R1 — input latency: the event loop drained input once per iteration, then
slept the whole metrics interval; keys and wheel events queued for up to
500ms (or the full interval at slower rates) and applied in bursts. The
input block is extracted to drain_input() and the tail sleep replaced by
a deadline wait in <=33ms poll slices that handles and repaints input the
moment it arrives. Verified: help modal opens <150ms into a 2000ms tick.
R2 — freeze-proofing: metrics/processes/disks requests had no timeout; a
half-dead connection left ws.next() pending forever and froze the TUI
with no way to quit (raw mode eats Ctrl+C as an unread key event). All
requests now carry a 5s budget.
C3 — desync: replies are matched to requests by order alone, so a timed-
out request's late reply would shift every subsequent reply off by one.
Any timeout now treats the stream as poisoned and goes through the
reconnect flow — a fresh stream is aligned by construction. The modal
endpoints additionally mark process details unsupported (flag resets on
modal close/selection change) so a detail-less agent doesn't cause a
reconnect loop. While disconnected the fetch path idles: recovery belongs
to the manual/auto retry paths instead of 5s-timeout hammering.
C7 — fit::truncate_middle_cols replaces util::truncate_middle: display-
width aware and char-boundary safe; the byte-slicing version panicked the
draw loop on non-ASCII device names.
Verified live: agent kill -9 mid-session -> error modal in <3s, q exits
while disconnected, r reconnects and resumes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Lightweight:
- GPU collection moves to a dedicated worker thread that owns the gfxinfo
handle for the process lifetime. gfxinfo's active_gpu() runs a full NVML
init/teardown (~20ms, blocking) and we were paying it on the async
runtime for every collect — measured at ~80% of the agent's entire
active CPU on a GPU machine. The handle holds Rc<Nvml> (not Send), so a
thread + mpsc/oneshot channel pair confines it; a zero-total-VRAM reply
is treated as a dead session (driver reload) and re-probed.
- journalctl now runs via tokio::process instead of blocking one of the
two runtime workers for the duration of the subprocess.
- TtlCell (state.rs) replaces the four hand-rolled static TTL caches; a
cached negative result now counts as fresh, so hosts with no matching
temp sensor or GPU stop rescanning every request. Single lock+clone on
the GPU cache hit path (was two).
Correctness:
- Process/child CPU times are now microseconds as documented; they were
milliseconds, rendering 1000x too small next to (correct) thread times.
- Non-Linux per-process CPU%% clamps AFTER dividing by core count; a
4-cores-busy process on an 8-core box reported 12.5% instead of 50%.
- Journal timestamps are real RFC 3339 UTC plus an additive timestamp_us
field (sorting is now numeric); the old strings were Debug-formatted
SystemTime mangled by string replace.
- Partition detection uses /sys/block on Linux: whole-disk filesystems on
names like nvme0n1 or zram1 are no longer misclassified as partitions.
One shared parent_disk_name() replaces two inline copies.
- New sampled_at_ms on the metrics payload (additive) records when the
snapshot was actually collected, so clients can compute exact rates
across the agent's TTL cache.
Security/robustness:
- key.pem is created 0600 (was umask default 0644, world-readable) and
pre-1.51 keys are tightened on startup.
- Per-PID detail/journal caches now evict (60s max age, 64 entries max);
they previously grew without bound under PID-walking clients.
- The two per-PID ws handlers collapse into one generic helper.
- /proc/<pid>/stat parsing unified in one comm-safe module.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Security: with --verify-hostname off (the default), the old NoVerify
verifier accepted ANY server certificate — the CA loaded from --tls-ca
was never consulted, so the documented pinning was a no-op and the
connection was trivially MITM-able. Replace it with PinnedCertVerifier:
the presented end-entity cert must be byte-identical to a cert in the
--tls-ca file (any cert in a multi-cert PEM matches, supporting
rotation). Signature validation now uses the ring provider's full
algorithm set instead of a hardcoded 3-scheme list. Empty PEM files
fail fast instead of failing closed per-handshake.
The --verify-hostname path is unchanged (WebPki root-store validation).
Also: the third argument of connect_async_tls_with_config is
tungstenite's disable_nagle flag, not a verification toggle — we were
passing verify_hostname there, leaving Nagle ON for default users. Pass
true unconditionally, and disable Nagle on the plain ws:// path too;
socktop exchanges small request/response frames where Nagle only adds
latency.
Client now consumes the connector via a dual path+version dep so these
fixes are in local builds and CI before the crates.io publish (cargo
strips the path on publish). Connector version -> 1.51.0.
Verified E2E: agent A's cert connects to agent A; agent B's cert
against agent A fails the handshake (the rpi-worker-1 wrong-PEM
scenario); --verify-hostname against a 127.0.0.1 SAN still connects.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Delete socktop_connector/src/connector.rs: orphaned since 08f248c removed
'pub mod connector;' during the modularization refactor. Never compiled
(verified under default, wasm, and workspace feature combos) but shipped
in the crates.io tarball and contained an outdated copy of the TLS
verifier — a trap for anyone patching the pinning bug in the dead copy.
