For easier root cause analysis when an invalid request is passed.

PiperOrigin-RevId: 613110312

I think this fixes crashes from the presubmit for unknown commit:

```
Error: Exception in thread "main" java.lang.NoSuchMethodError: 'java.lang.String org.eclipse.equinox.p2.core.IProvisioningAgent.getProperty(java.lang.String)'
	at org.eclipse.equinox.internal.p2.artifact.repository.simple.SimpleArtifactRepository.getAgentPropertyWithFallback(SimpleArtifactRepository.java:813)
	at org.eclipse.equinox.internal.p2.artifact.repository.simple.SimpleArtifactRepository.getMaximumThreads(SimpleArtifactRepository.java:1012)
```

I tested a PR with these changes and it was clean [1], so either this helps or
it was a transient issue, but either way this shouldn't hurt.

[1] https://github.com/google/google-java-format/actions/runs/8329274139

PiperOrigin-RevId: 616846344

PiperOrigin-RevId: 616908365

PiperOrigin-RevId: 620100015

Float was used so that forced breaks could be modeled with POSITIVE_INFINISTY, but an arbitrary large int works just as well.

PiperOrigin-RevId: 620395732

Cursory perf measurements show no detectable cost from scanning every token. If one should be discovered later, we can add guards to only scan potentially multiline tokens.

PiperOrigin-RevId: 620833035

… commit

Previously the formatting was adding a forced break at the beginning of text blocks, which caused issues like #1081. With the changes in the baseline CL it correctly handles the 'width' of text blocks containing newlines as infinity, instead of counting the number of characters and treating the newline as having width 1.

Fixes #1081

PiperOrigin-RevId: 620933964