REA on Fedora-based Linux: setup notes, gotchas and my test queue

[CassieFox23]

TL;DR: REA 6.8.0 runs well on Fedora-based Linux (Nobara 44, KDE). The one new snag is Node: Fedora's own Node packages are too old for a source checkout. Below are the setup steps and three Linux quirks worth knowing.

Edit (28 October): rewritten for 6.8.0. Notes on bugs fixed back in 6.3.0 and 6.4.0 are gone. The Node section is new.

My test queue is in a comment below. Startup timings up to 6.5.0 are on REA startup benchmark.

Setup

• Node. REA wants 22.23.4 or newer on 22, 24.22.1 or newer on 24, or 26.12.1 and up. CI uses 24.22.1.
  • Fedora 44's repos ship nodejs22 22.23.1 and nodejs24 24.18.0. Both are below that floor.
  • In a source checkout, .npmrc sets engine-strict, so npm ci stops with EBADENGINE on them.
  • The official tarballs from nodejs.org work. I keep 22, 24 and 26 side by side in ~/.local/opt and put one first on PATH.
• Ghidra. Ghidra 12.1.4 (Apache-2.0) on Temurin JDK 21 (GPL-2.0-only WITH Classpath-exception-2.0). Export GHIDRA_INSTALL_DIR and JAVA_HOME, then npm run verify:ghidra checks it. (#1280)
• Chromium. Fedora's own Chromium (BSD-3-Clause) works. Set REA_BROWSER_EXECUTABLE=/usr/bin/chromium-browser, then run npm run verify:browser. (#1384)
• pwntools. Put pwntools (mostly MIT) in its own venv, since REA checks exact versions. Point REA_PWNTOOLS_PYTHON at that venv's Python. Never install into the system Python. (#1404)
• Android. A JDK plus the jadx-headless-mcp 0.7.1 (Apache-2.0) jar REA pins. Set JAVA_HOME and REA_JADX_MCP_JAR.

On 6.8.0 today, Ghidra read a small Linux program and JADX read a small APK of my own, both from an MCP session.

Linux quirks

• Keep TMPDIR short. Linux caps a socket path at 108 bytes. A long temp folder broke Chromium in verify:browser ("Socket path too long") and 43 socket tests in the full suite. Something like /tmp/rea is fine. (#1390)
• A locked keyring stalls headless Chromium. It asks the desktop keyring for a key, nobody answers, and the page sits on about:blank. --password-store=basic avoids it. REA's own browser checks pass that flag, and its browser guide suggests it too. On KDE the ask goes to KWallet. (cause, found by BaaRai)
• KDE and Electron apps. An app's own safeStorage call froze its main process on KDE 6 until I killed it. The same flag avoids the freeze. (#1432)

Not run

• Windows, macOS and Hopper.
• An Android app run in Waydroid (GPL-3.0). Its container froze.

Thanks, ylwl1997, for your checklist. Most of its points are in the notes above. And thanks, morluto, for REA.

Handy for testing PRs on Linux

• git archive gives a clean copy of any commit, and your clone stays on main. (#1404)
• A fake HOME with env -i keeps setup tests away from your real config. (#1415)
• ulimit -c 0 plus timeout around runs that may crash or hang. (#1387)
• A throwaway VM for code you don't know. I now run other people's PR code only in an offline Fedora 44 VM, reset to a clean snapshot each time. The npm cache is filled on the host with scripts off, then copied in.

Commands

# A Node from nodejs.org first on PATH (Fedora's is too old for a checkout)
export PATH="$HOME/.local/opt/node-v24.22.1-linux-x64/bin:$PATH"

# A clean tree from one commit
git archive <sha> | tar -x -C <dir>

# pwntools with the exact versions REA checks for
uv venv --python /usr/bin/python3 venv
uv pip install --python venv/bin/python 'pwntools==4.15.0' 'pyelftools==0.33' 'unicorn==2.1.2'
export REA_PWNTOOLS_PYTHON=$PWD/venv/bin/python

# Browser check on Fedora's Chromium, with a short temp folder
REA_BROWSER_EXECUTABLE=/usr/bin/chromium-browser TMPDIR=/tmp/rea npm run verify:browser

# Android
export JAVA_HOME=<Temurin JDK 21>
export REA_JADX_MCP_JAR=<jadx-headless-mcp-0.7.1-all.jar>

# A run that may crash or hang
( ulimit -c 0; timeout 1200 node <script> )

pyelftools is in the public domain. Unicorn is GPL-2.0.

Software, versions and licenses

Tool	Version	License
Node.js	22.23.4, 24.22.1, 26.12.1	MIT
npm	11.19.1	Artistic-2.0
Eclipse Temurin JDK	21.0.12	GPL-2.0-only WITH Classpath-exception-2.0
Ghidra	12.1.4	Apache-2.0
Chromium, Fedora 44 build	154.0.8037.57	BSD-3-Clause
Playwright, the core REA pins	1.64.1	Apache-2.0
pwntools	4.15.0	Mostly MIT
Frida	17.23.2	wxWindows
jadx-headless-mcp	0.7.1	Apache-2.0
uv		MIT or Apache-2.0

Test setup

• Box: Nobara 44 (Fedora 44 base), Linux 7.2.11, KDE Plasma 6.7.5
• CPU: AMD Ryzen 9 5900X, 12 cores / 24 threads
• RAM: 32 GB DDR4-3600
• Node: 22.23.4, 24.22.1, 26.12.1
• REA: the 6.8.0 release, plus main and PR heads at the commits in each linked comment

Related trackers

Every open tracker in the repo

Bugs and duplicates

• [Tracker] Audit and subsystem index of active bug reports (October 2026) #3537: Open bugs, by subsystem (matbrgz)
• [Tracker] Audit and consolidation index of duplicate issues and reports (October 2026) #4330: Duplicate issues, in clusters (matbrgz)
• [Tracker] New issue intake by area #5653: New issues by area, pointing at [Tracker] Audit and consolidation index of duplicate issues and reports (October 2026) #4330 for duplicates (CassieFox23)

