v6.26.0
This commit simply changes a version label from '6.26.0-dev' to '6.26.0' in a single header file. It is a routine release-versioning change with no functional code modifications and no security relevance.
b2ca7248by XMRig+1−11 fileWidely deployed RandomX CPU/GPU miner with pool and network-facing code.
Every captured commit receives deterministic security triage and a separate communication-quality score. Security candidates and broader second-pass signals receive full-patch Ollama analysis.
Sep 27 → Mar 3147 seen35 candidatesCompleteMar 31 → Jul 290 seen0 candidatesCompleteJul 29 → Aug 280 seen0 candidatesCompleteAug 28 → Sep 270 seen0 candidatesCompleteMessage quality measures whether a commit identifies its scope, purpose, rationale, testing, and supporting references. It does not change the security-severity score.
Public Git author strings; identities are not independently verified.
Last scanned 35 minutes ago
This commit simply changes a version label from '6.26.0-dev' to '6.26.0' in a single header file. It is a routine release-versioning change with no functional code modifications and no security relevance.
b2ca7248by XMRig+1−11 fileThis is a large routine feature release for the XMRig cryptocurrency miner. The main changes add support for a new mining algorithm variant called RandomX v2 (used by an upcoming Monero network upgrade), improve performance on RISC-V and Z…
92705f2dby XMRig+10050−667286 filesThis commit is a routine version bump for the next development release of XMRig, a cryptocurrency mining program. It only updates the version number and copyright year, and adds a changelog listing new features and bug fixes from previous …
4f58a7afby XMRig+20−62 filesThis is a tiny one-line build fix for ARM64 processors. It changes a memory prefetch helper to accept a read-only (const) pointer, which allows the code to compile on ARM64 systems. There is no security issue visible in the change itself.
806cfc3fby xmrig+1−11 fileThis is a one-line build fix for ARM64 processors. It changes a memory prefetch helper to accept read-only (const) pointers, which is required by newer or stricter ARM64 compilers. There is no security bug being fixed here—just a compilati…
84352c71by SChernykh+1−11 fileThis is a minor build-compatibility fix for XMRig when compiled with the Clang compiler on 64-bit ARM (ARM64) processors. It changes a preprocessor condition so that Clang on ARM64 does not try to use a compiler feature (__builtin_prefetch…
27d535d0by xmrig+1−11 fileThis is a one-line build fix for compiling XMRig on ARM64 systems using the Clang compiler. It changes a preprocessor condition so that Clang on ARM64 does not try to use a GCC-specific memory prefetch instruction that Clang does not suppo…
9d296c7fby SChernykh+1−11 fileThis patch changes how XMRig resets its internal counter (nonce) when switching between the user's mining job and the built-in donation mining job. Previously, the nonce was reset during these switches, which could cause duplicate work and…
3b4e38baby xmrig+3−11 fileThis is a small bug-fix change in the XMRig cryptocurrency mining software. It changes how the mining 'nonce' (a kind of counter) is reset when the miner briefly switches to a 'donation round' (a short period where mined rewards go to the …
8b33d249by SChernykh+3−11 fileThis commit fixes how XMRig's RandomX mining code preloads data from the RandomX dataset before starting the main hashing loop. The old prefetch was only done inside assembly prologue code and only used one address, which was wrong for the…
a9356412by xmrig+16−124 filesThis commit moves where a data prefetch happens in the RandomX cryptocurrency mining code and adds support for a second prefetch in a newer RandomX variant. Prefetch instructions only hint to the CPU what memory might be needed soon; they …
976a08efby SChernykh+16−124 filesThis commit changes how the XMRig miner handles a new RandomX variant (RandomX v2). Previously, when switching to or from RandomX v2, the miner would regenerate its large dataset. Now it reuses the existing dataset, treating RandomX v2 and…
17aa97e5by xmrig+12−42 filesThis commit tweaks how the XMRig miner handles switching between two closely related RandomX algorithm variants (RX/0 and RX/V2). It avoids rebuilding the large RandomX dataset when the miner switches between these two variants, because th…
48b29fd6by SChernykh+12−42 filesThis commit is titled 'ARM64 fixes' and changes how the RandomX mining algorithm's ARM64 code handles AES lookup tables and memory checks. The code now stores AES encryption/decryption table addresses into a dynamically allocated code buff…
