本文へ移動
cccskills
無料GitHub で公開

update-site-for-release

Update/publish the Apache Tika website (tika-site SVN repo) for a release — step 17 of the Release Process. Handles the 4.x track (Changes page + aggregate javadoc + Antora docs branch) vs the 3.x maintenance track (full per-version apt docs + javadoc). Use for "update the site", "publish the site for X.Y.Z", "the website part of the release".

インストール方法を見る

含まれるファイル(5)

  • SKILL.md20.4 KB
  • scripts/extract-tika-contribs.py10.4 KB
  • scripts/extract-tika-issues.py7.4 KB
  • scripts/README.md2.8 KB
  • scripts/scaffold-stable-version.sh2.9 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

<!-- Licensed to the Apache Software Foundation (ASF) under one or more contributor license agreements. See the NOTICE file distributed with this work for additional information regarding copyright ownership. The ASF licenses this file to You under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. -->

Local override: $TIKA_SKILLS_LOCAL/update-site-for-release/LOCAL.md (default ~/.tika-skills), read after this file, wins on conflict.

Update the Tika website for a release

Step 17 ("Update Tika site") of the Release Process (https://cwiki.apache.org/confluence/spaces/TIKA/pages/109454070/Release+Process). Assumes the release (tag, artifacts, VOTE, dist promotion) is done; covers the website only.

First: cd $SITE && svn update. The checkout is usually behind (docs republishes, logo/security edits land between releases); a stale base only surfaces at the very end, after the 67 MB commit has uploaded, as E155011 ... is out of date, and the whole transaction rolls back.

Set these first:

  • $SITE — tika-site SVN checkout (not git): src/site/ (sources) + publish/ (generated, SVN-tracked, served; mvn install regenerates it and auto-runs svn add --force publish via antrun).
  • $SCRATCH — release working dir: unzipped src release, CHANGES-<NEW>.txt, built javadoc.

Scripts: ./scripts/. Local paths + toolchain (Maven binary, JDKs) live in a private companion skill under ~/.claude/skills/.

[HUMAN] gates — never do these yourself: final svn commit (outward-facing, irreversible), JIRA "release", s.apache.org shortlink, announce emails. Prepare everything, show svn status, hand off.


0. Inputs + release track

Two supported tracks since 4.0.0 (2026-08-21): 4.x is the current line, 3.x is the maintenance line. Which track a release is on decides the process; both tracks are "stable" — there is no preview slot unless a future 5.x preview reintroduces one.

InputExampleNotes
NEW_VERSION4.0.1the release
RELEASE_TRACK4.x4.x or 3.x (maintenance)
PREV_TAG / NEW_TAG4.0.0-rc1 / 4.0.1git tags, for the GitHub contributor query
JIRA fixVersion4.0.1may differ from the label (pre-releases often use the base version)
CHANGES file$SCRATCH/CHANGES-4.0.1.txtnotable-changes source
src release ziptika-<NEW>-src.zipjavadoc source (both tracks)
release date2026-08-21doap.rdf + news blurb

Confirm current values in pom.xml (repo root — tika.stable.version = 4.x line, tika.maintenance.version = 3.x line).

Before starting, check the final git tag exists (git ls-remote --tags https://github.com/apache/tika | grep <NEW>). The vote passes on an -rcN tag and nothing creates the plain <NEW> tag; 4.1.0 shipped with only 4.1.0-rc3. Ask the RM to tag the rc commit (git tag <NEW> <NEW>-rcN^{} + push) — the contributor query, the docs branch and the next release's PREV_TAG all want it. Until then use the rc tag for --tag.

Step4.x track3.x maintenance track
pom.xml <parent><version>leave at newest 3.x→ <NEW> (must stay a 3.x parent — Java 11 build)
pom.xml tika.stable.version→ <NEW>leave
pom.xml tika.maintenance.versionleave→ <NEW>
src/site/apt/<NEW>/only index.apt (Changes)full 8-file set (scaffold from prev 3.x)
site.xml entrysub-menu linking docs/<X.Y>.x/ pages + Changes + apifull legacy sub-menu, expanded
formats.aptn/a — docs/.../formats.adoc; its generated partial is test-enforcedregenerate from tika-app jar
javadocclean install -Pfast + javadoc:aggregate → publish/<NEW>/api (step 7)same
Antora docsnew minor → new docs/<X.Y>.x branch; patch → republish same branch (step 7)n/a
Download pageautomaticautomatic

doap.rdf, index.apt.vm news, verify, publish are common to both.

