Conversation
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2af2694f1f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 260d33549d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 260d33549d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 98b5d92d8d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| rzb_log(LOG_ERR, LOG_C_CNC, | ||
| "%s: Legacy dispatcher C&C runtime was removed for dispatcher-next; " | ||
| "use RzbNextRuntime instead", __func__); | ||
| return false; |
There was a problem hiding this comment.
Allow rzb-worker to initialize normal nuggets
When rzb-worker reaches module.initNug(), a typical nugget creates its normal inspection/collection context without CONTEXT_FLAG_DEV_TOOL; that path calls CommandAndControl_Start(), and this new unconditional failure for every non-dev context makes Razorback_Init_*_Context() return NULL, causing initNug() and the worker startup to fail before any dispatcher-next runtime handoff exists. Either keep a working non-dev startup path here or have rzb-worker initialize through RzbNextRuntime before rejecting legacy C&C.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 291c9842d0
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| goto cleanup; | ||
| } | ||
|
|
||
| (void)Block_MetaData_Add_FileName(inputItem->pEvent->pBlock, filePath); |
There was a problem hiding this comment.
Add filename metadata with a NUL-terminated helper
When rzb-dev prepares the local inspection block, this call adds filename metadata through Block_MetaData_Add_FileName(), whose current implementation stores only strlen(fileName) bytes even though it labels the value as NTLV_TYPE_STRING; unlike Metadata_Add_Filename(), the copied data is not NUL-terminated. Local inspections therefore receive malformed filename string metadata, and nuggets that read it as the existing string metadata convention permits can read past the buffer or see garbage. Use the metadata helper that includes the terminator, or fix the block helper before feeding this metadata to the nugget.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d3408249f4
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| block->data.fileName = fileName; | ||
| block->data.tempFile = tempFile; | ||
| block->data.pointer = mmap(NULL, block->pId->iLength, PROT_READ, MAP_PRIVATE, | ||
| fileno(block->data.file), 0); | ||
| if (block->data.pointer == MAP_FAILED) { |
There was a problem hiding this comment.
Clear block file state when mmap fails
If mmap fails after block->data.fileName has been assigned, this returns false with the block still pointing at that filename even though the FILE has been closed. The local-dev caller now has a special cleanup path, but the public helper is also used by RzbNextFileserver_FetchBlock, which frees the filename on failure; callers that then retry, inspect, or call RzbNextFileserver_FreeBlockData on the same block can hit a dangling pointer/double free. Move these assignments until after the map succeeds or clear fileName/tempFile on this failure path.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9a7ff0fbf8
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| bool live = Health_EvaluateInternal(RAZORBACK_HEALTH_LIVE); | ||
| bool ready = Health_EvaluateInternal(RAZORBACK_HEALTH_READY); | ||
| bool startup = Health_EvaluateInternal(RAZORBACK_HEALTH_STARTUP); |
There was a problem hiding this comment.
Keep probe endpoints from running unrelated checks
For a /livez, /readyz, or /startupz request this now evaluates all three probe kinds to render the JSON body, and Health_EvaluateInternal() runs the registered callbacks for each kind. A slow dependency/readiness callback can therefore block liveness probes (and the requested kind is evaluated twice), causing probe timeouts or side effects even though the endpoint status only depends on kind. Render these fields from cached/built-in state or avoid invoking unrelated custom checks on individual probe endpoints.
Useful? React with 👍 / 👎.
| <ClCompile Include="src\connected_entity.c" /> | ||
| <ClCompile Include="src\crypto.c" /> | ||
| <ClCompile Include="src\daemon.c" /> | ||
| <ClCompile Include="src\dev_mode.c" /> |
There was a problem hiding this comment.
Compile dispatcher-next sources in the MSVC project
When building with api.vcxproj, this adds only dev_mode.c; the same commit exposes public dispatcher-next APIs but this project still omits their implementations (config_next.c, fileserver.c, message_body.c, messages_next.c, and runtime_next.c) that are included in the autotools build. MSVC builds will produce a library without those new symbols, so Windows consumers using the new headers get unresolved externals. Add the same new sources to the project and filters.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a98d6cc5a6
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| blockId->pHash->iType == HASH_TYPE_SHA256 && | ||
| blockId->pHash->iSize == 32 && | ||
| (blockId->pHash->iFlags & HASH_FLAG_FINAL) != 0 && | ||
| blockId->iLength > 0; |
There was a problem hiding this comment.
Allow zero-length SHA-256 blocks
When a collector submits an empty file via BlockPool_AddData_FromFile, BlockPool_FinalizeItem can still produce a finalized SHA-256 block with iLength == 0, but all of the new fileserver store/fetch paths go through this helper and return RZB_NEXT_FILESERVER_LOCAL_ERROR before building the canonical ... .0 path or making the HTTP request. That makes dispatcher-next reject valid empty-file submissions that the previous file transfer path could store, so the identity check should not require a positive length (with zero-length mmap handling addressed separately).
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1c200167ae
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| return; | ||
|
|
||
| Mutex_Lock(sg_devCaptureLock); | ||
| (void)List_Remove(sg_devCaptures, context); |
There was a problem hiding this comment.
Remove dev captures using the capture object
When a dev-mode context is destroyed, this passes the RazorbackContext * as the list item to remove even though sg_devCaptures stores struct RazorbackDevContextCapture *; List_Remove() compares with the item comparator rather than DevCapture_KeyCmp, so DevCapture_Cmp interprets the context memory as a capture and normally never removes the real entry. After shutdownNug()/context teardown in rzb-dev or in tests that create dev contexts repeatedly, the stale capture retains old judgments/submissions and a freed context pointer, so a later context can leak capture state or reuse stale results if the address is recycled.
Useful? React with 👍 / 👎.
No description provided.