- Delete empty socktop/src/ws.rs, tracked editor backup ui/.modal.rs.backup,
and stray test_thiserror.rs at the repo root.
- Delete the two LEGACY #[allow(dead_code)] process input handlers; the
header-click render test now exercises the live _with_selection handler
instead (better coverage of the real path).
- Drop unused sysinfo dependency from the socktop client.
- Replace stale 'temporarily increased for testing' comment on
COMPRESSION_THRESHOLD (it already held the production value).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three panes painted two independent pieces of text onto one row with nothing
reserving space between them, so below roughly 105 columns the right-hand piece
was simply drawn over the tail of the left one:
socktop — host: cachyos-gaming | 🔒✗ TLS | (a: about⏱ 500ms metrics | 2000ms
The width arithmetic used str::len(), a byte count, so the emoji in these strings
overstated their width and left orphaned glyphs at the right edge as well. The
process table had the same problem in a different form: it handed the layout
solver a fixed, over-constrained column set, so a narrow pane crushed the
percentage-sized Name column to nothing while fixed-width PID and Mem % kept
their full width — losing the one field that identifies a process.
Add ui::fit (measure in terminal columns, truncate on character boundaries, pick
the richest wording that fits) and give each pane a priority ladder:
- Header: drop the key hints, then the TLS/token badges, then the
"socktop — host:" prefix, then the metrics/procs words, and only then shorten
the hostname. Hostname and intervals are what survive longest.
- CPU pane: drop the "CPU Temp:" label, then the now:/avg: labels, then the
average, then the temperature's decimal, and the temperature itself last — a
thermal warning outranks a second decimal place.
- Process table: Name is unconditional; CPU %, then Mem, then PID, then Mem %
are added as the pane widens, so Mem % is the first to go and Name the last.
Also fixes sort-header clicks, which resolved against a Layout that omitted the
column spacing the Table renders with, so a click landed off by up to four
columns. Covered by a test that clicks each label where it is actually drawn.
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
In a short window the fixed root layout runs out of rows and the CPU graph and
per-core bars are what collapse first: the header, gauges and process table hold
fixed heights, so at ~18 rows the top row is left with no drawable interior and
both panes disappear entirely.
Add a second layout, entered automatically once the Disks pane can no longer show
even one complete disk card:
- Disks is dropped — it is the pane that degrades worst when partly drawn.
- Memory and Swap move side by side into the row Disks vacated.
- GPU collapses to a single full-width line (utilisation and VRAM, no device
name), and is omitted entirely on a host that reports no GPU.
- The reclaimed rows go to the CPU panes, with the surplus above their floor
shared with the bottom half so the process table still grows with the window.
`--compact` pins the layout at any size.
The root layout was duplicated in three places (the draw path and both input
hit-testing paths), which would have drifted the moment a second layout existed.
Move it into ui::layout as the single source of truth and have all three callers
go through it.
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Demo mode spawns a socktop_agent child process, but the agent is a separate
crate that `cargo install socktop` does not pull in. When it was missing, the
raw spawn error propagated to main and printed as
`Error: Os { code: 2, kind: NotFound, message: "No such file or directory" }`,
which gives the user nothing to act on.
Introduce DemoAgentError so a NotFound spawn failure is distinguishable from
other io errors, and print the path we looked for plus the `cargo install
socktop_agent` fix. Other spawn errors still propagate as before.
Co-authored-by: Jason Witty <jason@localhost-live.localdomain>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Resolves all five open Dependabot alerts (GHSA-9f94-5g5w-gf6r,
GHSA-394x-vwmw-crm3, GHSA-hfpc-8r3f-gw53, GHSA-65p9-r9h6-22vj,
GHSA-vw5v-4f2q-w9xf) by moving aws-lc-sys from 0.33.0 to 0.42.0.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
The Windows CI matrix surfaced an `unused_variables` warning at
metrics.rs:846 — `let now = std::time::Instant::now();` was bound
unconditionally but only consumed inside a `#[cfg(feature = "logging")]`
tracing::debug! call.
This block lives in the non-Linux `collect_processes_all`, so the Linux
CI never compiles it and never sees the warning. Same shape as the
`processes_ttl_ms` Windows fix from earlier: a binding whose only
consumer is cfg-gated needs to be cfg-gated too.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Add Debian packaging support with cargo-deb
- Add cargo-deb metadata to socktop and socktop_agent Cargo.toml
- Create systemd service file for socktop_agent
- Add postinst/postrm maintainer scripts for user/group management
- Create GitHub Actions workflow to build .deb packages for AMD64 and ARM64
- Add comprehensive documentation in docs/DEBIAN_PACKAGING.md
- Packages will be available as artifacts on every push
- Automatic GitHub releases for version tags
* Add summary documentation for debian packaging
* fix unit test, move to macro cargo_bin!