8464d474by xmrig+32−2714 filesThis commit fixes two ARM64-specific issues in the XMRig cryptocurrency miner. First, it removes duplicate copies of AES lookup tables from the compiled RandomX code, instead reusing the existing tables already in memory. Second, it adds a…
26ee1cd2by SChernykh+32−2714 filesThis commit updates XMRig's mining code to support a new Monero block format called FCMP++ (a recent network upgrade). It changes how the software builds the 'Merkle tree' cryptographic proof that links the coinbase transaction to the bloc…
d9b39e8cby xmrig+116−486 filesThis commit updates XMRig's mining code to support a new Monero block format called FCMP++. It changes how the software builds the transaction hash tree (Merkle tree) used when mining. There is no clear security bug here; it appears to be …
05dd6dccby SChernykh+116−486 filesThis commit updates the XMRig mining software to support a new RandomX v2 protocol feature. When a mining pool asks for it, the miner now includes an extra 'commitment' value alongside its share submissions. The change is a feature additio…
316a3673by xmrig+49−156 filesThis commit updates the XMRig mining software to support a new 'commitment' field required by an upcoming RandomX version 2 protocol. It changes how miners package and submit proof-of-work results to mining pools. There is no direct eviden…
42c943c0by SChernykh+49−156 filesThis commit adds support for a new version of the RandomX mining algorithm (RandomX v2) on RISC-V processors. It is a feature/algorithm update, not a security patch. There is no indication in the commit that it fixes a vulnerability or tha…
c2c80807by xmrig+398−10310 filesExpand any commit for its author, full message, clarity score, changed files, triage signals, analysis, and source link.
cb600194 · Jan 31, 2026 · 4 filesInformational 15RandomX v2 (ARM64)
Why it was queuedThis commit updates the RandomX mining algorithm implementation for 64-bit ARM (ARM64/AArch64) processors. It adds support for a new variant of RandomX (v2), including AES-based mixing of register groups, a software fallback for devices without hardware AES, and a tweak to floating-point rounding behavior. These are normal, expected changes for a cryptocurrency miner to stay compatible with an evolving proof-of-work algorithm. There is no indication in the commit or supplied references that this is a security fix or vulnerability.
9d71358f · Jan 30, 2026 · 26 filesInformational 15RandomX v2 + commitments
Why it was queuedThis commit adds support for a new mining algorithm variant called 'RandomX v2' plus a 'commitments' feature in the XMRig cryptocurrency miner. It is a routine protocol/algorithm update: new benchmark hashes, new algorithm identifiers, new assembly and JIT compiler code paths, and a small change to how CPU workers compute and store an extra commitment value when the new variant is active. There is no indication of a security vulnerability, backdoor, or malicious behavior in the diff.
67cc6cfd · Jan 21, 2026 · 15 filesInformational 15fix: cross typos detail below:
What I did
Repository rules / exclusions
I didn’t find repo-specific spelling tooling already in place (no existing codespell config).
Given this is a C/C++ repo with vendored sources, I treated src/3rdparty/** as third-party and excluded it from typo fixing (and encoded that in the new .codespellrc).
Typos fixed (first-party only)
Docs
CHANGELOG.md: perfomance -> performance
doc/API.md: optionaly -> optionally, Offical -> Official
doc/BENCHMARK.md: parameteres -> parameters
doc/CPU.md: loosing -> losing, instuctions -> instructions
doc/CHANGELOG_OLD.md: multiple obvious text typos like Breaked -> Broken, singal -> signal, previos -> previous, secons -> seconds, automaticaly -> automatically, perfomance -> performance
Code comments / doc comments (safe text-only changes)
src/base/crypto/sha3.cpp: Inteface -> Interface (comment banner)
src/backend/opencl/cl/cn/cryptonight.cl: performe -> perform, crashs -> crashes (comments)
src/backend/opencl/cl/kawpow/kawpow.cl: regsters -> registers, intial -> initial (comments)
src/crypto/randomx/aes_hash.cpp: intial -> initial (comment)
src/crypto/randomx/intrin_portable.h: cant -> can't (comment)
src/crypto/randomx/randomx.h: intialization -> initialization (doc comment)
src/crypto/cn/c_jh.c: intital -> initial (comment)
src/crypto/cn/skein_port.h: varaiable -> variable (comment)
src/backend/opencl/cl/cn/wolf-skein.cl: Build-in -> Built-in (comment)
What I intentionally did NOT change
Anything under src/3rdparty/** (vendored).
