Introduces EVM version Osaka

[r0qs]

Depends on #15829 Merged

[clonker]

I guess this should be marked experimental? (ping @nikola-matic)

[cameel]

It definitely should.

Also, one thing I noticed recently, that we might want to fix (though maybe in a separate PR) is that we do have an experimental flag in metadata, but we do not set it for experimental EVM versions (only for some other experimental features, including EOF). So bytecode compiled for Prague is not currently marked as experimental.

[nikola-matic]

It's marked as experimental in the other PR (the one that moves EOF tests to Osaka) - https://github.com/ethereum/solidity/pull/15828/files#diff-c907e10ad187a2840ede6859989664668693753e8c7d860fdc1b60275e39f053R185

Not sure why it was done there; I also have a sneaking suspicion that this PR will start failing as soon as this PR and the linked one above are rebased, since Kamil said that we've introduced semantic tests for EOF, whereas evmone 13 (latest and now merged) release did not move the EOF implementation to Osaka.

[nikola-matic]

Then again, this PR is from 2 days ago - so not sure when the semantic tests PR was merged.

[cameel]

On Monday: #15666.

[nikola-matic]

Why wouldn't they be able to run? They are not marked as requiring Prague specifically.

I'm talking about semantic tests that are explicitly marked as EOF and thus requiring Prague (not even sure if we have such tests); if we do have such tests, bumping them to require Osaka, when the libevmone in our test backend does not yet have EOF in Osaka, would make the tests fail, no?

[cameel]

bumping them to require Osaka, when the libevmone in our test backend does not yet have EOF in Osaka, would make the tests fail, no?

Ah, that's what you mean. No, you don't have to worry about it. You can safely assume that it does support EOF in Osaka as long as it allows selecting Osaka at all (and it does since evmone 12 uses evmc 12, which introduced an enum value for it). Osaka by default includes everything from Prague so they'd have to go out of their way to make it not support EOF.

The thing it does not have yet is to disallow EOF features in Prague. But that's not a problem for us.

I'm talking about semantic tests that are explicitly marked as EOF and thus requiring Prague (not even sure if we have such tests);

We do have a few - those added by #15665. They were added specifically in cases where some of the existing tests did not pass on EOF as is due to differences in behavior and we needed a separate EOF variant.

Though note that they don't even require bumping to Osaka. We have a separate setting to mark them as requiring EOF (bytecodeFormat: >=EOFv1) and we only run them on EVM versions that satisfy it.

Well, we do seem to still have a few tests with EVMVersion: >=prague in them, but that has always been redundant and should be removed.

[cameel]

In any case @clonker's comment is still relevant - the comments about it being experimental should be added already here, not in #15830.

[nikola-matic]

Yeah, I've pushed the requested changes already. Can you take another look, and if it's OK, I'll merge it later today (have to leave now).

[cameel]

BTW, I also added an issue to fix the missing experimental flag: #15835.

[cameel]

Just to things to reduce diff noise.