URL scheme (decided 2026-08-19): Antora docs are per-minor (/docs/4.0.x/); the apt Changes page and javadoc are per-release (/4.0.0/, /4.0.0/api/) — javadocs are exact-version artifacts.


1. pom.xml versions (repo root) [AGENT]

  • 4.x: bump <tika.stable.version> to <NEW>.
  • 3.x: bump <tika.maintenance.version> and <parent><version> to <NEW>.
  • <parent> supplies build config only (no site content references it since 4.0.0). It must stay on the newest 3.x parent: the 4.x parent enforces Java 17, but maven-site-plugin 3.4 needs the Java 11 build (step 8).

Download page auto-reads these — no manual edit. (download.apt.vm also derives $docsLine = 4.1.x from tika.stable.version for its Antora links; check the artifact list against dist/release/tika/<NEW>/ if the distribution set changed.)


2. src/site/site.xml menu [AGENT]

Current 4.x + current 3.x expanded; older = collapse="true".

  • 4.x, new minor: new expanded block at the top; links go into the Antora tree plus the per-release Changes/api (repoint docs/<old>.x → docs/<new>.x when a new minor supersedes; a patch release only updates the Changes/api version numbers in the hrefs):
    <item name="Apache Tika 4.0.0" href="docs/4.0.x/index.html">
      <item name="Documentation Home" href="docs/4.0.x/index.html"/>
      <item name="Supported Formats" href="docs/4.0.x/formats.html"/>
      <item name="Using Tika" href="docs/4.0.x/using-tika/index.html"/>
      <item name="Getting Started (Java API)" href="docs/4.0.x/using-tika/java-api/getting-started.html"/>
      <item name="Pipes" href="docs/4.0.x/pipes/index.html"/>
      <item name="Configuration" href="docs/4.0.x/configuration/index.html"/>
      <item name="Migrating to Tika 4.x" href="docs/4.0.x/migration-to-4x/index.html"/>
      <item name="Changes" href="4.0.0/index.html"/>
      <item name="API Documentation" href="4.0.0/api/"/>
    </item>
    
  • 3.x: new expanded block above the previous 3.x; add collapse="true" to the old block:
    <item name="Apache Tika 3.3.2" href="3.3.2/index.html">
      <item name="Getting Started"                href="3.3.2/gettingstarted.html"/>
      <item name="Supported Formats"              href="3.3.2/formats.html"/>
      <item name="Parser API"                     href="3.3.2/parser.html"/>
      <item name="Parser 5min Quick Start Guide"  href="3.3.2/parser_guide.html"/>
      <item name="Content and Language Detection" href="3.3.2/detection.html"/>
      <item name="Configuring Tika"               href="3.3.2/configuring.html"/>
      <item name="Usage Examples"                 href="3.3.2/examples.html"/>
      <item name="API Documentation"              href="3.3.2/api/"/>
    </item>
    

3. Per-version apt docs src/site/apt/<NEW>/ [AGENT]

4.x — create only index.apt (the Changes page; title Apache Tika <NEW>; fill step 4). No other apt files — everything else lives in the Antora docs.

3.x — scaffold (these docs are version-string-identical across 3.x):

./scripts/scaffold-stable-version.sh $SITE 3.3.1 3.3.2

Copies+bumps configuring/detection/examples/parser/parser_guide/gettingstarted.apt. Then:

  • formats.apt — two parts: a hand-written top (license, intro, %{toc}, ~25 prose format-family sections) down to the header line Full list of Supported Formats in "standard" artifacts, then a generated flat list below it. --list-parser-details-apt regenerates only the flat part. Copy prev formats.apt, bump versions, replace everything below that header:
    java -jar <path>/tika-app-<NEW>.jar --list-parser-details-apt
    
    Do NOT truncate at the %{toc} line — keep the prose sections. (Scaffold skips formats.apt for this reason.)
  • index.apt — step 4.

4. Per-version index.apt: notable changes + contributors [AGENT + HUMAN]

Shape (see src/site/apt/3.3.1/index.apt): license+title; "most notable changes…" bullets; "The following people have contributed…" bullets; "See {{https://s.apache.org/XXXX}} …".

Notable changes — review output, keep only notable items:

./scripts/extract-tika-issues.py CHANGES-3.3.2.txt out-3.3.2.apt 3.3.2

Mirrors CHANGES verbatim; TIKA-####/Github-#### auto-linked; ALL-CAPS headers (incl. SERVER, PIPES AND OPERATIONS, INFERENCE (EXPERIMENTAL)) → apt sections; apt markup chars in prose (<= 0, type->parser, {x}, [], ~35%) are backslash-escaped — an unescaped < fails the site build with Error parsing '.../index.apt': line [N] missing '>'; output is wrapped at 78 cols without splitting hyphenated words or links.

