Skip to content

Releases: fallen020/comic-dl

comic-dl v0.0.6

Choose a tag to compare

@github-actions github-actions released this 11 Oct 18:21
v0.0.6
ab1571b

[v0.0.6] - 2026-10-11

Added

  • Global egress override for IP-reputation blocks: --proxy flag,
    [http] proxy key, $COMIC_DL_PROXY / $HTTPS_PROXY fallback.
    Proxy credentials are masked in config show and resume hints.
  • Per-host image concurrency cap (default 3, [http] per-host-concurrency), so single-node hosts stop absorbing the full
    eight-wide burst.
  • Session headers locked to the TLS impersonation profile; gallery
    sessions send page-load fetch metadata, image requests send the image
    set.
  • Shell completions with per-flag descriptions, PowerShell support, and
    a tabular logo.
  • RichPopup gallery scraper; FSIComics es/de/fr/it language mirrors.
  • new-site.py scaffolds new built-in sites; plugin validation names
    the repair for each failure.

Changed

  • Site adapter bumps: fsicomics 2.0.0 (mirrors), ehentai 1.0.4,
    imhentai 1.0.2, mangadex 1.0.3.
  • Dependencies: platformdirs 4.13.0, cryptography 50.0.2,
    ruff 0.17.0, iniconfig 2.3.1.
  • Generic fallback: shape-based series detection, honest chapter
    numbers, and a decline reason naming why a URL was skipped.
  • Release packaging runs in CI on PRs: vendored wheels per arch, native
    arm64 builders, distro package cache, uncached clean-container
    validation.
  • Verbose output retagged ([context]/[trace]), per-image headers
    folded onto the request line, dry-run URLs redacted.

Fixed

  • Challenge solves no longer replay POSTs; the solve still clears the
    way for later requests.
  • Per-page fan-outs stay resilient instead of fail-fast (ehentai,
    imhentai); MangaDex at-home assignments fetch fresh.
  • Download cooldown keyed per host; rate-limiter slot reserved before
    sleeping; each redirect hop validated exactly once; streamed block
    bodies classified by peek.
  • Dry-run index width zero-padded cross-platform.

Security

  • Release checksum verification before privileged install.

What's Changed

  • test: add site scraper contract test with violation allowlist by @fallen020 in #56
  • chore(site): gate, fix, and tune the docs site by @fallen020 in #66
  • feat: harden image downloads against WAFs and throttling CDNs by @fallen020 in #67
  • chore: drop duplicated chown comment in release.yml by @fallen020 in #68
  • chore: release v0.0.6 by @fallen020 in #69

Full Changelog: v0.0.5...v0.0.6

comic-dl v0.0.5

Choose a tag to compare

@github-actions github-actions released this 04 Oct 19:21
v0.0.5
af90e74

[v0.0.5] - 2026-10-04

Added

  • Stable site_error_code values now surface a fix hint: a central
    per-code hint map fills in whenever a failure carries a code but no
    hand-written hint, so blocked, locked, or moved chapters always print
    the cause plus what to do.
  • Interrupt resume hint echoes the real command run (canonical program
    name, prompt-entered URL included).
  • Downloaded images are structurally verified; truncated and polyglot
    pages are rejected instead of archived.
  • Help screen rebuilt from the parser; self site live-checks test URLs
    for 27 adapters with kind-aware checks.
  • Single shared retry path (retry_transient) for scraper-side fetches.
  • Hivetoons series roster parses the site's TanStack stream payload.
  • Website: llms.txt generation, Explain-with-AI dropdown, namespaced
    tab sets, responsive shell, one-row-per-site tables with live filter.

Changed

  • Site adapter bumps: asurascans 1.0.3, tapas 1.0.2, flamecomics 1.0.3,
    ehentai 1.0.3, kagane 1.0.3, hivetoons 1.0.3.
  • Dependencies: attest-build-provenance v4.2.2, hatchling >= 1.32.4,
    mypy 2.4.0, ruff 0.16.10, platformdirs 4.12.3, astro 7.3.5,
    @lucide/astro ^1.49.0.
  • Arch package renamed to avoid an AUR collision; CI artifact retention
    capped; config starter slimmed.

