| Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Plan-only document covering the ~230 Linux syscalls ior does not yet
trace. For each not-yet-covered syscall the plan records whether it
takes/returns a file descriptor, whether it has a meaningful bytes or
size dimension that could feed the existing read/write/transfer
accumulators, the suggested TracepointKind (existing or new), extra
per-event metrics, and a priority tier. Also enumerates the taxonomy
extensions required (new TracepointKind values, retClassification
additions, family/dimension tags), risk areas (high-cardinality
syscalls like futex/clock_gettime), and a phased rollout from "free
wins" (network bytes already mapped but blocked by shouldIgnore)
through process lifecycle to the long tail.
No code changes.
|
|
Change tabDescriptor.InitCmd from func() tea.Cmd to func(*Model) tea.Cmd
so the initial tick for the flame and stream tabs uses the model's
fastRefreshEvery interval rather than hardcoded 200 ms constants.
Both call sites in Init() and postKeyTransitionCmd() now pass the model
pointer. The now-unused streamTickCmdFn/flameTickCmdFn package-level
adapters are removed; the TabFlame and TabStream registry entries use
inline closures that delegate to m.flameTickCmd()/m.streamTickCmd().
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
Add a bullet explaining the default (250ms), what it controls (flamegraph
and stream tab refresh cadence), and that setting it to 0 disables
high-frequency refresh.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
Add fastRefreshMs parameter to NewModelWithConfig so callers can supply
the high-frequency tick cadence for stream and flame tabs. Convert the
streamTickCmd/flameTickCmd package-level functions to model methods that
honour fastRefreshEvery (falling back to the 200 ms constants when zero
for backward-compatibility). Add SetFastRefreshInterval setter so
RunWithTraceStarterConfig can apply cfg.TUIFastRefreshInterval after
construction. Update all 68 test call sites to pass fastRefreshMs=200
and add three new tests covering zero-fallback, stored value, and setter
behaviour.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
Add TUIFastRefreshInterval time.Duration to flags.Config struct with a
default of 250ms and register it as the -tui-fast-refresh CLI flag so
callers can tune or disable the high-frequency TUI refresh cadence for
the flamegraph and stream tabs.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
Split 22 production files across the codebase — event loop, TUI models,
probe manager, dashboard, export, flag parsing, code generation, and
ioworkload scenarios — so that no function body exceeds 50 lines. Each
extracted helper carries its own comment explaining its role.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
The dashboard model's View() was mutating sub-model state on every render:
it called streamModel.SetFooterVisible() and flameModel.SetViewport() on
local copies instead of keeping those fields in sync through Update().
Moved the sync points to the Update() handlers that trigger each change:
- handleWindowSize: syncs stream footer visibility alongside SetViewport
- handleHelpToggleKey: already updated flame viewport; now also updates
stream footer visibility when the help bar is toggled
- handleKey: syncs flame viewport when switching to the flame tab, covering
the case that window-resize and help-toggle do not
Aligned the stream model's initial showFooter value with the dashboard's
default showHelp=false in NewModelWithConfig so there is no mismatch on
the first render.
Also extracted numeric tab shortcuts into the tab registry via ShortcutKey
fields so handleShortcutKey no longer needs updating when tabs are added.
Updated two tests that bypassed Update() to set dimensions directly; they
now use WindowSizeMsg so sub-model viewports are initialised correctly
under the new pure-View contract.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
The printCb and warningCb function fields on eventLoop bundled two
distinct concerns (pair emission and warning delivery) directly on the
event-processing struct. This commit extracts them into a dedicated
outputFormatter type that owns these callbacks plus emit() and
notifyWarning() helper methods.
outputFormatter is embedded (not pointed-to) in eventLoop so that
existing call sites — including tests that write el.printCb = ... and
el.warningCb = ... directly — require no changes beyond the three
struct-literal sites in eventloop_filter_test.go that used field
initialiser syntax.
fdTracker and commResolver were already proper collaborator types; only
the output concern needed extraction.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
snapshot-building
Define statsengine.Accumulator (Ingest + Reset) to represent the
event-accumulation responsibility separately from runtime.SnapshotSource
(Snapshot), which handles the read side. This reduces the SRP violation in
Engine: callers that only push events now hold an Accumulator; callers that only
read statistics hold a SnapshotSource.