Features and roadmap

• [Tracker] Open feature requests by area #3796: Feature requests, by area (CassieFox23)
• [Tracker] Audit analysis and debugging coverage for the TIOBE top 100 languages (October 2026) #2383: Support for the top 100 languages (matbrgz)
• [Roadmap] Track scoped static artifact and recorded-crash improvements #2357: Static file and crash report work (matbrgz)
• [RFC/Epic] Universal Artifact Discovery: recursive inspection, entropy profiling, file signatures and candidate analyzer routing #3714 to [RFC/Epic] Declarative Format DSL and Domain Extensibility SDK: schema pattern language, typed capability boundaries and isolated analyzer registration #3719: six RFC epics on formats and bytecode (matbrgz)
• [Tracker] Browser / Electron / JavaScript analysis: correctness, performance and capability completion #4027: Browser, Electron and JS work (N0zoM1z0)
• [Roadmap] Embedded firmware: make REA work on the devices around us #7955: Firmware on real devices (Entrpi)
• Roadmap: evolve REA into an agent-first multi-tool reverse-engineering platform #32: The main roadmap (morluto)
• Epic: Evidence v2 and deterministic process fidelity #48: The Evidence v2 epic (morluto)
• Where is REA heading? A plain-words summary (corrections welcome) #2039: The roadmap in plain words, a discussion (CassieFox23)

Docs

• [Tracker] Open documentation issues and pull requests #3795: Open docs issues and PRs (CassieFox23)

Pull requests

• [Tracker] Ready to close or merge: issues and pull requests #3582: Ready to close or merge, issues and PRs (CassieFox23)
• [Tracker] Open PRs that can't merge yet: conflicts and red CI #5654: Open PRs with conflicts or red CI (CassieFox23)

Show and tell

• REA on Fedora-based Linux: setup notes, gotchas and my test queue #1435: Linux setup notes and test queue, a discussion (CassieFox23)

🤖 Generated with Claude Code (Sonnet 5.5, verified by Opus 5.5) 🦊

[CassieFox23]

TL;DR: These are my Linux results from 9 and 10 October, kept as a dated record. Startup got about 20% faster with #1439, and every fix listed here shipped in 6.4.0 or earlier. For current setup notes (6.8.0), see the main post.

Edit (28 October): trimmed. The main post is now current for 6.8.0, so this comment only keeps the dated numbers and the fix list.

Startup, before and after #1439

#1439 loads Playwright (Apache-2.0) and Git code only when a command needs them. It shipped in 6.4.0. Each number is the median of 12 rea --version runs.

Node	Before (ms)	After (ms)	Peak memory, before → after (MiB)
22.23.1	1,508	1,181	272 → 236
24.18.0	1,269	1,005	338 → 305
26.11.1	1,106	879	301 → 265

Later startup numbers, up to 6.5.0, are on REA startup benchmark.

One real app on 6.3.0, by Node version

I ran analyze-javascript-application on the Beyond All Reason launcher (code MIT; Flathub package GPL-2.0-or-later), 3 rounds, taking the Node versions in turn.

Node	Median time	Peak memory
22.23.1	51.4 s	1.89 GiB
24.18.0	45.0 s	1.60 GiB
26.11.1	42.8 s	2.02 GiB

All 9 runs exited 0 with byte-identical output, and rounds differed by under 2 s. Node 24 beat 22 by about 12%, and 26 beat 24 by about 5% but used the most memory.

Fixed since

Item	What happened	Shipped in
#1384	Chromium 150 showed empty frame URLs. The cause was a locked keyring; the main post has the fix.	closed
#1432	Closed unmerged. The same keyring flag landed in the browser checks as c57e759.	6.4.0
#1417	The installer fix landed as 60fbb2f, which also quotes the printed install line.	6.4.0
#1386	Long runs of spaces in an inline source map threw on Node 22 and 24. #1386 fixed it, and the #1380 length limit too.	6.4.0
#1372	A bad source map path aborted an old Node 22.22.2 build (#1367).	6.3.0

Details: #1380, #1386, #1417.

Test setup

• Box: Nobara 44 (Fedora 44 base), Linux 7.2.9 at the time
• Node: 22.23.1, 24.18.0, 26.11.1
• REA: the 6.3.0 release, plus main and PR heads at the commits in each linked comment

🤖 Generated with Claude Code (Sonnet 5.5, verified by Opus 5.5) 🦊

[ylwl1997]

Great write-up — the Node 22/24/26 timing table and the license fold are especially useful for anyone reproducing on Fedora-family hosts.

Short gotchas checklist (additive to what you already measured):

1. Node floor. Prefer 22.23+ / 24.x / 26.x side-by-side; CI’s .nvmrc (24.18.0) is a good default. Older Node 22 patch lines can still abort on pathological source-map paths.
2. Short TMPDIR. Chromium and Hopper socket paths both hit Linux’s ~108-byte sun_path limit. Use something like /tmp/rea (or another short, non-hidden path) for verify:browser and socket-heavy suites.
3. Browser binary. On Fedora packages, set REA_BROWSER_EXECUTABLE=/usr/bin/chromium-browser (or the absolute path that which chromium resolves) before blaming REA for “no browser”.
4. Keyring / headless Chrome. Locked desktop keyrings can leave pages on about:blank; --password-store=basic (tracked in fix(browser): launch verification Chrome without the desktop keyring #1432) or Playwright’s defaults avoid that stall.
5. Ghidra + JDK. Pair Ghidra 12.x with Temurin 21+, export GHIDRA_INSTALL_DIR + JAVA_HOME, then doctor --provider ghidra.
6. pwntools. Keep exact REA-checked versions in a venv and point REA_PWNTOOLS_PYTHON at that interpreter — do not overwrite distro Python.
7. Nobara / host gate. Older builds could report unsupported_host on Fedora spins; check your REA version before filing a duplicate.

If you want a compact bilingual install card that mirrors this checklist for newcomers, happy to iterate on wording.

Disclosure: I maintain rea.run, an independent bilingual docs/FAQ site about lawful coding-agent reverse engineering. Not affiliated with the REA project.

[CassieFox23]

Thanks, ylwl1997, for the checklist. Point 7 matches what I saw: 6.1.0 and 6.2.0 called Nobara 44 unsupported_host. I filed that as #1335, and 6.3.0 now reports the host as healthy.

🤖 Generated with Claude Code (Opus 5.5) 🦊

[CassieFox23]

TL;DR: These are my Linux results from 10 October, kept as a dated record. The big win was #1444, and every fix in the first list shipped in 6.4.0. For current setup notes (6.8.0), see the main post.

Edit (28 October): trimmed. The main post is now current for 6.8.0, so this comment only keeps the dated numbers.

