fix: stop polling the debug port once the browser has answered

[nathanfallet]

The bug

The wait loop in DefaultBrowser.start() reads as "try until it connects, up to maxTries times":

delay(config.browserConnectionTimeout)
repeat(config.browserConnectionMaxTries) {
    if (testConnection()) return@repeat
    delay(config.browserConnectionTimeout)
}

return@repeat returns from the lambda, not from the loop — it is a continue, not a break. The loop never stopped early. And because the line it skipped is the delay, a browser that answered on the first try still got every remaining attempt fired at it back to back, with no wait in between.

This is the successful path, not an error path. Every browser start did it.

Standalone reproduction of the loop, run as-is:

testConnection() appelé : 60 fois
delay() exécuté        : 0 fois

So with the defaults, each successful start sends 60 GET /json/version in a burst at a Chrome that has just come up.

The fix

The polling rule moves to awaitConnection(), so it can be tested without a real browser, and uses a genuine early return.

start() also now logs "Connection to browser established" only when the connection actually was established. It used to log it unconditionally — one line above the error saying the browser never opened its port, which is confusing when reading logs of a failed start.

Not changed

The timeouts themselves. 60 tries × 500 ms is still 30 s of budget. It is worth revisiting (zendriver waits 2.5 s, pydoll 10 s), but that is a behaviour change and this is a bug fix — happy to do it in a follow-up.

Tests

AwaitConnectionTest, written red first against a copy of the old loop:

test	old loop	new
answers immediately → stops polling	❌ 60 attempts, returns false	✅ 1 attempt
answers on the 3rd try → stops there	❌	✅ 3 attempts
never answers → exactly maxTries attempts	✅	✅
waits one interval per failed attempt	✅	✅

The last two passed against the buggy version too — they pin the behaviour that must not change, so the fix cannot trade one bug for another.

Locally: all five targets compile, :core:jvmTest green, detekt findings identical to main (317, same set).

[codecov]

Codecov Report

❌ Patch coverage is 85.71429% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
.../kotlin/dev/kdriver/core/browser/DefaultBrowser.kt	66.66%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!