- Add Accumulator interface and compile-time assertion in statsengine/engine.go
- Add EventIngester type alias (= statsengine.Accumulator) in runtime/runtime.go
with a compile-time assertion, so callers in the runtime layer can reference
the ingestion contract without importing statsengine directly
- Split tuiRuntime.engine field into accumulator + snapSource so the event-loop
callback holds Accumulator and wireRuntimeBindings passes SnapshotSource to
SetDashboardSnapshotSource — making each consumer's dependency explicit
- Simplify resetDashboardSnapshotSource in tui.go to cast for interface{ Reset() }
independently of Snapshot(), removing the combined ad-hoc interface check
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
setupEventChannel now returns the *bpf.RingBuffer handle alongside the
event channel. Both callers (runTraceWithContext, runHeadlessParquet)
defer rb.Stop() so the libbpfgo polling goroutine and C ring_buffer
struct are released promptly when the trace context is cancelled,
rather than waiting for bpfModule.Close() to eventually call rb.Close().
rb.Stop() and rb.Close() are both idempotent, so the double-call from
the defer and from bpfModule.Close() is safe.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
- Add tracepoints.Selector type with ShouldAttach method and ParseSelector
constructor, replacing the raw TracepointsToAttach/TracepointsToExclude
regex slices on flags.Config.
- Add flags.BuildTraceFilter as a standalone function replacing the
Config.TraceFilter() method, keeping filter-building logic out of the
config struct.
- Remove stale ShouldIAttachTracepoint noise-filter entry from Magefile.
- Add selector_test.go with full coverage of ParseSelector and ShouldAttach.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
The testRace and goBuildRace targets already existed in Magefile.go
(added in commit 70acff7) but were missing from the build/test command
reference in AGENTS.md. Add the testRace entry so contributors know
how to run the full test suite under Go's race detector.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
Four near-identical sorted-rows functions (sortedFileSnapshots,
sortedDirSnapshots, sortedSyscallSnapshots, sortedProcessTableRows)
each repeated the same guard-clone-sort-tiebreak-apply pattern.
Replace them with a single generic sortedWithState[T,K] in sort.go
that accepts per-tab byKey and tiebreak comparators as parameters.
Remove now-unused slices imports from syscalls.go and processes.go.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
fileRankHeap Len/Less/Swap converted to pointer receivers to match
Push/Pop; bubbleChart HasNodes and AnimationState Settled converted
to pointer receivers to match all other methods on their types.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
Replace sync.WaitGroup with errgroup.Group in buildSubSnapshots so errors from
sub-builders (buildSyscallSnapshots, buildFileSnapshots, buildProcessSnapshots,
buildHistogramSnapshot) are captured and propagated rather than silently dropped.
Change Engine.Snapshot() to return (*Snapshot, error), update runtime.SnapshotSource
and dashboard.SnapshotSource interfaces accordingly, and adjust all callers in
tui.go, dashboard/model.go, and the test helpers.
Each sub-builder now returns (result, error); the error return is currently
always nil but establishes the contract for future validation. The per-type
Snapshot() convenience methods (histogram, syscall, file, process) panic on
error since they are internal helpers where failure would be a programming bug.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
paths
Swap out ad-hoc += string concatenation in the flamegraph toolbar/status
lines, dashboard filter summary, bubble/treemap status lines, eventstream
view, processes tab, and probes list for strings.Builder, eliminating
redundant allocations on every render tick. Also update dashboard/model_test.go
fake SnapshotSource implementations to match the updated interface signature.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
|
Introduce kindregistry.go with a kindMeta struct (structName, enterAccepted)
and a kindRegistry map keyed by TracepointKind. Replace the switch in
isEnterRejected (codegen.go) and the switch in eventStructName (bpfhandler.go)
with lookupKind registry lookups. Adding a new TracepointKind now only
requires a single registry entry — no switch statements need to be touched.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|