Commit Graph
30 Commits
Author SHA1 Message Date
imSp4rky 3027592e31 fix: drop null vault keys, cap snapshot restore size, isolate test globals
Vaults.add_keys and kv add skip null content keys instead of failing the batch; is_null_key moves to core.vault. A manifest snapshot restore has one 32 MiB inflate budget. Tests reset the speed limiter and config between runs.
2026-11-02 22:01:02 -07:00
imSp4rky 96c3840b82 fix(session): keep cookie paths so same-name cookies stay on their own path
RnetCookieAdapter stores cookies by domain, path and name and sends path, secure and a leading-dot domain to rnet. It also takes server-rotated values, and worker children and track sessions get the cookies with their paths.
2026-10-31 22:57:22 -06:00
imSp4rky dfc99783b5 fix(downloader): report spent segment retries in one line without the URL
The session message names the host, not the signed URL, and a segment that fails every attempt prints one line with no traceback. API job errors come from the exception.
2026-10-23 19:47:12 -06:00
imSp4rky bfda31ce4a fix(proxy): keep service traffic on --proxy and warn about requests that bypass it
--proxy now overrides HTTP(S)_PROXY and the Windows system proxy while keeping the env CA bundle. Core fallback sessions use the run's proxy, and a warn-only audit hook names service code that sends a request or opens a raw socket outside the proxy; a session with proxy_exempt = True opts out.
2026-10-16 16:12:42 -06:00
imSp4rky 9f82777394 ci: run the unit suite in parallel and keep it off the network
Test servers shut down without the 0.5s poll wait, unit tests refuse non-loopback lookups, and three load-sensitive tests are steadied. Pushes to dev test Python 3.11 only; the docs job builds with uv.
2026-10-13 19:31:23 -06:00
imSp4rky 80f2b8f5b4 test(downloader): cover truncation and retry spin guards
Assert that an empty body of unknown length, a 206 with no complete length, and a 206 starting past the resume offset each fail instead of finalizing, and that every restart path spends an attempt.
2026-09-06 20:05:24 -06:00
imSp4rky fe7b3bb651 test(downloader): cover session state carried into download children
Assert that verify, params, cert and auth reach a spawned child, and that a mounted adapter the child cannot reproduce falls back to the single-process path. Documents --download-processes in the CLI reference.
2026-09-06 20:04:25 -06:00
imSp4rky 85503fd87c test(downloader): cover probe judging and the submission window
Assert that a probe is judged on the trailing tick, that a backoff is temporary and climbs back to the cap, and that a worker always finds its next segment staged. Documents adaptive-workers behaviour in the downloading guide.
2026-09-06 20:03:46 -06:00
imSp4rky 22cc739e83 fix(downloader): correct truncation, ranged sizing, and session carry
Check a sequential download against the complete length whenever one is known, and settle the unknown-total case on the 416. An empty body of unknown length, a 206 with no complete length, and a 206 starting past the resume offset all passed the size checks before and were renamed into place. Four restart paths spent no attempt, so a host that always answered the same way cycled forever.
2026-09-06 20:03:08 -06:00
imSp4rky dffb8b27e3 fix(serve): lock shared session state and throttle progress writes
Handlers share one service instance, so a per-session lock keeps two requests out of it at once. Track threads merge and write job state under one lock, and the progress sink coalesces ticks that say the same thing.
2026-09-06 20:01:38 -06:00
imSp4rky 2f28aa8e64 fix(dl): end a cancelled track by raising instead of returning short 2026-09-05 12:52:02 -06:00
imSp4rky d4690e5c6b fix(downloader): bound requests after a cut and stop retry nesting
Request 4 MiB sub-ranges once a host cuts a body, skip the session's own urllib3 Retry inside download(), fail the ranged path at once on a 4xx, and cap resumes at 50 of 1 MiB each. Adds a deny fault profile.
2026-08-24 15:10:36 -06:00
imSp4rky 5cece9672a fix(dl): accept a clamped 206 and report audio-description-only titles
RFC 9110 clamps a range end past the resource to its last byte; the guard
rejected that valid response. The descriptive-audio filter also had no
emptiness check, so a title with only described audio failed with a
misleading missing-language error.
2026-08-23 16:32:30 -06:00
imSp4rky dd9dafef91 fix(downloader): harden ranged reads, teardown and multiprocess sessions
Validates the 206 range end, bounds the DASH collapse, and preserves cookie domains and pool size in spawned children.
2026-08-16 17:07:28 -06:00
imSp4rky 7f73d9c469 fix(downloader): validate Content-Range before placing 206 bytes
A 206 was trusted from its status alone, so a server answering a different range than requested silently corrupted output.
2026-08-16 14:23:39 -06:00
imSp4rky e3c1ce3511 refactor(tests): drop leading underscores from test helper functions
Completes the core underscore sweep for the test files this branch added; make_ctx/make_track/seg_urls avoid shadowing locals.