Contributors — candidate list, RM curates:

./scripts/extract-tika-contribs.py 3.3.2 --prev-tag 3.3.1 --tag 3.3.2 > contribs.txt
# beta (fixVersion differs from label):
# ./scripts/extract-tika-contribs.py 4.0.0 --prev-tag 3.3.1 --tag 4.0.0-beta-1

Merges JIRA (reporters/assignees/comment authors) + GitHub commit/PR authors (uses gh auth; resolves logins→names; case-insensitive sort; filters bots/AI). Over-reports drive-by commenters, misses GitHub-issue-only commenters. [HUMAN] prune / normalise / add.

Shortlink [HUMAN] — s.apache.org/XXXX → the JIRA "issues fixed in <NEW>" query; needs s.apache.org login. Ask the RM.


5. src/site/resources/doap.rdf [AGENT]

New <release> at the top. Ordering is by date, not version (a stable point release can sit above an older-dated preview — 3.3.2/Jul-16 above 4.0.0-beta-1/Jul-3):

      <release>
        <Version>
          <name>Apache Tika 3.3.2</name>
          <created>2026-07-21</created>
          <revision>3.3.2</revision>
        </Version>
      </release>

6. Home page src/site/apt/index.apt.vm [AGENT + HUMAN]

  1. New Latest News block at the top; its CHANGES link uses dist.apache.org/repos/dist/release/... (live mirror):
    [21 July 2026: Apache Tika Release]
     Apache Tika 3.3.2 has been released! <one or two sentence summary>.
     Please see the {{{https://dist.apache.org/repos/dist/release/tika/3.3.2/CHANGES-3.3.2.txt}CHANGES.txt}}
     file for the full list of changes in the release and have a look at the download page for more information
     on how to obtain Apache Tika 3.3.2.
    
  2. Repoint the superseded release's CHANGES link (its artifacts get svn rm'd from the live mirror at release): dist.apache.org/repos/dist/release/tika/<PREV>/… → archive.apache.org/dist/tika/<PREV>/…. [HUMAN] confirm which version was removed.

7. Docs / Javadoc

Javadoc — both tracks [AGENT]. NOT the wiki's javadoc:aggregate-no-fork (runs against tika-parent, its relative <sourcepath> fails → No source files for package org.apache.tika; wrong goal, not a JDK issue). From the unzipped src release (its ./mvnw is broken — use system mvn):

unzip tika-3.3.2-src.zip && cd tika-3.3.2
mvn clean install -Pfast        # ~4 min; module artifacts + full dep classpath
mvn javadoc:aggregate           # FORKING goal (NOT -no-fork)
mkdir -p $SITE/publish/3.3.2
mv target/reports/apidocs $SITE/publish/3.3.2/api

Both steps matter: without install javadoc dies on package org.slf4j does not exist; the forking aggregate (@aggregator) runs once on the root, -no-fork breaks per-pom-module. 3.x: any modern JDK (11 and 25 verified); 4.x: Java 17+ (21 verified on 4.1.0, ~3,900 html / 67 MB). A -Pfast install that prints a Felix/OSGi stack trace still succeeds — check the <NEW> artifacts landed in the local repo, not the tail of the log. (tika-server miredot docs discontinued — skip.)

4.x — Antora docs [AGENT]. Built from the tika git repo (main checkout), not the src zip — the playbook pulls every docs/{0..9}* branch as a content source. New minor: create docs/<X.Y>.x from the tag (or main), set version: '<X.Y>.x' + tika-version and tika-javadoc-url attributes in that branch's docs/antora.yml, and make sure main's antora.yml has prerelease: true and its version/tika-version match main's new <revision> (the release plugin bumps the poms, not antora.yml — after 4.1.0 main was 4.1.1-SNAPSHOT in tika-parent/pom.xml but still 4.1.0-SNAPSHOT in antora.yml) [HUMAN commits]. Cut the branch from the release tag, not main: main's docs may already describe post-release changes. Patch: commit doc changes + tika-version bump to the existing branch (bump tika-javadoc-url too). Then:

cd tika-main
rm -rf docs/target/site            # Antora never prunes: a stale <old>-SNAPSHOT dir from an earlier build gets published
./mvnw package -Papache-release -pl :tika-docs -DskipTests
./docs/publish-docs.sh $SITE/publish