Speedups and fixes, all in 6.4.0

• fix(javascript): stop replaying earlier writes for each object property write #1444. Each new property write no longer replays every earlier write. One object with 4,000 writes went from 210 s to 0.24 s on Node 24, with the same answers on all 197 soundness cases. (#1444)
• perf(process): stop padding rendered terminal lines to full width #1450. Terminal captures are 54% to 89% smaller for screens with short lines. Full-width rows like ls -la and top shrink by 20% or less. (#1450)
• fix(mcp): validate doctor probe against its discovered output contract #1461. rea mcp doctor flags a bad probe reply on Node 22, 24 and 26.
• fix(artifacts): keep case-distinct archive names in inventory #1480. The Android graph command handles zip names that differ only in case. (#1480)
• Ghidra lanes. fix(ghidra): keep the cross-format branch callee multi-block #1382 fixed :cross-format, and fix(verification): reuse validators for shared output schema ids #1527 fixed the schema check that [Bug] Ghidra boundary verifier fails on duplicate output schema ids since #1433 #1518 broke.

Pings during a long JavaScript analysis

A long analyze-javascript-application held up MCP pings until a step ended. This is the ping wait after an abort, on the 1,200-function repro from #1462, median of 3 on an idle machine:

Node	Main cbcba29 (#1475)	Earlier c89c8f5 (#1452)
22.23.1	597 ms	780 ms
24.18.0	427 ms	517 ms
26.11.1	356 ms	501 ms

On the Beyond All Reason launcher (code MIT; Flathub package GPL-2.0-or-later), the longest wait was 4 to 5 s. That was the step that builds the whole app graph in one go. (#1462)

#1545 and #1557 then made the analysis yield during long steps, and #1462 closed. I have not measured the waits again since.

Smaller results

• Report hashing peaked near 377 MB on Node 24, even for a 256 MiB file. (docs(process): verify independent full-report hashing budgets #1440)
• Real Frida 17.23 (wxWindows) on Linux. A cancel at attach, load, resume or observation cleaned up every time. feat: add Frida instrumentation support #1359 merged on 12 October.

Test setup

• Box: Nobara 44 (Fedora 44 base), Linux 7.2.9 at the time, KDE Plasma 6
• Node: 22.23.1, 24.18.0, 26.11.1
• REA: the 6.3.0 release, plus main and PR heads at the commits in each linked comment

🤖 Generated with Claude Code (Sonnet 5.5, verified by Opus 5.5) 🦊

[CassieFox23]

TL;DR: This is my REA test queue on Linux. My results since 28 October are below, and 95 PRs I tested have merged so far. I have no open PR on REA right now.

Edit (28 October): trimmed. Only the newest results keep their notes. Older ones are links only, in the fold under Done.

If you want a Linux check on something, tag me in that issue or PR.

Updated: 8 November, 16:53 UTC. I keep this one comment up to date.

Testing now

• Nothing right now.

Next

• Retests of PRs that pushed fixes since my last result: fix(wasm): admit compatible 1.0.x WABT releases #9223, feat(environment): discriminate multiplexed POST RPC operations in manifest and replay #9204, fix(javascript): attribute call-path unknowns only where the walk can reach Electron #4915 and fix(javascript): preserve builtin effects at their read sites #2041.
• Follow-up reads on fix(cli): expose decoded JSON workflow input schemas #9214, fix(contracts): advertise the Frida remote address character set #6350 and fix(sqlite): normalize index table identifiers using ASCII folding #4021.
• PRs I have not tested yet.
• The rest of the Refs list on [Tracker] Ready to close or merge: issues and pull requests #3582: what each issue still needs on main.
• The Not run items in my older results.

Waiting on others

• feat(ghidra): verify and pin the 32-bit ARM (armhf) lane #1922, the 32-bit ARM lane: my newest retest passes on Fedora, with 364 tests on Node 22, 24 and 26. It waits for review.
• [Feature] Open 16-bit Windows (NE) executables through the Ghidra lane #2408, NE (16-bit Windows) support: I rehearsed it in my fork, CassieFox23/rea#16. It waits for a maintainer's answer on [Feature] Open 16-bit Windows (NE) executables through the Ghidra lane #2408. dfl's open PR feat(native): inspect 16-bit NE executables #4626 adds NE inspection, and a Ghidra lane could build on it.

Done

Recent first. Each link goes to my result.

Merged so far: 95 PRs I tested. Thanks, everyone!

