This is made more obvious by #41877, which corrects the spans of tags to their full spans, making the error spans even more ungainly. Here's one example; see the errors.txt diffs in the linked PR for more.

Expected behavior:

    /** @type {Gioconda} */
        ~~~~~~~~~~~~~~~~ // before #41877
        ~~~~~~~~~~~~~~~~~ // after #41877 (both are wrong)
!!! error TS8030: The type

I'm Writing a ThemeColor Plugin, use Visitor.

In fn:visitColor , node.parent sometimes is null (box-shadow:0 0 0 1px #999 inset). I can't find the parent Declaration Node of Color Node.
so I want to [less.js/lib/less/visitors/visitor.js Visitor.prototype.visit](https://github.com/less/less.js/blob/0e268598c8327b9e8d36d9bb2edfde95d8d787be/packages/less/src/less/visitors/visitor.js

### Example
let unreachableProxy = "http://127.0.0.1:8888"
try:
  let client = newHttpClient(proxy=newProxy(unreachableProxy), timeout=1)
  var res = client.getContent("https://github.com")
  echo res
except:
  echo "Timed out"

Current Output

timeout variable is ignored, program stuck until default socket timeout will happen

Expected Output

Timed out a

We already have a specialized checked exponentiation when the base is a literal and exponent can only be known at compile time.

Implement the case where exponent is a literal and base is only known at compile time. Vyper already has this: vyperlang/vyper#2072

This is a follow-up to #4097 where we used an AtomicInteger as a poor-mans replacement for a Semaphore with tryAcquire (or maybe even drainPermits).

Should suppress --strict warnings.