* hotfix for issue with socktop agent not creating ssl certificate on first launch after upgrade of axum server version.
* Add helpful post-install message to guide users on enabling socktop-agent service
* Fix CI build by installing libdrm development dependencies
* Fix package rename script - cargo-deb already includes architecture in filename
* Make GPU support optional to enable RISC-V builds without libdrm
- Add 'gpu' feature flag (enabled by default)
- Make gfxinfo dependency optional
- Provide no-op GPU metrics when gpu feature disabled
- Disable GPU support for RISC-V builds in CI (libdrm unavailable)
- All other architectures (amd64, arm64, armhf) still get GPU support
* feature gate GPU stats for arm v7
* specify correct package names.
* install aarch64-linux-gnu-gcc build dep
* specify correct package name
* add RISC-V GCC compiler
* add .cargo to gitignore to elimicate issue with riscv64-linux-gnu-gcc linker in config.toml
* add gcc-arm-linux-gnueabihf linker fore armv7
* set correct x-compile lib gcc-aarch64-linux-gnu for arm64 builds.
* add ports.ubuntu.com to sources
* Add ARM64 as a foreign architecture
* fixe for ARM64 build.
* security.ubuntu.com` aNNOYING
* apt repo github page
* copy output to apt repo
* fix secrets path
* fix secrets path
* change build dep
* Fix postinst message box alignment
* ci(deb): restrict APT publish to v* release tags
Previously the workflow built and published on every push to master
and feature/debian-packaging in addition to v* tags. That meant the
gh-pages APT repo got overwritten on every commit with same-version
.debs, causing apt clients to see a phantom "update available" each
time and burning ~5-10 min of cross-compile CI per merge.
After this change:
- PRs into master still cross-build .debs as a sanity check.
- v* tags build, publish to gh-pages, and create a GitHub release.
- workflow_dispatch remains as the manual escape hatch.
- master pushes no longer trigger this workflow (ci.yml still runs).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore: update ratatui from 0.28 to 0.30
* style: cargo fmt
* fix: replace manual zero-guarded divisions with checked_div
* fix: collapse nested if into match guard
* style: cargo fmt
* bump crossterm and optimize various types, remove stale code.
* fix windows build
* only show parent level processes on main tui
- Add Up/Down arrow key handling in help modal
- Display scrollbar when content exceeds viewport
- Update title to indicate scrollability
- Fixes content cutoff on small terminal windows
This commit implements several major improvements to the TUI experience:
1. CPU Average Display in Main Window
- Show average CPU usage over monitoring period alongside current value
- Format: "CPU avg (now: 45.2% | avg: 52.3%)"
- Helps identify sustained vs momentary CPU spikes
2. Max Memory Tracking in Process Details Modal
- Track and display peak memory usage since monitoring started
- Shown as "Max Memory: 67.8 MB" in yellow for emphasis
- Helps identify memory leaks and usage patterns
- Resets when switching to different process
3. Fuzzy Process Search
- Press / to activate search mode with bordered search box
- Type to fuzzy-match process names (case-insensitive)
- Press Enter to auto-select first result
- Navigate results with arrow keys while typing
- Press c to clear filter
- Press / again to edit existing search
Search box features:
- Yellow bordered box for high visibility
- Active mode: "Search: query_"
- Filter mode: "Filter: query (press / to edit, c to clear)"
Technical implementation:
- Centralized filtering with get_filtered_sorted_indices()
- Consistent filtering across display, navigation, mouse, and auto-scroll
- Proper content area offset calculation for search box
- Real-time filtering as user types
4. Code Quality Improvements
- Created ProcessDisplayParams and ProcessKeyParams structs
- Created MemoryIoParams struct for process modal rendering
- Reduced function arguments to stay under clippy limits
- Exported get_filtered_sorted_indices for reuse
Files Modified:
- socktop/src/app.rs: Search state, auto-scroll with filtering, max memory tracking
- socktop/src/ui/cpu.rs: CPU average calculation and display
- socktop/src/ui/processes.rs: Fuzzy search, filtering, parameter structs
- socktop/src/ui/modal.rs: Updated help modal with new shortcuts
- socktop/src/ui/modal_process.rs: Max memory display, MemoryIoParams struct
- socktop/src/ui/modal_types.rs: Added max_mem_bytes field
Testing:
- All tests pass
- No clippy warnings
- Cargo fmt applied
- Tested search, navigation, mouse clicks, and auto-scroll
- Verified on both filtered and unfiltered process lists
Breaking Changes:
- None (all changes are additive features)
Closes: (performance monitoring improvements)
This check in offers alpha support for per process metrics, you can view threads, process CPU usage over time, IO, memory, CPU time, parent process, command, uptime and journal entries. This is unfinished but all major functionality is available and I wanted to make it available to feedback and testing.
jasonwitty
jason