Fixed

  • WAF/challenge misclassification: Asura, Tapas, and FlameComics
    Cloudflare challenges now report SITE_BLOCKED with a solver hint
    instead of a network error; Kagane API 404s report removal instead of
    request failure; Stonescape maps non-404/403 HTTP errors; E-Hentai
    raises on retry exhaustion and API errors.
  • site_error_code coverage across the remaining adapters (malformed
    URLs, locked chapters, ValiScans platform).
  • Interrupt hint no longer leaks the launcher path; library errors no
    longer leak SQLite driver text.
  • Gedecomix test URL repointed to a live chapter; FSIComics series
    scraping restored on the Foxiz theme; self site columns corrected.
  • SSRF, content-length, referer, and cookie-permission hardenings; rate
    limiter keyed by hostname; host park lifted on success.

What's Changed

  • fix(fsicomics): restore series scraping on Foxiz theme, full summaries by @fallen020 in #47
  • fix(self-site): correct check columns, usage errors, status provenance by @fallen020 in #48

Full Changelog: v0.0.4...v0.0.5

comic-dl v0.0.4

Choose a tag to compare

@github-actions github-actions released this 27 Sep 18:33
v0.0.4
3ef6a98

[v0.0.4] - 2026-09-27

Added

  • Ten new built-in sources — DivaScans, HD Porn Comics, IMHentai,
    KodokuEasyAccess, ManhuaTo, Nyx Scans, Tapas, ToonVerse, ValirScans, and
    Vortex Scans — taking built-in support from 21 to 29 sites.
  • First-run legal notice.
  • Stable site_error_code failure classification across scrapers, surfaced
    in errors and batch output.
  • Per-page download state manifest with a resume-safe retry pass; partial
    chapters tracked separately in batch footers.
  • Connectivity probe before blaming the network for failures.
  • Library schema v4: one global database with per-row output roots, global
    library commands, and library update --dry-run.
  • Leading m. mobile subdomains normalized away before matching.

Changed

  • KodokuStudio replaced by KodokuEasyAccess (new domain).
  • System-webview solver off by default.
  • Archive writes stream to disk instead of buffering whole image sets.
  • Website design refreshed; dev guides rewritten as end-to-end processes.

Removed

  • ManhwaZ support dropped.

Fixed

  • Kagane reports the Cloudflare challenge instead of a network error.
  • Batch output honesty: Saved: gated on success, per-page failure reasons,
    chapter/partial accounting; single-URL polish (short chapter titles, one
    completion line).
  • Tapas cover/creator/description; Asura Scans full series summary; GenzToons
    listing guard, locked chapters, metadata cleanup; Vortex/Nyx cover,
    synopsis, genres.
  • Cookie collision resolution, charset-aware cache decode, cookie-replay
    probing with a JSON cache-poisoning guard; .partial markers restored to
    on-disk page names so reruns skip them.

Security

  • Cookie jar encrypted at rest.
  • Outbound fetches fail closed on DNS/validation errors; rebinding window
    narrowed.
  • 2xx media payloads never treated as antibot challenges; DataDome markers
    tightened.

Full Changelog: v0.0.3...v0.0.4

comic-dl v0.0.3

Choose a tag to compare

@github-actions github-actions released this 20 Sep 11:29
v0.0.3
b4149b2

[v0.0.3] - 2026-09-20

Added

  • comic-dl self version and comic-dl self update — detects whether comic-dl
    was installed by apt, dnf, pacman, pip, uv, a source checkout, or a standalone
    binary and updates it through the owning package manager (with --yes,
    sudo-when-needed). Never touches a source checkout or unwritable environment;
    it reports and instructs instead.
  • comic-dl self site list / check / update — per-site support versioning:
    every built-in adapter ships with a semantic version and minimum core version,
    checked against the site-support.json manifest attached to each release.
  • Eight new built-in sources: HiveToons, GenzToons, Qimanga, StoneScapes,
    EN-ThunderScans, KingOfShojo, ManhwaTop, and LGBTics (taking built-in support
    from 13 to 21 sites).
  • Built-in scrapers are auto-discovered from the registry; a broken adapter
    module fails fast instead of silently disappearing from --list-sources.

