Fix uncaught exception when using empty string for yul optimization flags

[matheusaaguiar]

fix #14946.

[nikola-matic]

Bugfixes need changlog entries :)

[matheusaaguiar]

I am wondering whether we should have a better error message for this case. Although the docs explain what should be an empty optimizer sequence, it can be a bit confusing. Or in other words, it is understandable mistake to think that an empty or blank string is the same as the empty sequence.

[nikola-matic]

Can you add CLI tests as well?

[nikola-matic]

Also, please add a short line to describe what an empty optimizer sequence is (:) here.

[nikola-matic]

Suggested change

"--strict-assembly",

[cameel]

It would actually not hurt to test both as long as it can be done concisely, e.g. by looping over

{
    ["--strict-assembly", "--yul-optimizations="],
    ["--strict-assembly", "--yul-optimizations=    "],
    ["--yul-optimizations="],
    ["--yul-optimizations=    "],
}

which would be used to construct the command. Both assembling and compilation go through the same path in command-line parser now, but it's not guaranteed that it will always be like this and the issue affects both.

I'd say it's optional, but @nikola-matic has a point that currently you're assembling with input file called input.sol.

[matheusaaguiar]

Actually, this should had been input.yul. I was testing using the example from the issue.
I will add the variations.

[cameel]

It would actually not hurt to test both as long as it can be done concisely, e.g. by looping over

{
    ["--strict-assembly", "--yul-optimizations="],
    ["--strict-assembly", "--yul-optimizations=    "],
    ["--yul-optimizations="],
    ["--yul-optimizations=    "],
}

which would be used to construct the command. Both assembling and compilation go through the same path in command-line parser now, but it's not guaranteed that it will always be like this and the issue affects both.

I'd say it's optional, but @nikola-matic has a point that currently you're assembling with input file called input.sol.

[matheusaaguiar]

I noticed here that the metadata was not preserved. The optimizerSteps value is just the colon instead of a whitespace before it and two whitespaces after it.
The cause is

solidity/libsolidity/interface/CompilerStack.cpp

Line 1732 in ae9bcab

details["yulDetails"]["optimizerSteps"] = ":";

being defined if the sequence is the empty optimizer sequence, although we accept whitespaces and linebreaks.

[matheusaaguiar]

I fixed that in a455829.

[cameel]

True. But changing it so that this is preserved also technically changes the metadata so I'd keep it out of this PR.

Like I said in #14967 (comment), this is fine to do, but I'd rather do any such changes explicitly, in a separate PR.

Also, I think that the right way to make it consistent is to be always stripping whitespace from the sequence, means we'd leave this bit as is and instead change the rest to match :)

[matheusaaguiar]

Ok, I will separate this commit apart.

[cameel]

The only thing that still needs to be addressed is #14967 (comment). Other than that there are no big problems and I pushed some fixups to clean up some of the smaller ones without going through extra review rounds.

[matheusaaguiar]

Rebased and squashed commits.

[nikola-matic]

Also, please add a short line to describe what an empty optimizer sequence is (:) here.

Push docs and I'll re-approve so we can merge.

[matheusaaguiar]

Also, please add a short line to describe what an empty optimizer sequence is (:) here.

Push docs and I'll re-approve so we can merge.

Sorry, I missed that when checking pending comments. Fixed.