Repository navigation
Releases
Releases · Fahry-a/ODM
v1.9.2
1.9.2
Changed
- Foenui PM release mirror — the release workflow publishes all six cross-compiled ODM archives to the personal Foenui PM namespace (fahry-a/odm), providing a real end-to-end test of package upload during a release.
v1.9.1
1.9.1
Changed
- Download engine hygiene — filename helpers (
uniqueName,
sizeOrUnknown,formatSegSize,flattenFilename,deriveFilename)
extracted fromtask.gointotask_filename.go(move-only, no behavior
change); duplicate doc comments left over from thetask.gosplit cleaned
up. (internal/download) - Faster engine tests — test-only timeout/delay trims (
Timeout30s→10s,
work-stealing throttle 120ms→60ms).TestWorkStealingnow runs in ~0.6s
and stays stable across repeated-raceruns; no production code touched.
(internal/download)
v1.9.0
1.9.0
Fixed
- RPC shutdown returns fake task ID —
Scheduler.Enqueuenow returns an
error when the scheduler is winding down or the pending queue is full (128
task cap), soDaemon.AddURLpropagates the failure to the RPC layer instead
of returning a task ID that will never complete. (internal/scheduler/queue.go,
internal/scheduler/daemon.go) - Config validation bypassed by config-file values —
Validate()checked
IsSet()(CLI-explicit only) for profile-specific rejections, allowing
config-file values to bypass semantic rules. Now validates final merged
values against defaults. (internal/config/config.go) - Negative
--split-fileaccepted — values like-sf -1silently passed
validation. Now rejected. (internal/config/config.go) --rpc-listen-allwithout authentication — binding to 0.0.0.0 with no
--rpc-secretor TLS was only warned, not blocked. Now rejected.
(internal/config/config.go)- Known-size single-stream short read — a single-stream download could
reach EOF before the probedTotalSizeand still succeed. Now detects the
mismatch. (internal/download/task_io.go) - Global 429 cooldown lived per-task — the adaptive 429 back-off timestamp
was on eachTask, allowing one task's success to prematurely restore the
shared rate. Now consolidated intoratelimit.Limiter. (internal/ratelimit/bucket.go)
Changed
- Dynamic connection reallocation — when a task completes, its connection
budget is redistributed to remaining tasks instead of sitting idle. The
scheduler now callsAdjustConnsat runtime. (internal/scheduler/)
Security
- WebSocket resource limits — inbound message size capped at 1 MiB,
read/write deadlines enforced (60s/10s), ping/pong keep-alive added.
(internal/rpc/ws.go) - Rate parser overflow protection — rejects NaN, Inf, negative rates, and
values exceedingint64or platformintcapacity.SetRatevalidates
before mutating. (internal/ratelimit/bucket.go) - CI self-modifying workflow removed — workflow no longer requests
contents: writeor pushes from CI runners. (.github/workflows/)
v1.8.1
1.8.1
Fixed
- Summary line loses speed/ETA on narrow terminals — the compact fallback
mode usedBarWidth(30) instead of the adaptivebarWfromlayoutFor,
causing padding to go negative and triggering aggressive degradation that
dropped speed/ETA entirely. The compact tier now uses the adaptive bar width,
keeps speed/ETA when they fit, and right-aligns like the per-file task lines.
(internal/ui/summary.go)
v1.8.0
1.8.0
Changed
odm updatedelegates to install.sh via subprocess — instead of downloading
and replacing the binary internally (which fails on Termux due to TLS/DNS issues),
odm updatenow downloads install.sh and executes it in update mode. This unifies
the update path across all non-AUR platforms and leverages curl (system DNS/certs).
(internal/update/update.go)CheckLatest()uses curl instead of Go HTTP client — avoids TLS certificate
verification failures caused by Go's pure-Go DNS resolver on Android/Termux.
(internal/update/update.go)
Added
--updateflag andODM_UPDATEenv var for install.sh — install.sh now
accepts--updateflag (orODM_UPDATE=1env var for curl|sh pipeline fallback)
to enable update mode: auto-proceeds past confirmation, shows "Updating" messages
instead of "Installing", skips config install and PATH warnings.