Changed

  • Adapter versioning rules and update flow documented in
    docs/usage/site-support.md; self-update strategies in
    docs/usage/self-update.md.
  • Docs and website refreshed for v0.0.3 (SEO/AI discoverability, accessibility,
    brand icons, regenerated supported-sites tables).

Fixed

  • self update asset selection and --yes paths race-free and arch-correct on
    Apple Silicon CI runners.
  • Update checks no longer mis-report a dev snapshot as "up to date" against the
    published release with the same number.

Full Changelog: v0.0.2...v0.0.3

comic-dl v0.0.2

Choose a tag to compare

@github-actions github-actions released this 15 Sep 10:14
v0.0.2
ac78079

[v0.0.2] - 2026-09-15

Added

  • WeebCentral as a built-in source.
  • Bare positional URL shorthand: comic-dl <url> instead of comic-dl -u <url>.
  • --chapters flag now accepts canonical chapter numbers (including 0).
  • comic-dl plugin list, validate, and scaffold commands for plugin
    management.
  • Typo-aware "Did you mean" suggestions when a flag is misspelled.
  • Plugin and Madara series URLs routed via matches_series_url.

Fixed

  • Cloudflare replay sessions re-use cookies on blocked downloads.
  • FlameComics: correct series metadata derivation, deduplication, and chapter
    sort order.
  • E-Hentai: URL parsing hardened, throttle recovery improved, size estimates
    use display images.
  • Dry-run mode: series previews point at the real series folder path.
  • Asura Scans: friendly 404s and CDN host hardening.
  • WebTOON: episode number read from data attribute; duplicate helpers removed.
  • Diagnostic output sanitized at the error boundary; unified error taxonomy
    across scrapers.
  • PyInstaller binary now bundles banner.txt.
  • asyncio.as_completed replaced in dry-run overlay to fix hanging tasks.
  • Parse errors, summary, and progress styling polished.

Security

  • Probe redirects validated before dispatch.
  • Cookie jar rejects public-suffix cookies; Secure flag enforced on replay.
  • Identity hashes use SHA-256 (replaced MD5).
  • Solver URL validated before use.
  • Download-path deduplication prevents staging-directory collisions across
    chapters.

comic-dl v0.0.1

Choose a tag to compare

@github-actions github-actions released this 06 Sep 17:52
v0.0.1
6587065

comic-dl v0.0.1

Downloads comic/manga galleries from supported websites and
compiles them into verified CBZ, ZIP, and CBT archives.

What's in the box

  • 12 built-in sources plus a plugin API (comic-dl --list-sources)
  • Parallel downloads paced by a per-host rate limiter, with retry
    and resume (--concurrency, SQLite library, .partial markers)
  • Verified archives: magic-byte checks, SHA-256 dedup, atomic
    writes, ComicInfo.xml metadata at series and chapter level
  • Cloudflare handling: TLS/HTTP impersonation and a system-webview
    solver for managed challenges
  • SSRF-hardened outbound traffic; retries back off politely
  • Offline-safe automated test suite (no live network in CI)

Install

  • Python package: download the wheel or sdist below and
    pip install <file> (Python 3.11+). PyPI publishing is enabled
    by a trusted publisher only when one is configured; the artifacts
    on this release are always installable directly.
  • Windows: comic-dl-v0.0.1-windows-amd64.zip (standalone exe)
  • Linux: .deb (Debian/Ubuntu), .rpm (Fedora/RHEL),
    .pkg.tar.zst (Arch) for amd64 and arm64
  • See docs/ for per-platform commands

Known limitations

  • Windows ARM64 and macOS binaries are not shipped yet — use uvx
  • The Windows binary is unsigned
  • Early maturity — read the README status banner before use

All artifacts are checksummed in SHA256SUMS.

Full Changelog: https://github.com/fallen020/comic-dl/commits/v0.0.1