Fix miss-reported node reuse in types.

[dragomirtitian]

Node reuse is not being reported correctly. When using a cached node builder result the original node was being set as a synthetic node (the node that was in the cache) rather than an original source node.

Ex:

var v = (a: ) => {
>v : (a: any) => void                         //<- correct, no node reuse
>  : ^ ^^^^^^^^^^^^^^
>(a: ) => {   } : (a: any) => void
>               :                        //<-  second time around nodes are reused?

[typescript-bot]

This PR doesn't have any linked issues. Please open an issue that references this PR. From there we can discuss and prioritise.

[dragomirtitian]

cc: @weswigham

[weswigham]

If range already has a .original set, this is going to unset it if location is undefined or lacks a .original of its own, which, while is what fixes the issue you witness, doesn't seem desirable. Unsetting .original on a copy is probably not a good thing. If this issue is that when we clone a node from the cache, we .original it, we should update the underlining requirements in nodeIsFullySynthetic in typeWriter.ts to instead of just checking for a .original, specifically look for a getParseTreeNode instead. So,

function nodeIsFullySynthetic(node: ts.Node) {
    return ts.nodeIsSynthesized(node) && !ts.getParseTreeNode(node);
}

[dragomirtitian]

That also works. My thought was that the node you get should be the same regardless of if it comes from the cache or from the original operation.

[dragomirtitian]

Changed to using the parse tree node.

[weswigham]

I don't think we need the new nodeIsSynthesized check, either - all that's going to do is block setting .original on the range when the location has no position data, but it may still have a .original of its' own that is a parse tree node, and we wanna keep that 3 node chain (and is what getParseTreeNode looks up)

[weswigham]

🤦 I meant the one just added in this PR in checker.ts - though that one is probably redundant, too (since we generally set original node now), I'd rather keep it, since a node made with a pos/end set, but no .original set to a parse node is probably still based enough on an input node, since it has concrete positions (which have to have come from one or more input nodes).