(docs/public/install.sh)
Fixed
- Progress bar invisible until Enter/new line and stacked lines — when stdout
is block-buffered (piped, wrapped, or fully buffered stdio) the pacman bar's
writes stayed invisible until a newline flushed the buffer, causing stacked
37.3K/?lines and requiring Enter to appear. The renderer now flushes after
every frame (Begin/End/Frame/Interject) and prefers the TTY writer
(stderrwhenstdoutis piped) so the bar is live immediately.
(internal/ui/progress.go,cmd/odm/main.go) - Temp directory resolution for Termux —
runInstallScript()now resolves the
temp directory via$TMPDIR→os.TempDir()→~/.odm-tmpfallback chain,
avoiding Termux's/tmpsymlink which may not exist or be writable.
(internal/update/update.go)
v1.7.3
1.7.3
Fixed
- Skip interactive prompt when install script is piped — when
odm updateis
invoked viacurl|sh, stdin is not a tty; the[y/N]prompt now auto-proceeds
in that case with an informational message. (docs/public/install.sh) - Force Cloudflare DNS (1.1.1.1) in update client for Android/Termux — Go's
pure-Go DNS resolver on Android tries[::1]:53(IPv6 localhost) which does not
exist, causingconnection refusedwhen calling the GitHub API; the API client
now always dials Cloudflare over UDP. (internal/update/update.go) - Isolate failed chunk progress rollback — under concurrent multi-worker
downloads, a failing chunk could subtract bytes written by other workers via the
shared atomic counter; rollback now uses a per-attempt byte tracker to prevent
progress corruption. (internal/download/task_io.go)
v1.7.2
1.7.2
Fixed
- Install script: handle
armv8larchitecture — Android/Termux devices with
32-bit mode on ARMv8 CPUs (e.g. Samsung A13) reportedarmv8lfrom
uname -m, which was unrecognized. Now maps toarm. (docs/public/install.sh) - Install script: checksum verification filename mismatch — tarball was saved
asodm.tar.gzbutchecksums.txtexpectsodm_VERSION_OS_ARCH.tar.gz,
causingsha256sum -cto fail. (docs/public/install.sh) odm updatecrash on Termux/Android —isAUR()called
exec.LookPath("pacman")which triggersfaccessat(AT_SYMLINK_NOFOLLOW),
a syscall blocked by Android's seccomp sandbox (SIGSYS). Now checks
pacmanexistence viaos.Statfirst. (internal/update)
v1.7.1
1.7.1
Added
odm updatesubcommand — self-update that auto-detects the install method
(AUR viapacman -Qi, self-installed via install.sh, or manual) and dispatches
to the appropriate update path: AUR prompts for yay/paru, self-installed
downloads + replaces the binary in-place, manual prints instructions.
(internal/update,cmd/odm)- Version hint in
odm -h— when a newer release exists on GitHub, the help
output shows→ update available: vX.Y.Z (odm update). Cached 24h via
~/.config/odm/.last-update-checkto avoid hitting the API on every--help.
(internal/update,cmd/odm) - One-line install script (
curl -fsSL https://odm.orynix.id/install.sh | sh)
— auto-detects prefix (writable/usr/local→ system-wide, else~/.local),
installs binary + man page + config, verifies checksum. (docs/public/install.sh)
Changed
- README rewritten — cleaner structure with features overview, one-line
install, installation guides (AUR, pre-built binaries, source, systemd),
and a more scannable flag reference table. (README.md)
Fixed
- Duplicate doc comments in
task_io.goandtask_checksum.go(merge artifact
from thetask.gosplit into 6 files). (internal/download)
Contributors
v1.7.0
1.7.0
Fixed
bothprofile: region1 now really runs on HTTP/1.1 — the h2-enabled
client was handed to the whole task, so region1's work-stealing engine
collapsed onto one multiplexed TCP stream and silently wasted half the
connection budget. Client routing is now by engine kind: static-split
(aria2c-model) engines ride h2 when available; fixed-chunk work-stealing
engines always keep N separate h1 connections — including degraded
single-region layouts whose profile string still readsboth.