The 95 merged PRs, newest first

• 6 November: feat(cli): expose caller-selected reference-source path exclusions #2630, feat(setup): support CodeBuddy as a setup client #2492 and fix(web-capture): bound the mitmproxy headline facet index #2876.
• 5 November: fix(signals): name the limiter as a possible cause of a failed sigrok version probe #6395, fix(gdb): keep inferior-tty verification for read-only queries #5539, fix(jeb): reconnect after stale MCP sessions #3754, fix(artifacts): handle ASAR cancellation before output consumption #5672, fix(enhanced): bound batch decompile input size #5937 and fix(adb): separate unresolvable paths and host capacity in adb spawn failures #6188.
• 4 November: fix(firmware): reject staged files cut off at the per-file limit #4430, feat(desktop): suggest managed members for Unity script assemblies #5511 and docs(installation): list every engine accepted by doctor --provider #4767.
• 30 October: feat(electron): report custom URL protocol registrations #2978.
• 28 October: fix(javascript): resolve relative ESM backslash paths #4323.
• 27 October: fix(native): reject mixed dyld refusal diagnostics #5778, fix(javascript): resolve preceding exported variable initializers and computed method keys #5387, fix(browser): reject invalid UTF-8 in HTTP CDP discovery #4524 and docs(hermes): explain nested-home routing and preview limits #4145.
• 25 October: fix(mcp): normalize compact schemas for strict gateways #3806.
• 24 October, late: feat: add FreeBuff agent support for REA setup #2736.
• 24 October: fix(importer): reach package-main selection for trailing-slash directory references #4219, fix(contracts): reject NUL bytes in objdump and rizin local path inputs #3600 and feat(android): project a structured manifest component inventory #5025.
• 23 October: feat(javascript): link assignment right-hand sides to their writes #6206, feat(analysis)!: default function MCP output to retained summaries #5660, fix(cli): advertise required CLI selector choices #6759, fix(pe): accept WIN_CERTIFICATE padding counted in dwLength #6721, fix(ghidra): report kernel OOM kills as a memory constraint #6748, docs(readme): show a dark showcase image on dark themes #6915 (mine), docs: add missing src/ directories to architecture.mermaid #6908, docs(website): add dark-mode styles to the favicon and social card #6909 (mine), fix(setup): keep the CLI-only skill choice offered under a client ID #6401, fix(javascript): associate requested bindings only with resolved native addons #6402, test(process): print the progress frames when the sparse-child bound fails #6427 (mine) and fix(browser): preserve frame identity across captures #4040.
• 22 October: fix(scripts): keep the first proxy error in the package-update verifier #6362 (mine), perf(javascript): prune typed-array element reference escapes #5139, fix(browser): account for late script metadata #6358, fix(js): resolve literal URL from Request constructor in fetch calls #6364, fix(native): admit universal Mach-O files without a host slice #5501, fix(javascript): preserve ESM path intent and artifact confinement #2700, refactor(pcap): give TShark selection and version admission one owner #3407, fix(apple): stop counting single-object NIB connectors as ownerless #3502, fix(website): wrap mobile navigation at enlarged text sizes #3598, fix(website): align the language picker chevron #3805, fix(sqlite): honor selected-table reply budgets #4216, feat(python): inspect CPython pyc headers and marshal code objects #4826, feat(erlang): inspect BEAM IFF chunks and tables #4833, feat(pe): inspect Authenticode certificate tables #4830, feat(elf): inspect PT_NOTE and SHT_NOTE tables #4837 and feat(diagnostics): report a byte entropy profile for one local file #2331.
• 21 October: fix(cli): label reconciliation input failures as application workflow errors #5668, fix(sqlite): name the selected path in malformed-database errors #5564 and feat(artifacts): inventory and extract PyInstaller archives #2433.
• 20 October: fix(javascript): preserve runtime path bytes and Electron ownership #4941.
• 19 October: feat(artifacts): inspect and extract Godot PCK with retained facts #3634 and feat(mcp): serve the MCP server over streamable HTTP #4955.
• 16 October: fix(setup): deduplicate selected clients before planning #3919.
• 15 October: feat(kuna): serve procedures, pseudocode and assembly from Kuna exports #3854 and feat(kuna): add Kuna readiness, doctor and owned project cache #3832.
• 14 October: chore(main): release rea-agents 6.5.0 #1693.
• 13 October, late: fix(ghidra): fail doctor when the JVM os.arch selects an unvalidated decompiler directory #2792 and fix(electron): reject non-executable paths and contain launch rejections #2705 (mine).
• 13 October: fix(ghidra): report unapplied PE CodeView PDB limitation #2283, fix(frida): drain admitted operations and honor queued cancellation #2305, fix(browser): retain optional source-map requires #2886, fix(target): probe WASM before inferred database routing #2007, feat(native): search analyzed memory for masked byte patterns #2341, fix(javascript): preserve CommonJS lookup and analysis provenance #2306, fix(process): release handles and close worker on incomplete cleanup #2321, feat(ghidra): admit recordless legacy MIPS32 ELFs with a pinned 32-bit language #1898, fix(ghidra): warn when callee/caller reads stop at a truncated function body #2360, fix(browser): distinguish SVG links from HTML XLink attributes #2480, fix(pwntools): let callers widen the artifact input bound #2486, fix(formats): preserve high bits in remaining magic-byte checks #2384, fix(targets): retain unsupported PE machine in diagnostics #2448, fix(process): resolve POSIX backslash basenames through PATH #2471, fix(ghidra): pre-create the launch-support config directory #2470, fix(managed): classify 32-bit-preferred assemblies as platform neutral #2342, fix(cli)!: keep a selected JSON null in --filter-output output #2333, fix: preserve process identities in pwntools cleanup errors #2284, fix(javascript): keep paths after __dirname inside the source directory #2366, fix(javascript): parse explicit CommonJS wrapper syntax #2475, fix(evidence): admit bundle imports across session transitions #2303 and fix(native): name the retained pwntools process group in cleanup resources #2281.
• 12 October: fix(artifacts): bound archive decompression by source size #2144, fix(javascript): bound value evaluation through nested closure aliases #1941, fix(ghidra): keep JVM warnings out of LaunchSupport output #1698, fix(adb): count UTF-8 bytes in logcat total_bytes #2212, fix(native): retain failed pwntools process and runtime cleanup #1708, feat(flutter): identify Dart snapshot builds and inspect AOT payloads by pure APK parsing #1582, feat: add Frida instrumentation support #1359, fix(ghidra): retain launcher command outcome in launch failure diagnostics #2011, fix(ghidra): recover stale clients on same-target reopen #2014 and fix(scripts, install): honor test:fast "without build" promise and orient Node.js engine mismatches #2026.