A few remaining codespell hits are either:
Upstream/embedded sources we excluded (groestl256.cl, jh.cl contain Projet)
Potentially valid identifier/name (Carmel CPU codename)
Low-risk token in codegen comments (vor inside an instruction comment)
These are handled via ignore rules in .codespellrc instead of modifying code.
Added: .codespellrc
Created /.codespellrc with:
skip entries for vendored / embedded upstream areas:
./src/3rdparty
./src/crypto/ghostrider
./src/crypto/randomx/blake2
./src/crypto/cn/sse2neon.h
./src/backend/opencl/cl/cn/groestl256.cl
./src/backend/opencl/cl/cn/jh.cl
ignore-words-list for:
Carmel
vor
Verification
codespell . --config ./.codespellrc now exits clean (exit code 0).
Signed-off-by: rezky_nightky <with.rezky@gmail.com>
This commit only fixes spelling mistakes in documentation and code comments. No program logic, configuration handling, or executable code was changed, so it cannot affect security or functionality.
db24bf51 · Jan 21, 2026 · 22 filesInformational 12Revert "Merge branch 'pr3764' into dev"
This reverts commit 0d9a372e494bd602d7eb40704c91ca7e6dc47f76, reversing
changes made to 1a04bf2904b8cb29320f577c78ad88f097c253a7.
This commit is a wholesale revert of a previous merge (pr3764). It undoes a large set of changes across documentation, comments, and several source files. The reverted code appears to have been a mix of typo corrections, minor cleanups, and some functional changes to latency tracking, job-result handling, thread cleanup, and memory management. Because the commit removes those changes rather than introducing new behavior, it does not by itself create a security vulnerability. It may restore older, potentially less safe patterns (for example, calling thread join without checking joinability, or removing safeguards in the job-results queue), but the commit message gives no security rationale and there is no direct evidence of an exploitable flaw being introduced here.
5ca48282 · Jan 21, 2026 · 7 filesModerate 53feat: stability improvements, see detail below
Key stability improvements made (deterministic + bounded)
1) Bounded memory usage in long-running stats
Fixed unbounded growth in NetworkState latency tracking:
Replaced std::vector<uint16_t> m_latency + push_back() with a fixed-size ring buffer (kLatencyWindow = 1024) and explicit counters.
Median latency computation now operates on at most 1024 samples, preventing memory growth and avoiding performance cliffs from ever-growing copies/sorts.
2) Prevent crash/UAF on shutdown + more predictable teardown
Controller shutdown ordering (Controller::stop()):
Now stops m_miner before destroying m_network.
This reduces chances of worker threads submitting results into a network listener that’s already destroyed.
Thread teardown hardening (backend/common/Thread.h):
Destructor now checks std::thread::joinable() before join().
Avoids std::terminate() if a thread object exists but never started due to early exit/error paths.
3) Fixed real leaks (including executable memory)
Executable memory leak fixed (crypto/cn/CnCtx.cpp):
CnCtx::create() allocates executable memory for generated_code via VirtualMemory::allocateExecutableMemory(0x4000, ...).
Previously CnCtx::release() only _mm_free()’d the struct, leaking the executable mapping.
Now CnCtx::release() frees generated_code before freeing the ctx.
GPU verification leak fixed (net/JobResults.cpp):
In getResults() (GPU result verification), a cryptonight_ctx was created via CnCtx::create() but never released.
Added CnCtx::release(ctx, 1).