(internal/download)- UI data race between engine logs and the redraw loop — log lines
interjected during a live frame could read torn renderer state
(cur.live/startedAt) written unlocked by the frame loop; under the Go
memory model this is a real race even if it never visibly corrupted a
frame. All last-seen state is now written only under the renderer lock.
(internal/ui) - aria2c display cap race —
Startpublished the raw connection budget
before capping it to the segment count, so UI/RPC pollers could briefly
see e.g.[x16]for a 4-segment file; the cap is computed first and
stored once. Also de-flakedTestAria_ConnsDisplayCappedToSplit, which
sampled that window ~50% of the time locally. (internal/download)
Changed
- Development workflow is now TDD (red → green → refactor), documented
in AGENTS.md. - Housekeeping: removed dead code (
ExecOptions.MaxConn,
PermanentErroralias, a duplicated Content-Range verification block),
resume failures now warn once per task instead of being swallowed, and
task.gosplit its resume/checkpoint half intoresume_impl.go
(movement only).
Fixed
Results of audit round 2 (1 critical, 4 high — all fixed with regression
tests):
- Path traversal via server-controlled filename — a Content-Disposition
or URL basename containing../(or separators) reachedfilepath.Join
raw, so a hostile server (or an RPC client on--rpc-listen-all) could
write the download body anywhere the process can — including dotfiles in
the user's home. Filenames are flattened to a single path component before
joining; explicit-ooverrides are untouched. (internal/download) - RPC
odm.shutdownbypassed auth — shutdown detection ran on the raw
request while auth was checked inside dispatch, so an unauthenticated POST
got an auth-error response and still killed the daemon, cancelling all
in-flight downloads. Detection is now gated on the same secret check.
(internal/rpc) - RPC
changeOption connections ≤ 0caused silent data loss — target 0
retired every worker via the graceful drain while chunks remained queued;
Start reported the half-downloaded file as completed and deleted its
control file, making resume impossible. Values < 1 are rejected at the RPC
boundary and clamped inAdjustConns. (internal/rpc,internal/download) - Negative
--retryfaked success — a config file withretry=-1made
the per-chunk attempt loop run zero iterations: chunks "succeeded" without
a single byte downloaded and control files were deleted. Rejected by
config validation; clamped defensively inNewTask.
(internal/config,internal/download) - Last-completion tally race in the scheduler — completion reports post
to the scheduler channel after the WaitGroup release, so when the final
tasks finished back-to-back, Run could return on its done path with one
report still buffered — wrong summary line, wrong exit code, missing RPC
completion event. The done path now drains the buffer first.
(internal/scheduler)
v1.6.2
1.6.2
Fixed
Results of audit round 2 (1 critical, 4 high — all fixed with regression
tests):
- Path traversal via server-controlled filename — a Content-Disposition
or URL basename containing../(or separators) reachedfilepath.Join
raw, so a hostile server (or an RPC client on--rpc-listen-all) could
write the download body anywhere the process can — including dotfiles in
the user's home. Filenames are flattened to a single path component before
joining; explicit-ooverrides are untouched. (internal/download) - RPC
odm.shutdownbypassed auth — shutdown detection ran on the raw
request while auth was checked inside dispatch, so an unauthenticated POST
got an auth-error response and still killed the daemon, cancelling all
in-flight downloads. Detection is now gated on the same secret check.
(internal/rpc) - RPC
changeOption connections ≤ 0caused silent data loss — target 0
retired every worker via the graceful drain while chunks remained queued;
Start reported the half-downloaded file as completed and deleted its
control file, making resume impossible. Values < 1 are rejected at the RPC
boundary and clamped inAdjustConns. (internal/rpc,internal/download) - Negative
--retryfaked success — a config file withretry=-1made
the per-chunk attempt loop run zero iterations: chunks "succeeded" without
a single byte downloaded and control files were deleted. Rejected by
config validation; clamped defensively inNewTask.
(internal/config,internal/download) - Last-completion tally race in the scheduler — completion reports post
to the scheduler channel after the WaitGroup release, so when the final
tasks finished back-to-back, Run could return on its done path with one
report still buffered — wrong summary line, wrong exit code, missing RPC
completion event. The done path now drains the buffer first.
(internal/scheduler)