked_add_*, checked_sub_* etc) but I can't convince myself that they are actually needed there and I don't just want to put in checks (that cost gas) without having tests that prove we need them.

[g-r-a-n-t]

I know that Solidity performs way more of these cleanup operations

I've noticed that too. It could be precautionary, but I'm not sure.

The PR looks good, but I'll spend some more time reviewing and approve it when I feel comfortable with it.

[cburgdorf]

In this issue it is noted that they could perform less checks and it also sounds that some of these checks are just in place to avoid users messing things up with inline assembly which we don't have.

but I'll spend some more time reviewing

Thanks, I'll sleep better if you check this over thoroughly. 😅

[cburgdorf]

Btw, here's a blog post on a recently found Solidity bug related to these cleanups. Haven't read it yet but felt like leaving it here for later: https://blog.soliditylang.org/2021/09/29/signed-immutables-bug/

[g-r-a-n-t]

LGTM