2026-08-16 12:52:36 -06:00
imSp4rky ee9d3edfe3 Merge branch 'dev' into feat/downloader-speed
# Conflicts:
#	unshackle/core/api/download_manager.py
#	unshackle/core/downloaders/requests.py
#	unshackle/core/manifests/dash.py
#	unshackle/core/session.py
#	unshackle/core/utils/webvtt.py
2026-08-16 12:43:01 -06:00
imSp4rky 1af77d395d fix(downloaders): submit segments in a window so progress can drain
Submitting every segment up front convoyed on ThreadPoolExecutor's process-global lock, starving the event drain so per-track progress bars only advanced once the download had finished.
2026-08-16 11:06:17 -06:00
imSp4rky 21d5f1c5e1 refactor(core)!: drop leading underscores from internal definitions
Renames 314 defs and 41 module/class-level constants. Keeps 11 names underscored where the stripped form collides in scope, plus 5.5.0 shims for the three symbols external service repos call.
2026-08-15 22:07:41 -06:00
imSp4rky 9466db37a6 style: tighten comments added on this branch 2026-08-07 11:43:13 -06:00
imSp4rky 397cbae221 fix(downloader): accept a 200 whose body is exactly the requested byte-range slice
RFC 9110 permits a server to ignore Range and answer 200 with the whole
resource. When the item's slice starts at byte 0 and spans exactly the
response body, that 200 is byte-identical to the 206, so keep it instead of
burning all five attempts and failing the track. Every other 200 (nonzero
start, span mismatch, open-ended, encoded) still fails the attempt to
protect the merge.
2026-08-07 11:38:52 -06:00
imSp4rky e45cf15dd7 fix(downloader): fail closed on rnet content codings outside the old allowlist
The rnet branch zeroed content_length only for gzip/deflate/br, so an
unknown coding (zstd, future tokens) kept a bogus length and, worse, the
drop-the-partial-instead-of-Range-resuming guard could never fire on the
rnet path because content_encoded stayed False. Use the same fail-closed
_is_content_encoded check the requests branch uses.
2026-08-07 11:37:54 -06:00
imSp4rky 90a6164c56 test(downloader): make speed-limit tests immune to leaked cancel state
Any batch failure sets the process-global DOWNLOAD_CANCELLED, which
TokenBucket.consume waits on, so a test elsewhere could void every timing
assertion here; add the autouse clear fixture the other downloader test
modules already use. Also capture the debt sleep via a stub instead of
timing a real Event.wait, whose occasional spurious early wake tripped the
1.0s lower bound.
2026-08-07 11:36:53 -06:00
imSp4rky 431aa8fa54 fix(downloader): keep speed-limited batches in a single process
Spawned children re-import the module and start with no TokenBucket, so
--download-processes silently dropped a configured --speed-limit /
global_speed_limit for any 24+ segment batch. Fall back to the in-process
path while a limit is set (same gate as the ranged-parallel paths), log the
fallback, and document the interaction.
2026-08-07 11:36:17 -06:00
imSp4rky d7e08e979b fix(downloader): surface failed downloads instead of returning silently 2026-08-07 11:31:11 -06:00
imSp4rky cfae8f8882 Merge branch 'dev' into feat/downloader-speed
Catch the branch up with dev (45 commits) so testers can run the perf work
against current dev.

Conflict resolutions:
- downloaders/requests.py: keep the branch's racing-gated read1, adaptive
  controller speed reporting and byte-range slice guards; take dev's
  content-encoding checks (raw reads are skipped on an encoded body) and its
  rolling SpeedWindow for the no-controller speed fallback. The single-URL
  ranged-parallel gate now requires no speed limit and no caller Range.
- service.py / session.py: keep the branch's TimeoutHTTPAdapter and retry
  config, raise POOL_MAX_SIZE to 64 so dev's pool-starvation fix survives
  (pool_block=True stalls threads on a pool smaller than the in-flight count).
- docs: dev split configuration.md into per-area pages, so the branch's rnet
  network option rows moved to reference/configuration/network.md.
2026-08-07 10:45:47 -06:00
imSp4rky d448ee8ff9 fix(downloader): decode content-encoded bodies instead of writing raw
Reading .raw skips urllib3's decoding, so brotli-served subtitles landed compressed and converted to empty files. Also drops a partial rather than Range-resuming an encoded body, since a Range indexes the encoded bytes.
2026-08-05 17:56:33 -06:00
imSp4rky 734df87bad feat(dl): add --speed-limit and serve global_speed_limit
Shared token bucket across all download threads; decimal suffixes (5M = 5.0 MB/s) or plain bytes/sec
2026-07-31 12:48:46 -06:00
imSp4rky b54aadea7a fix(downloader): honor Retry-After with cancel-safe backoff, add fault-injection bench and tests 2026-07-18 20:21:30 -06:00
imSp4rky 2adf8bad87 fix(downloader): validate 206 on byte-range segments and force identity encoding
Segments carrying their own Range (DASH SegmentBase, HLS EXT-X-BYTERANGE) now
fail the attempt when a server ignores the Range and returns 200, instead of
writing the whole parent resource as the segment and corrupting the merge. The
resume path no longer clobbers such a slice's Range, and retries rewrite the
slice whole. All segment GETs request Accept-Encoding: identity and the rnet
path mirrors the gzip guard, so transparent CDN (de)compression can't break the
Content-Length truncation check. Adds tests pinning these plus the adaptive
start-at-cap wiring and the small-segment tail-probe gate.
2026-07-16 00:00:26 -06:00