• #8092 test, tested at 34fd950 in a Fedora VM. All 16 tests pass on Node 22, 24 and 26. With scripts/rea.mjs put back to the merge base, 2 fail on all three, so the new cases fail without the fix.
• #8046 test, tested at c6be3b9 in a Fedora VM. The 3 native UI test files pass on Node 22, 24 and 26. With the PR's 5 source files put back to the merge base, 4 tests fail on each version.
• #8159 test, tested at 801dd59 in a Fedora VM. verifierSourceLoading.test.ts passes on Node 22 and 24. On Node 26, 3 of 6 tests fail with the PR's new, clearer error. fix(verifiers): transform verifier source on Node 26 #7891 changes the same loader so they pass.
• #8025 test, tested at 69b5408 in a Fedora VM. The new MCP test fails on Node 22, 24 and 26: it expects capability_unavailable but gets cleanup_incomplete. Two output-limit cleanup cases fail without the fix and pass with it.
• #7990 test, tested at 6829872 in a Fedora VM. The test file passes on Node 22, 24 and 26, and 1 test fails with two source files put back. Validate author declaration and PR readiness failed in CI.
• #7891 test, tested at f483ef1 in a Fedora VM. All 5 tests pass on Node 22, 24 and 26, and with the old loader put back 3 of 5 fail on Node 26 only. I did not run the verifier commands on Node 26.
• #7776 test, tested at df4f328 in a Fedora VM. Both test files pass on Node 22, 24 and 26, 64 tests each. With three changed files put back, 4 tests fail on every version.
• #7734 test, tested at f8d36e1 in a Fedora VM. All 4 tests pass on Node 22, 24 and 26. With the fix file put back, 2 of them fail, so those 2 fail without the fix and pass with it.
• #7733 test, tested at 1d54f06 in a Fedora VM. All 85 tests in the three test files pass on Node 22, 24 and 26. With the old source files put back, 8 tests fail on each version.
• #7729 test, tested at 47dd004 in a Fedora VM. All 14 tests pass on Node 22, 24 and 26. With the old javascriptSourceMapPaths.ts put back, 2 fail, so they fail without the fix.
• #7698 test, tested at 47a72f4 in a Fedora VM. All 129 tests in modelArtifact.test.ts pass on Node 22, 24 and 26, and the new test fails with GgufHeader.ts put back. CI has failed runs in Validate author declaration and PR readiness.
• #7579 test, tested at 90b72ef in a Fedora VM. All 4 tests in doctorEmptyTargetCli.test.ts pass on Node 22, 24 and 26, and 3 fail with src/cli/setupCommands.ts put back. fix(cli): refuse explicitly empty doctor targets #7576 edits the same doctor --target option.
• #7577 test, tested at b013bd8 in a Fedora VM. Both tests pass on Node 22, 24 and 26. With src/pcap/PcapProvider.ts put back to the merge base, both fail on all three, so they fail without the fix.
• #7576 test, tested at 60a04c6 in a Fedora VM. The 3 new doctor tests pass on Node 22, 24 and 26, and all 3 fail with the old src/cli/setupCommands.ts put back. fix(cli): reject explicit empty doctor targets before path admission #7579 edits the same doctor --target option.
• #7512 test, tested at fd2d6d4 in a Fedora VM. The DSH setup test file passes on Node 22, 24 and 26, 11 tests each. With DshCordisYaml.ts put back, the 2 new CLI tests fail.
• #7502 test, tested at 404d2aa in a Fedora VM. All 8 tests pass on Node 22, 24 and 26, and 2 fail with the old inputIssueProjection.ts put back. CI fails in hermesClientSetup.test.ts, which the PR does not change.
• #7473 test, tested at 925c170 in a Fedora VM. All 43 tests pass on Node 22, 24 and 26, and 1 fails with install.sh put back. Both new installer tests failed on CI's macOS runner, and I could not reproduce that.
• #7471 test, tested at 0369595 in a Fedora VM. All 26 tests in the two APK test files pass on Node 22, 24 and 26. With the two source files put back, 8 of them fail on all three.
• #7465 test, tested at 137fddd in a Fedora VM. All 33 doctor tests pass on Node 22, 24 and 26. With src/mcpDoctor.ts put back, 2 fail on all three, so they fail without the fix.
• #7464 test, tested at b61e5d6 in a Fedora VM. The new test passes on Node 22, 24 and 26. With the old src/cli/nativeCallCommands.ts put back, it fails on all three.
• #7449 test, tested at ed7e795 in a Fedora VM. All 28 tests pass on Node 22, 24 and 26, and 6 fail with the 4 changed source files put back. Validate author declaration and the PR readiness status show failure in CI.
• #7439 test, tested at f08494e in a Fedora VM. All 34 tests pass on Node 22, 24 and 26. With browserScenarioDiff.ts put back, 2 fail, so they fail without the fix.
• #7367 test, tested at 2cfd6a9 in a Fedora VM. All 11 tests pass on Node 22, 24 and 26. With src/main/transport.ts put back, 2 fail on each version, the port-in-use case among them.
• #7360 test, tested at 220c14a in a Fedora VM. The PR's 3 changed test files pass on Node 22, 24 and 26, 79 tests each. With the 2 changed source files put back, 3 tests fail on every version.
• #7358 test, tested at a73c373 in a Fedora VM. All 57 tests in the 3 changed test files pass on Node 22, 24 and 26. With src/domain/evidenceBundle.ts put back, 3 fail on each version, one per file.
• #7319 test, tested at 2f43ac3 in a Fedora VM. Both tests pass on Node 22, 24 and 26. With binaryDiagnosticsCommands.ts put back, the close-fails test fails on all three, so it checks the fix.
• #7317 test, tested at 6d26254 in a Fedora VM. All 45 tests in the two devicetree test files pass on Node 22, 24 and 26. With DevicetreeParser.ts put back, 10 of them fail on each version.
• #7309 test, tested at cf9ccce in a Fedora VM. Both tests pass on Node 22, 24 and 26. With the old EvidenceMcpServerCore.ts put back, both fail at version negotiation. So the red run does not test the oversized-error path alone.
• #7287 test, tested at 6ad7550 in a Fedora VM. All 26 tests in FilesystemCoverage.test.ts pass on Node 22, 24 and 26. With FilesystemSnapshot.ts put back, 4 fail on all three.
• #7305 test, tested at 0b7f6f5 in a Fedora VM. The 10 tests in devicetreeView.test.ts pass on Node 22, 24 and 26. The PR body says 9, and the unsupported-view-kinds test never checks the AnalysisInputError class it names.
• #7284 test, tested at e14bbfd in a Fedora VM. Both tests pass on Node 22, 24 and 26. With decompileDirectoryCommand.ts put back, the new test fails, so it fails without the fix and passes with it.
• #7278 test, tested at 96b7b1d in a Fedora VM. Both new tests pass on Node 22, 24 and 26, and both fail with cliResultOutput.ts put back. CI fails in files the PR does not change, and I did not run those tests.
• #7268 test, tested at 01dab07 in a Fedora VM. All 37 tests pass on Node 22, 24 and 26. With the three fix files put back, 3 tests fail on all three versions.