4) JobResults: bounded queues + backpressure + safe shutdown semantics
The old JobResults could:
enqueue unlimited std::list items (m_results, m_bundles) → unbounded RAM,
call uv_queue_work per async batch → unbounded libuv threadpool backlog,
delete handler directly while worker threads might still submit → potential crash/UAF.
Changes made:
Hard queue limits:
kMaxQueuedResults = 4096
kMaxQueuedBundles = 256
Excess is dropped (bounded behavior under load).
Async coalescing:
Only one pending async notification at a time (m_pendingAsync), reducing eventfd/uv wake storms.
Bounded libuv work scheduling:
Only one uv_queue_work is scheduled at a time (m_workScheduled), preventing CPU starvation and unpredictable backlog.
Safe shutdown:
JobResults::stop() now detaches global handler first, then calls handler->stop().
Shutdown detaches m_listener, clears queues, and defers deletion until in-flight work is done.
Defensive bound on GPU result count:
Clamp count to 0xFF inside JobResults as well, not just in the caller, to guard against corrupted kernels/drivers.
5) Idempotent cleanup
VirtualMemory::destroy() now sets pool = nullptr after delete:
prevents accidental double-delete on repeated teardown paths.
Verification performed
codespell . --config ./.codespellrc: clean
CMake configure + build completed successfully (Release build)
Signed-off-by: rezky_nightky <with.rezky@gmail.com>
This commit is a hardening patch for the XMRig cryptocurrency miner. It fixes several stability problems that could also have security consequences: unbounded memory growth in network statistics, a shutdown crash where worker threads could use an already-destroyed network object, a leak of executable memory used by mining code, a leaked cryptography context during GPU result checking, and unsafe unbounded queues in the job-results processor. The patch adds size limits, safer teardown ordering, and proper cleanup of executable memory and contexts. There is no claim in the commit that these issues are exploitable for code execution or were reported as security vulnerabilities.
cb751150 · Jan 21, 2026 · 15 filesInformational 15fix: cross typos detail below:
What I did
Repository rules / exclusions
I didn’t find repo-specific spelling tooling already in place (no existing codespell config).
Given this is a C/C++ repo with vendored sources, I treated src/3rdparty/** as third-party and excluded it from typo fixing (and encoded that in the new .codespellrc).
Typos fixed (first-party only)
Docs
CHANGELOG.md: perfomance -> performance
doc/API.md: optionaly -> optionally, Offical -> Official
doc/BENCHMARK.md: parameteres -> parameters
doc/CPU.md: loosing -> losing, instuctions -> instructions
doc/CHANGELOG_OLD.md: multiple obvious text typos like Breaked -> Broken, singal -> signal, previos -> previous, secons -> seconds, automaticaly -> automatically, perfomance -> performance
Code comments / doc comments (safe text-only changes)
src/base/crypto/sha3.cpp: Inteface -> Interface (comment banner)
src/backend/opencl/cl/cn/cryptonight.cl: performe -> perform, crashs -> crashes (comments)
src/backend/opencl/cl/kawpow/kawpow.cl: regsters -> registers, intial -> initial (comments)
src/crypto/randomx/aes_hash.cpp: intial -> initial (comment)
src/crypto/randomx/intrin_portable.h: cant -> can't (comment)
src/crypto/randomx/randomx.h: intialization -> initialization (doc comment)
src/crypto/cn/c_jh.c: intital -> initial (comment)
src/crypto/cn/skein_port.h: varaiable -> variable (comment)
src/backend/opencl/cl/cn/wolf-skein.cl: Build-in -> Built-in (comment)
What I intentionally did NOT change
Anything under src/3rdparty/** (vendored).
A few remaining codespell hits are either:
Upstream/embedded sources we excluded (groestl256.cl, jh.cl contain Projet)
Potentially valid identifier/name (Carmel CPU codename)
Low-risk token in codegen comments (vor inside an instruction comment)
These are handled via ignore rules in .codespellrc instead of modifying code.
Added: .codespellrc
Created /.codespellrc with:
skip entries for vendored / embedded upstream areas:
./src/3rdparty
./src/crypto/ghostrider
./src/crypto/randomx/blake2
./src/crypto/cn/sse2neon.h
./src/backend/opencl/cl/cn/groestl256.cl
./src/backend/opencl/cl/cn/jh.cl
ignore-words-list for:
Carmel
vor
Verification
codespell . --config ./.codespellrc now exits clean (exit code 0).
