[gmpassos]

Problem

To find the target class of a ClassProxy, InputAnalyzerResolved.findCandidateClassElements listed every library known to the resolver (thousands: the SDK and all the packages) and collected all their classes, then searched by name. It did this again for each ClassProxy.

In a project with 69 ClassProxys in one library, that library took 4.5s of the builder's ~7s of "building". About 3.7s of it went to collecting classes (~53ms per proxy, over ~2,400 libraries).

Fix

• The resolver's classes are indexed by name once per build step (InputAnalyzerResolved), each library when it's first seen. The resolver gets more libraries as more are resolved (630 → 2,410 after the first proxy here), so each call still reads the resolver's current library list, at the same point as before, and indexes only the new ones.
• The candidates are the same as before: the class sets are per library, so their union over the libraries is unchanged. The generated code is byte-for-byte identical (checked on the 75 generated files of that project).

Measured

Clean build_runner build of that project, local, on AC power, 2 runs each:

	2.12.1	this PR
reflection_factory "building"	7s	4s
"analyzing"	4s	4s
Built with build_runner/aot in	23–24s	20s

What's left, for the record:

• "analyzing" (~4s): the analyzer's first resolution of the program, shared by all inputs.
• Compiling builders (~11s locally): build_runner's own cost; a trivial builder takes the same time.
• Formatting (~1.5s) and the 225 @EnableReflection classes (~0.8s).

Also

• CHANGELOG 2.12.1: don't name a private project.
• Test: many ClassProxys in one library, the second with a libraryPath that brings in libraries the first one didn't index.
• VM 565 and Chrome 439 tests pass; dart analyze --fatal-infos clean.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.55%. Comparing base (5e99421) to head (8023182).

Additional details and impacted files

@@           Coverage Diff           @@
##           master      #60   +/-   ##
=======================================
  Coverage   93.54%   93.55%           
=======================================
  Files          14       14           
  Lines        6538     6544    +6     
=======================================
+ Hits         6116     6122    +6     
  Misses        422      422           

Flag	Coverage Δ
unittests	93.55% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.