[romtsn]

📜 Description

When debug = false, an SDK log call still allocates at the call site. The disabled logger then discards the result. This PR removes two sources of these allocations.

1. A varargs array on each call. javac makes a new Object[] for each varargs call, also for a call without arguments. ART does not remove it. ILogger now has default overloads for 0, 1 and 2 format arguments. DiagnosticLogger and NoOpLogger override them, and they check isEnabled before an array exists. The SDK call sites need no change: when they compile again, they use the new overloads. About 95% of the ~1,300 SDK log calls have 2 arguments or fewer.

2. Messages built before the call. 7 log calls on normal code paths built their message with string concatenation or toString(): 2 in DefaultCompositePerformanceCollector, called on each transaction stop, and 5 in SentryGestureListener, for example on each scroll. They now send %s arguments, so the SDK builds the message only when logging is enabled. Calls in catch blocks and at init are not changed.

The new ILogger methods are additions with default bodies, so existing ILogger implementations continue to work.

💡 Motivation and Context

Apps in production pay for SDK debug logs that they never see.

💚 How did you test it?

I used an androidx.benchmark microbenchmark on a Pixel 2 XL (API 30, debug = false, median). It was a temporary module and is not part of this PR. The logger is the real DiagnosticLogger that wraps AndroidLogger.

Benchmark	main	Overloads	Overloads + call sites
log(level, msg)	48.6 ns / 1 alloc	17.8 ns / 0	17.9 ns / 0
log(level, msg, a)	55.2 ns / 1	18.5 ns / 0	18.5 ns / 0
log(level, msg, a, b)	70.2 ns / 1	18.7 ns / 0	18.7 ns / 0
log(level, msg, long)	77.9 ns / 2	39.3 ns / 1	39.0 ns / 1
log(level, "msg " + s)	602.8 ns / 5	598.1 ns / 4	612.0 ns / 4
captureMessage	598 allocs	594	594
Transaction with one child span	416 allocs	412	408

• Each call: a disabled log call is about 3× faster and does not allocate.
• SDK operations: they make about 1–2% fewer allocations. Their time changes by less than the noise between runs.
• log(level, msg, long): the Long box stays, because the overloads take Object.

All unit tests pass (9,074 tests in 46 modules). Some tests verified the varargs overload with anyVararg(). Because the SDK calls now use the new overloads, those tests failed, or, in 4 never() checks, passed always. They now verify the overload that the SDK calls.

📝 Checklist

• I added GH Issue ID & Linear ID
• I added tests to verify the changes.
• No new PII added or SDK only sends newly added PII if sendDefaultPII is enabled.
• I updated the docs if needed.
• I updated the wizard if needed.
• Review from the native team if needed.
• No breaking change or entry added to the changelog.
• No breaking change for hybrid SDKs or communicated to hybrid SDKs.
• Public API changes reviewed by another Mobile SDK team member or implemented according to the develop docs spec.

🤖 Generated with Claude Code

[runningcode]

LGTM!

[sentry]

📲 Install Builds

Android

🔗 App Name	App ID	Version	Configuration
SDK Size	io.sentry.tests.size	8.64.0 (1)	release

⚙️ sentry-android Build Distribution Settings