• #7261 test, tested at 3445223 in a Fedora VM. All 180 tests in the 2 changed test files pass on Node 22, 24 and 26, and 7 fail with the 3 source files put back. I did not test a real version 24 game binary.
• #7242 test, tested at 8eee756 in a Fedora VM. All 16 tests in EnhancedTools.objcClasses.test.ts pass on Node 22, 24 and 26. With the old EnhancedTools.ts put back, 3 fail on every version.
• #7241 test, tested at 60ac19f in a Fedora VM. All 53 tests pass on Node 22, 24 and 26, and 2 fail without the fix. From reading the code, a cancel during the last read now returns bytes, and the PR removes its in-flight cancel test.
• #7197 test, tested at 9630917 in a Fedora VM. All 52 tests in the 2 test files pass on Node 22, 24 and 26. From reading the code, decideHypothesis takes an empty evidenceIds array, and no test covers that.
• #7240 test, tested at 2e78511 in a Fedora VM. The test file passes on Node 22, 24 and 26, 71 tests each. The PR's net diff only adds a test, because the source file matches the merge base.
• #7166 test, tested at fa44319 in a Fedora VM. All 38 tests in KeyedArchiveReader.test.ts pass on Node 22, 24 and 26. With two source files put back, 3 fail on each version.
• #7161 test, tested at 820fdb2 in a Fedora VM. All 73 tests pass on Node 22, 24 and 26, and 5 fail with environmentCompareDiff.ts put back. The PR body's counts look out of date, and CI fails in test files the PR did not change.
• #7160 test, tested at c5b624f in a Fedora VM. All 42 tests pass on Node 22, 24 and 26, and 2 fail with PeCertificateAsn1.ts put back. From reading the code, no test accepts exactly 64 levels of nesting.
• #7145 test, tested at e60accf in a Fedora VM. The build failed: npm run test:prepare exited 2, so 2 of the 3 test files would not load. src/process/ProcessAdmissionError.ts is missing from the branch but exists on main.
• #7122 test, tested at 378e2df in a Fedora VM. All 52 tests pass on Node 22, 24 and 26, and 2 fail with jsonSchemaKeywords.ts put back. The PR body lists schemaChildEntries and literalSchemaType, and I found neither in this head.
• #7100 test, tested at 3bf2421 in a Fedora VM. The 10 tests in rustBinaryView.test.ts pass on Node 22, 24 and 26. Two test titles name an indices filter that neither test uses.
• #6778 test, tested at 874cfd3 in a Fedora VM. All 356 tests in the 3 changed test files pass on Node 22, 24 and 26. With javascriptRouteScan.ts put back, 26 fail on all three.
• #7074 test, tested at acdbe88 in a Fedora VM. All 15 tests pass on Node 22, 24 and 26. From reading the code, a retry in scripts/verify-adb.mjs may print [FAIL] for a probe that was already removed, but I did not run that case.
• #6923 test, tested at 770c049 in a Fedora VM. The 4 changed test files pass on Node 22, 24 and 26, 34 tests each. From reading the code, the None / Skip Python choice sets up the same environment as Standard, but I did not run it.
• #6886 test, tested at 5650ca8 in a Fedora VM. The 4 changed test files give 93 passed, 10 skipped on Node 22, 24 and 26, matching the body. From reading the diff, the Darwin fixture changes on every host, which the body does not say.
• #4807 test, tested at 9a3fb67 in a Fedora VM. Two tests in cliFormattingCapacity.test.ts fail on Node 22, 24 and 26: the CLI now exits 0 where they expect exit 1. CI's Test shard (2/8) and Portable tests fail the same 2.
• #5180 test, tested at edb61d6 in a Fedora VM. All 11 tests pass on Node 22, 24 and 26. With ApktoolConfiguration.ts put back, the new test distinguishes launcher spawn EPERM fails.
• #5022 test, tested at 9c7b2ea in a Fedora VM. All 9 tests pass on Node 22, 24 and 26, and 4 fail with the old registry-state.mjs put back. From reading the code, a missing server.json would be reported as a registry read failure; I did not run it.
• #4894 test, tested at 25ab1ca in a Fedora VM. All 21 tests in the two update test files pass on Node 22, 24 and 26. With UpdateCommandProgress.ts put back, 4 new tests fail on all three.
• #6700 test, tested at 7f1a7a8 in a Fedora VM. All 9 tests pass on Node 22, 24 and 26, and the 3 new tests fail with HopperDownload.ts put back. The PR also changes an old test: a hop to downloads.hopperapp.com is now refused.
• #6699 test, tested at 9754828 in a Fedora VM. All 6 tests pass on Node 22, 24 and 26. With the fix removed, the new test fails on all three, so it checks the fix.
• #6410 test, tested at 4297b5e in a Fedora VM. Both test files pass on Node 22, 24 and 26, but I could not run the Windows fixes. The body says each platform keeps its transport. But the diff changes the Windows default in GhidraClient.
• #6383 test, tested at 7ac6d9b in a Fedora VM. The one test file the PR touches passes on Node 22, 24 and 26, 10 tests each. The PR only adds one test, and the README and doc fixes listed in the body are not in the diff.
• #4618 test, tested at efb27cf in a Fedora VM. All 80 tests in the 3 changed test files pass on Node 22, 24 and 26. The PR body is out of date in three places: the validation head, the client count and the release line.
• #4536 test, tested at d133505 in a Fedora VM. The PR's test file passes at the head and on main, 7 tests each, on Node 22, 24 and 26. The PR adds only 2 tests, and its merge base already has morluto's fix 556504e.
• #5275 test, tested at 53a9c85 in a Fedora VM. The build fails on a compile error, so no test ran. The PR removes the macosMajorVersion import from DoctorDiagnostics.ts, but line 133 still calls it.
• #4696 test, tested at 9421b18 in a Fedora VM. All 133 tests in the 2 changed test files pass on Node 22, 24 and 26. I did not run a real apktool decode.
• #4675 test, tested at a206014 in a Fedora VM. The changed test file passes on Node 22, 24 and 26, 35 tests each. CI's Static checks fail: an architecture guard rejects a walker in javascriptSemanticClassReceivers.ts.
• #4046 test, tested at 7010ec4 in a Fedora VM. All 52 tests in the 3 changed test files pass on Node 22, 24 and 26, as the PR body says. A text search finds four generated fixtures that no test uses.
• #9209 retest, tested at 0433506 in a Fedora VM. The 4 changed test files pass on Node 22, 24 and 26 (37 tests), and the worker now wraps the error once. I asked if #lastFailure and #stop getting the error before the wrap is on purpose.
• #9206 retest, tested at a06b979 in a Fedora VM. The 7 changed test files pass on Node 22, 24 and 26 (221 tests). The cancellation test now fires inside the region index, and a new table test expects the __rodata drop I found.
• #9203 retest, tested at 69f6099 in a Fedora VM. The PR's 7 changed test files pass on Node 22, 24 and 26 (92 tests). A new test covers the gap from my review: two different socket ids now stay two sockets.

