)

* Add .worktrees to .gitignore

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Add load testing and benchmarking framework

Automated benchmark suite comparing java-http against JDK HttpServer,
Jetty, Netty, and Apache Tomcat. Includes wrk (C-based) and
fusionauth-load-tests (Java-based) as benchmark tools with structured
JSON output, README auto-updater, and GitHub Actions workflow.

Also adds HTTP/1.1 compliance smoke tests for RFC 9110/9112 coverage.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Improve update-readme.sh and drop invocationCount from new compliance tests

- update-readme.sh now prefers hello scenario, adds high-concurrency
  stress table, includes machine specs and reproducibility CLI
- Drop invocationCount=100 from 12 new compliance tests (default 1 is
  sufficient for protocol validation that closes the connection)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Add JDK HttpServer commentary and disable Tomcat compression

- update-readme.sh now adds a note about JDK HttpServer not being
  production-ready when high-concurrency results are present
- Disable compression in Tomcat server.xml for fair benchmarking
  (other servers don't enable compression)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Remove possessive FusionAuth phrasing from load-tests README

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Remove FusionAuth branding from load-test files

Position java-http as an independent open-source project.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Add elapsed timer to benchmark runner

Shows a live-updating elapsed time counter during each benchmark run
so you can see how long the current test has been running.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Add start time and total duration to benchmark runner

Banner shows start time, final line shows total elapsed minutes/seconds.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Show elapsed time per benchmark in results output

Each test result line now ends with the wall-clock time it took to run.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Add Duration label to per-test elapsed time

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Set ulimit at top of run-benchmarks.sh

Ensures wrk has sufficient file descriptors for high-concurrency
benchmarks. macOS defaults can be as low as 256, which causes
connection errors with -c1000.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Document ulimit usage in run-benchmarks.sh

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Stabilize vendor order in README performance tables

java-http always first, remaining servers sorted alphabetically.
Prevents row order from shifting between runs when RPS changes.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Fix blob repeat formula, misleading comments, and test inconsistencies

Fix incorrect string repeat formula that produced wrong-sized blobs for
the /file endpoint across all five server implementations. Fix misleading
InputStream comments, correct cross-reference pointers, and restore
invocationCount on case_insensitive_header_matching test.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Remove benchmark target from top-level build.savant

The run-benchmarks.sh script requires CLI args, so run it directly
from the load-tests directory instead.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Add machine model and OS version to benchmarks, add performance commentary

Capture machine model (e.g. MacBook Air) and OS version (e.g. macOS 15.7.3)
in benchmark output and README footer. Add performance commentary to the
README and update-readme.sh template explaining project positioning.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Remove redundant project goals paragraph, fix HTTP to HTTPS in Savant URL

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Daniel DeGroff <robotdan@Mac.localdomain>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

…y-handling

added better directory checking