Signed-off-by: rezky_nightky <with.rezky@gmail.com>
This commit is a spelling and typo cleanup. It fixes misspelled words in documentation files and code comments, and adds a configuration file for an automated spell-checker. No program logic, configuration parsing, network handling, or cryptographic calculations were changed. There is no security impact.
6e6eab17 · Jan 20, 2026 · 1 fileTriage 0Fix keepalive timer logic
- Reset timer on send instead of receive (pool needs to know we're alive)
- Remove timer disable after first ping to enable continuous keepalives
45d0a15c · Jan 16, 2026 · 1 fileInformational 15Optimized VAES code
Use only 1 mask instead of 2
This is a small performance cleanup in a cryptocurrency mining algorithm's AES encryption helper. It replaces two different bit masks with a single shared mask and swaps the argument order where needed so the result stays the same. There is no security bug being fixed here.
ed80a8a8 · Jan 16, 2026 · 3 filesInformational 18RandomX: added VAES-512 support for Zen5
+0.1-0.2% hashrate improvement.
This commit adds a new performance-optimized code path for AMD Zen 5 processors in the RandomX cryptocurrency-mining algorithm. It uses wider AVX-512 AES instructions to slightly increase mining speed. There is no indication this change fixes a security bug or introduces a known vulnerability; it is a routine optimization.
e41b28ef · Jan 15, 2026 · 3 filesInformational 12Improved RISC-V code
Why it was queuedThis commit refactors the RISC-V RandomX mining code in XMRig. It removes a dynamic analysis that tried to insert scratchpad prefetch instructions at an optimal point in each generated program, and instead always places the prefetch at a fixed, predictable location in the assembly template. The change is described by the author as an improvement/cleanup of RISC-V code. There is no indication in the commit or supplied references that this fixes a security vulnerability.
8ccf7de3 · Dec 31, 2025 · 10 filesInformational 15RISC-V: use vector hardware AES instead of scalar
Why it was queuedThis commit is a performance optimization for XMRig on RISC-V processors. It replaces a slower scalar (one-value-at-a-time) hardware AES implementation with a faster vector (multiple-values-at-once) AES implementation. There is no security vulnerability here; it is purely about making cryptocurrency mining faster on newer RISC-V CPUs.
d3a84c4b · Dec 29, 2025 · 6 filesInformational 15RISC-V: detect and use hardware AES
Why it was queuedThis commit adds support for RISC-V processors that have built-in AES hardware instructions. It is a performance optimization for the RandomX mining algorithm on RISC-V, not a security fix or vulnerability. There is no indication it introduces a security issue.
e1efd3dc · Dec 27, 2025 · 6 filesInformational 15RISC-V: auto-detect and use vector code for all RandomX AES functions
Why it was queuedThis commit is a performance optimization for XMRig on RISC-V processors. It adds new code paths that use RISC-V vector instructions (RVV) to speed up the AES-based hashing routines used by the RandomX mining algorithm. There is no indication this change fixes a security bug or introduces a known vulnerability; it is purely about making mining faster on RISC-V hardware.
f661e1eb · Dec 26, 2025 · 18 filesInformational 18RISC-V: vectorized RandomX main loop
Why it was queuedThis commit adds a new, faster RISC-V code path for the RandomX mining algorithm in XMRig. It is a performance optimization: it detects RISC-V vector (RVV) and optional bit-manipulation/prefetch extensions, then generates vectorized machine code for the main mining loop. There is no indication in the commit that it fixes a security bug or introduces a known vulnerability.
99488751 · Dec 23, 2025 · 1 fileInformational 15v6.25.1-dev
Why it was queuedThis commit only changes the software version number from 6.25.0 to 6.25.1-dev in a single header file. There are no functional code changes, bug fixes, or security-related modifications visible in the diff.
753859ca · Dec 23, 2025 · 1 fileInformational 15v6.25.0
Why it was queuedThis commit is a routine version bump from '6.25.0-dev' to '6.25.0' in a single header file. It changes only the displayed version string and contains no functional code changes, bug fixes, or security-related modifications.