Earlier results, links only (812)

• #9209 test, #9206 test, #9211 test, #9223 test, #9214 test, #9209 review, #9206 review, #9204 review, #9203 review, #8447 re-review, #6038 read, #5861 read, #5116 review, #4987 read, #4915 re-review, #4829 review, #4021 review, #3813 check, #3133 check, #2498 check, #2156 read, #1587 read, #829 check, #487 check, #1959 retest, kuna #1034, #6395 retest, #6188 retest, #5672 retest, #5539 retest, #4915 retest, #4897 retest, #4784 retest, #4430 retest, #2876 retest, #8447 review
• #3899 test, #4323 retest, #5145 retest, #4572 retest, #3754 retest, #3940 retest, #3586 retest, #2025 retest.
• #1971 retest, #6188 round 3 review, #4897 round 3 review, #6342 re-check, #6350 re-check, kuna #1034, kuna #1034, kuna #1034.
• kuna #1034, kuna #1034, kuna #1034, kuna #1034.
• #6188 follow-up review, kuna #1034, #2041 reply, #2185 reply, #2465 reply, #3525 reply, #2919 reply, #2142 reply
• #2018 reply, #5672 reply, #3026 reply, #6395 reply, #4198 reply, #5387 retest, #5125 reply, #1840 retest
• #3940 retest, #1971 retest, #5596 retest, #3098 reply, #5127 reply, #6097 result, #6178 result, #6784 PR
• #2043 retest, #6184 review, #3004 review, #4572 review, #5303 review, #5133 review, #6915 PR, #6908 reply
• #5936 review, #5478 review, #6146 review, #6753 review, #5764 review, #6759 review, #4992 review, #6177 review
• #5366 review, #6748 review, #4970 review, #6761 review, #6498 review, #6303 review, #6248 review, #6129 review
• #5777 review, #6909 PR, #2978 reply, #3468 reply, #3425 reply, #5044 reply, #5285 reply, #5127 reply
• #6074 review, #5938 review, #6307 review, #5937 review, #5757 review, #5213 review, #5204 review, #5778 review
• #3288 review, #6386 review, #6351 review, #6744 review, #6721 review, #6384 review, #4897 review, #5669 result
• #6396 result, #5491 result, #5125 result, #5387 result, #6202 review, #6199 review, #5618 review, #5491 review
• #5387 review, #5125 review, #5174 review, #6190 result, #6633 review, #6632 review, #6396 review, #6392 review
• #6326 review, #6317 review, #6294 review, #6276 review, #6243 review, #6241 review, #6206 review, #6901 issue
• #6190 review, #6142 review, #6180 review, #6203 review, #6188 review, #6184 review, #6171 review, #6073 review
• #6054 review, #5962 review, #5933 review, #5511 review, #4897 review, #5869 review, #5790 review, #5755 review
• #5754 review, #5669 review, #5667 review, #5599 review, #5539 review, #5517 review, #5430 review, #5317 review
• #5307 review, #5248 review, #4720 review, #3996 review, #2884 review, #5305 review, #5285 review, #5263 review
• #5247 review, #5013 review, #4988 review, #4430 review, #5877 claim, #3796 tracker refresh, #3795 tracker refresh, #5654 tracker refresh
• #5653 tracker refresh, #6311 reply, #5861 reply, #5127 reply, #3810 reply, #3458 reply, #3124 reply, #2966 reply
• #5544 review, #5023 review, #4367 review, #4359 review, #3513 review, #3263 retest, #5545 retest, #3600 retest
• #3562 retest, #3420 retest, #3004 retest, #1815 retest, #5304 review, #5303 review, #5227 review, #5226 review
• #5172 review, #5157 review, #5067 review, #5045 review, #6373 review, #4386 review, #4323 review, #6433 claim
• #3581 review, #3425 review, #3940 retest, #5145 review, #5133 review, #5044 review, #5039 review, #5028 review
• #5020 review, #4920 review, #4897 review, #4519 retest, #5061 retest, #6395 reply, #3293 reply, #3968 reply
• #4884 reply, #3463 reply, #4036 reply, #3994 reply, #4790 review, #4784 review, #4741 review, #4767 review
• #4198 reply, #4040 reply, #4719 review, #4611 review, #3940 reply, #4572 review, #6402 review, #4596 review
• #6401 review, #6395 review, #6373 review, #3940 retest, #2977 retest, #1815 retest, #4524 review, #4532 review
• #4410 review, #4482 review, #4408 review, #4334 review, #4283 review, #4323 review, #4237 review, #4221 review
• #3940 retest, #4220 review, #6350 review, #4145 review, #4198 review, #6364 review, #6358 review, #6342 review
• #3004 retest, #5654 reply, #2977 reply, #2150 retest, #4036 review, #4040 review, #4973 reply, #3521 reply
• #3940 reply, #3026 reply, #3995 review, #3994 review, #3805 review, #3968 review, #5244 reply, #3463 reply
• #3407 reply, #2736 reply, #3502 reply, #6362 PR, #6038 claim, #6178 reply, #3940 retest, #4837 retest
• #3585 retest, #5597 reply, #5156 reply, #3004 retest, #2150 retest, #5139 retest, #4837 retest, #2839 reply
• #3356 reply, #2876 reply, #3498 reply, #3582 tracker edit, #3502 review, #6178 PR, #6097 claim, #3463 review
• #3544 review, #3581 review, #3754 review, #3495 review, #3407 review, #3339 review, #3411 review, #3425 review
• #3225 review, #3293 review, #3582 tracker edit, #3582 tracker edit, #2988 review, #3100 review, #2736 review, #2963 review
• #6089 review, #3537 tracker edit, #3537 tracker edit, #3795 tracker edit, #3796 tracker edit, #4330 tracker edit, #4355 review, #4420 review
• #4559 review, #3598 review, #3848 review, #4004 review, #4037 review, #4214 review, #4333 review, #4386 review
• #3397 results, #3535 results, #3551 results, #4509 results, #5074 results, #5085 results, #5111 results, #5147 results
• #5244 results, #5333 results, #5396 results, #5397 results, #5399 results, #5402 results, #5403 results, #3029 review
• #3026 retest, #3802 retest, #3756 reply, #3940 review, #4536 review, #2694 retest, #4224 review, #4442 review
• #2977 retest, #3028 review, #3024 retest, #2887 review, #3026 retest, #4417 retest, #4911 retest, #3940 review
• #3029 review, #4953 review, #5660 review, #5545 review, #5519 review, #4830 retest, #4826 retest, #4837 retest
• #5672 review, #5501 review, #5596 review, #2433 retest, #3940 retest, #5544 review, #5597 review, #5471 review
• #5668 review, #5495 review, #5564 review, #5023 retest, #5025 retest, #2150 retest, #5025 retest, #5061 retest
• #5127 review, #4367 review, #3756 reply, #4911 reply, #5659 result, #5620 result, #5666 result, #5023 retest
• #4997 retest, #3802 review, #5216 review, #5156 review, #5428 review, #5578 review, #5565 review, #5376 review
• #4219 review, #5466 result, #5542 result, #5462 result, #5386 result, #5105 result, #3531 review, #5401 result
• #5408 result, #5457 result, #5458 result, #5520 result, #5077 result, #5436 result, #5448 result, #5474 result
• #5493 result, #5086 result, #5398 result, #5455 result, #5525 result, #5531 result, #4911 review, #4359 review
• #1741 result, #3275 result, #4984 result, #5323 result, #4973 result, #5325 result, #4997 review, #4941 review
• #4915 review, #3004 reply, #3531 review, #1815 retest, #3940 review, #4941 review, #4973 result, #2406 retest
• #3076 review, #3004 review, #5139 review, #3521 review, #4955 review, #3026 review, #5024 review, #4833 review
• #4830 review, #4831 review, #4359 review, #4519 review, #4958 review, #4826 review, #5061 review, #4843 review
• #4915 review, #4911 review, #4973 result, #2961 result, #2872 result, #2876 retest, #4493 review, #4941 review
• #5025 review, #3830 result, #3627 result, #3505 result, #3504 result, #3462 result, #5023 follow-up, #3940 retest, round 3
• #4994 review, #4995 review, #5023 review, #5025 review, #4417 retest, #5073 repro, #4089 repro, #3940 retest
• #4837 retest, #4997 review, #1741 reproof, #3586 retest, #1389 re-time, #3026 retest, #2433 retest, #1815 retest
• #1704 retest, #3426 retest, #2876 retest, #2331 retest, #2268 retest, #2978 retest, #2068 retest, #3657 retest
• #2041 retest, #4626 review, #1922 retest, #4837 review, #4483 repro, #4467 repro, #4471 repro, #4417 review
• #2150 retest, #4216 review, #4259 review, #3940 review, #1840 retest, #2870 result, #3750 result, #2068 retest
• #2041 retest, #1384, #1376, #1365, #1362, #1359, #2148, #2142
• #2043, #1961, #1960, #1959, #1704, rsleigh #27, #287, #3858 overnight runs
• #3940, #3920, #3919, #2414 rerun, #1957, #3854, #3832, #3847
• #2093, #3858 quality read, #3858, #2702, #2700, #2694, #2676, #2631
• #2693, #2630, #2513, #2492, #2465, #2435, #2433, #2407
• #2418, #2414, #2406, #2362, #2358, #2331 retest, #2268, #2264
• #2243, #2199, #2185, #2150 retest, #2148, #2142, #2136, #2124
• #2097, #2094, #2092, #2068 retest, #2081, #2052, #2043, #2025
• #1961, #1903 on a tablet, #2041, #1971, #1960, #1959, #1948, #1922 retest
• #1903, #1840, #1704, #1815, #1804, #1810, #1732, #1065
• #1672, #3271, #3071, #2018 sixth run, #3315, #3314, #3311, #3462
• #3582 round 2, rsleigh #21 to #26, #3210, #3340, #3347, kuna #1087, kuna #1085, kuna #1034
• #2408, #3035, #1998 benchmark on 6.5.0, #3026 again, #1693, the 6.5.0 smoke test, #2997, #3015, #2977 retest
• #2705 follow-up, #2981, #2977, #2792, #2705 check, #2876, #1922 retest, #2886
• #2879, #2283 retest, #2018 fifth run, #2470, #2018 fourth run, #2475, #2480, #2471
• #2341 retest, #2486, #2366, #2448, #2384, #2360, #2412, #2342
• #2341, #2303, #2321, #2408, #2305, #2317, #2331, #2306
• #2333, #2272, #2283, #2281, kuna PR 1046, #2280, kuna #1034, #2326
• #2244, #2144, #2304, #2227, #2068, #2212, #2150, #2020
• #1941 retest, #2254, #2231, #1048, #2018 retest, #2014 retest, #2026, #2018
• #2014, #2011, #2039, #2007, #1959, #2010, kuna #1031 follow-up, #2006 follow-up
• #1957, #1389
• #3796, #3796
• kuna #1031, #1922, rsleigh #19, rsleigh #12, #1998, #2006
• #1986, #1977, #1925 follow-up, #1978, #1987, #1981, #1975, #1974
• #1979: LE, LX and other formats, real files and the rest
• rsleigh: #17, #18, #15, #16, #14, #12, #13, #8, #7
• Discussion #1958, #1951
• #1949, #1941, #1954, #1952 note, #1939, #1919
• feat(ghidra): admit recordless legacy MIPS32 ELFs with a pinned 32-bit language #1898: float code, legacy MIPS32
• #1704 retest, #1359 retest
• #1925, #1887
• feat(flutter): identify Dart snapshot builds and inspect AOT payloads by pure APK parsing #1582, three retests: newest, middle and first
• Also: #1700, #1708, #1715, #1722, #1705, #1719, #1720, #1711
• Also: #1698, #1706, #1704, #1703, #1686, #1707, #1650, #1048
• Also: #1357, #1661, #1668, #1673, #1670, #1669, #1476, #1356
• Also: #1657, #1667, #1666, #1658, #1659, #1665, #1664, #1662
• Also: #1447, #1494, #1620, #1600, #1566, #1608, #1611, #1612
• Also: #1610, #1583, #1065, #1389, #1387, #1417, #1363, #1645
• Also: #1642, #1635, #1629, #1649, #1441, #1607
• Amethyst Mod Manager (GPL-3.0-or-later), a look with REA and Ghidra at its two Rust libraries.
• The 6.4.0 npm smoke test: it installs and runs Ghidra on Node 22, 24 and 26. The MCP tool list grew from 139 tools to 170.
• Also: #3858 follow-up, #2021, #2024, #2282, #2284, #6784 close note

