manga-termux 1.0.6 — fixed

Repairs a broken pip install manga-termux (v1.0.5). The tool could not build a chapter list, so every download failed, and it reported network failures as "no results", which made search look broken too.

Verified against the live site on 2026-09-27.

Root cause

1. Chapter list — the fatal bug

weebcentral.com switched its markup from absolute to relative links:

<!-- what 1.0.5 expected -->
<a href="https://weebcentral.com/chapters/01M3DVDYA933SQQ6703XQYMMGQ">

<!-- what the site actually serves now -->
<a href="/chapters/01M3DVDYA933SQQ6703XQYMMGQ">

The 1.0.5 regex in scraper.py hard-coded the absolute prefix:

r'href="(https://weebcentral\.com/chapters/[^"]+)"[^>]*>.*?<span class="">\s*([^<]+?)\s*</span>'

So it matched nothing. get_all_chapters() returned [], and the CLI looped on [!] Could not fetch chapter list. Try again. — no download could ever start. A live measurement confirmed it: One Piece parsed 0 chapters before the fix, 1194 after.

2. Errors were silently swallowed

Every scrape call was wrapped in except Exception: return []. A DNS failure, a timeout and a genuine "no matches" were indistinguishable — so "search doesn't work" and "chapter list is empty" both surfaced as a bare "no results".

3. Wrong default save path off Android

DEFAULT_DOWNLOAD_DIR was hard-coded to /sdcard/Manga, which cannot exist on Linux/macOS/WSL.

What changed

Everything else — chapter-range syntax, CBZ/PDF/images output, aria2c 16-connection downloading, resume/skip, the prompt_toolkit live-search UI — is unchanged.

Install

Option A — patch the installed package (fastest, recommended)

cd manga-termux-fix
bash apply_fix.sh

It resolves the installed package through its metadata (safe to run from inside this folder), backs up the originals as *.bak-<timestamp>, copies the fixed files in, and clears stale __pycache__. If your interpreter is python3:

PYTHON=python3 bash apply_fix.sh

Option B — install from this source tree

cd manga-termux-fix
pip install --force-reinstall .

Option C — build a wheel, copy to the phone

cd manga-termux-fix
pip install build && python -m build --wheel
# then on Termux:
pip install --force-reinstall manga_termux-1.0.6-py3-none-any.whl

Revert

The original files are kept next to the patched ones as *.bak-<timestamp>:

TARGET=$(python -c "from importlib.metadata import distribution as d; print(d('manga-termux').locate_file('manga_termux'))")
cd "$TARGET"
for bak in *.bak-*; do cp "$bak" "${bak%.bak-*}"; done
rm -rf __pycache__

Verified after the fix

Known limits

• It is a scraper, so it will break again whenever weebcentral changes its markup. The parser now tolerates relative/absolute hrefs and several title layouts, but no regex scraper is future-proof.
• Some series' early chapters sit under a 0 / prologue number, so ranges are keyed to whatever the site labels them.
• Downloads depend on third-party image hosts (temp.compsci88.com, hot.planeptune.us). If one starts refusing requests, pages fail individually and the per-page URLs are now printed in the error log.
• You are responsible for what you download and for complying with copyright law and the site's terms in your jurisdiction.