712a5a5e · Dec 23, 2025 · 1000 filesInformational 15Merge branch 'dev'
Why it was queuedThis is a routine merge of the 'dev' branch into the main XMRig cryptocurrency mining project. The commit adds the entire project source tree (about 319,000 lines across roughly 1,000 files) to the repository. There is no indication of a security patch, vulnerability fix, or malicious change in the supplied materials.
290a0de6 · Dec 23, 2025 · 3 filesInformational 15v6.25.0-dev
Why it was queuedThis commit is a routine version bump for the next development release (6.25.0-dev). It only updates the version number, adds a changelog entry, and shortens the license header in one file. No code behavior changes and no security fixes or vulnerabilities are present.
482a1f0b · Dec 11, 2025 · 8 filesInformational 21Linux: added support for transparent huge pages
Why it was queuedThis commit adds Linux-only support for 'transparent huge pages' in the XMRig cryptocurrency miner. Transparent huge pages are a normal Linux performance feature that can speed up memory-heavy mining work. The change only calls the standard Linux madvise() system call with MADV_HUGEPAGE when regular huge-page allocation fails, and does nothing on Windows. There is no direct security vulnerability visible in the diff, but enabling huge pages can sometimes make certain advanced memory-based side-channel attacks slightly easier in very specific scenarios. That is a general property of huge pages, not a flaw introduced by this commit.
23da1a90 · Dec 5, 2025 · 5 filesInformational 15RISC-V: added vectorized soft AES
Why it was queuedThis commit adds a new, faster software-based AES implementation for RISC-V processors that support the RISC-V vector extension. It is a performance optimization for the RandomX mining algorithm and does not appear to fix or introduce a security vulnerability.
7ef5142a · Nov 30, 2025 · 9 filesInformational 14RISC-V: added vectorized dataset init (activated by setting `init-avx2` to 1 in config.json)
Why it was queuedThis commit adds a new optional performance optimization for RandomX mining on RISC-V processors that support vector instructions. It introduces a vectorized way to initialize the RandomX dataset, controlled by a config setting called `init-avx2`. There is no indication in the commit that this fixes a security bug or introduces a vulnerability; it appears to be a normal performance feature for a specific CPU architecture.
e88009d5 · Nov 12, 2025 · 983 filesInformational 15add detection for MSVC/2026
Why it was queuedThis commit is a massive initial import or full repository snapshot of the XMRig cryptocurrency miner project, not a focused security patch. The title mentions adding detection for MSVC/2026, but the diff shows 983 files being added including the entire codebase, build system, documentation, third-party libraries, and a Windows kernel driver. There is no evidence of a security vulnerability being fixed or introduced in this commit itself.
5115597e · Nov 6, 2025 · 2 filesInformational 17Improved compatibility for automatically enabling huge pages on Linux systems without NUMA support.
Why it was queuedThis commit is a routine compatibility improvement for XMRig's huge-page memory setup on Linux. It makes the miner better at enabling large memory pages on systems that do not expose per-CPU-node (NUMA) huge-page statistics. There is no indication this change introduces a security vulnerability or fixes one; it simply broadens the paths the program reads from and writes to in the Linux /sys filesystem.
b02519b9 · Nov 4, 2025 · 6 filesInformational 15feat: initial support for haiku
Why it was queuedThis commit adds build and runtime support for the Haiku operating system to the XMRig cryptocurrency miner. It is a straightforward portability patch: it teaches the build system to recognize Haiku, adjusts compiler flags, adds Haiku to existing OS-specific code branches for byte swapping and thread affinity, and provides Haiku-specific memory allocation fallbacks. There is no indication of any security vulnerability being fixed or introduced.
a44b21ce · Oct 27, 2025 · 3 filesInformational 15Cleanup
Why it was queuedThis commit is a routine cleanup with no apparent security relevance. It merges two separate conditions for ARM and RISC-V into one, fixes a punctuation typo in the README, and removes the HAVE_ROTR compiler flag from ARM builds. There is no indication of a vulnerability being fixed.