suggest new catch/finally parents

[eric-therond]

No description provided.

[ircmaxell]

Looks reasonable. The only thing missing I think is discussion around ensuring finally has a link to the parent or not...

[ircmaxell]

I think this is still needed, isn't it? For the case where you have nested finally, it would propagate the parent's catch context:

try {
    try {
        //...
    } finally {
        return true;
    }
} finally {
    return false;
}

[eric-therond]

Can we try propagating the parent's catch context while putting the try body block as the parent of the finally block?

Actually, the problem is more serious in the case of nested try: we must have all possible nested catch contexts within a block? This means that a block must contain an array of catch contexts, not just the closest one in the flow, as it's the case today.

[ircmaxell]

That shouldn’t be a problem. Since we always propagate a finally block (whether try has one or not), and that finally block contains the catch context of the parent (with the original implementation), we can reconstruct based upon what’s in the propagated finally. So basically hit finally block, if the exception is still valid at the end, find the next in the parent catch context for the finally block (inherited from the parent).

this is preferable to merging catch blocks, as it respects finally by design.

[eric-therond]

Okay, that's clear to me.
I've updated the PR.

[ircmaxell]

Yeah, this is a much better approach...