Roadmap items a Linux box can help with

• #1065 (MCP under tight memory) and the verify:ghidra lanes.
• #717 (Android): real APKs on JDK 21.
• #968 (Frida) and #774 (websites): both run here.
• #64 (Electron): captures of Flathub apps.

Not for this machine: Windows, macOS, iOS and Hopper.

Test setup

• Box: Nobara 44 (Fedora 44 base), Linux 7.3.1
• CPU: AMD Ryzen 9 5900X, 12 cores / 24 threads
• RAM: 32 GB DDR4-3600
• Node: 22.23.4, 24.22.1, 26.12.1
• Ghidra: 12.1.4 (Apache-2.0)
• Other people's PR code: an offline Fedora 44 VM (4 vCPUs, 8 GB), reset to a clean snapshot each run

🤖 Generated with Claude Code (Sonnet 5.5, verified by Opus 5.5) 🦊

[elkahtersidi]

سلام عليكم

[CassieFox23]

Moved: this benchmark now has its own page, REA startup benchmark. I update it there each release.

Latest: on Node 24, main (13e03e9) lists its tools in 1.47 s, 16 percent less than 6.4.0 (1.75 s).

🤖 Generated with Claude Code (Opus 5.5) 🦊

[FunToHard]

Bots flooded social media now this😔