Antora reads only LOCAL branches of the clone (a remote-only origin/docs/4.0.x is ignored — verified on 4.1.0): before building, git branch docs/4.0.x origin/docs/4.0.x for every released line, and have main checked out so HEAD supplies the SNAPSHOT. Run a trial build with the new docs/<X.Y>.x checked out (its worktree = HEAD, so uncommitted antora.yml edits are exercised) to validate before the human commits; do NOT publish that trial output — it lacks the other versions, so the version dropdown, sitemap and /docs redirect would be wrong. publish-docs.sh copies target/site into publish/docs/, flattens URLs, rewrites the search index (has its own guards). First 4.x publish after the SNAPSHOT era: svn rm publish/docs/<old>-SNAPSHOT (nothing prunes it) and verify publish/docs/index.html redirects to the released line, not a SNAPSHOT. Full procedure: docs/modules/ROOT/pages/maintainers/site.adoc.


8. Build + verify [AGENT]

tika-site has no ./mvnw → system mvn. Build with Java 11 — it pins maven-site-plugin 3.4 (2014), unreliable on newer JDKs; a Doxia error here means wrong JDK, not a content problem (separate from step 7's JDK-agnostic javadoc).

cd "$SITE"
mvn clean install

ALWAYS clean install, never bare install — an incremental build leaves publish/css/ stale → pages render with no CSS/sidebar. Fix is a clean rebuild, not a CSS edit.

Build auto-copies target/site → publish/, strips timestamps, svn add --force publish. Check: new version in the menu; news + download versions right; pages styled (CSS + sidebar); per-version pages + javadoc/Antora resolve. Preview: mvn site:run → http://localhost:8080.


9. Stage + hand off the commit [HUMAN]

cd "$SITE"
svn status
svn add src/site/apt/<NEW>            # + any other new files
# hand to the RM — do NOT run yourself:
# svn commit -m "Update website for <NEW> release."

Big-commit caveat (javadoc): publish/<NEW>/api is ~3,000 files / ~55 MB; a single commit often times out / E000104 Connection reset by peer — this is size, NOT auth (bad password = Authentication failed/403, and cached creds won't re-prompt). Fixes:

  • http-timeout = 1800 in ~/.subversion/servers [global].
  • Else commit the api in chunks, then the rest:
    svn commit --depth=empty publish/<NEW> publish/<NEW>/api \
        publish/<NEW>/api/org publish/<NEW>/api/org/apache \
        publish/<NEW>/api/org/apache/tika -m "<NEW> site: api dir skeleton"
    for d in publish/<NEW>/api/org/apache/tika/*/; do
      svn commit "$d" -m "<NEW> javadoc: $(basename "$d")" || break   # parser/ ~1,300 files
    done
    svn commit publish/<NEW>/api -m "<NEW> javadoc: remaining api files"
    svn commit -m "Update website for <NEW> release."
    
  • Atomic per invocation → a failed commit rolls back; retry. Locked (E155004) → svn cleanup.

10. Confirm published + re-kick [HUMAN]

svnwcsub maps /www/tika.apache.org ← %(ASF)s/tika/site/publish: only a commit touching publish/ triggers a republish, and it publishes the whole tree at HEAD. Verify (cache-buster hits the origin, not Varnish):

curl -s -o /dev/null -w "%{http_code}\n" "https://tika.apache.org/<NEW>/index.html?cb=$(date +%s)"

Want 200. Still 404/old minutes later → the web-node svn up choked on the big commit. Re-kick with a trivial whitespace commit to a file under publish/ (e.g. a blank line in publish/index.html):

svn commit publish/index.html -m "Nudge svnwcsub to republish."

Re-fires svnwcsub → svn up to HEAD (harmless; next build regenerates it). A commit outside publish/ won't trigger. Still stuck ~30 min → ping #asfinfra.


11. Post-site [HUMAN] (context)

  • JIRA: "release" <NEW>; move stragglers to the next version.
  • Announce to user@, dev@, announce@apache.org.
  • Log at https://reporter.apache.org/addrelease.html?tika.
  • CVEs fixed → update security.apt / per-version security notes, republish.

Checklist

  • svn update done before any edits
  • 4.x vs 3.x track decided
  • pom.xml versions (stable for 4.x; maintenance+parent for 3.x)
  • site.xml: new entry added, previous same-track entry collapsed
  • per-version apt docs (index.apt only for 4.x / full set for 3.x)
  • formats.apt regenerated (3.x only; 4.x formats page is in Antora, partial test-enforced)
  • index.apt: notable changes + curated contributors + shortlink
  • doap.rdf entry
  • index.apt.vm: news block + superseded CHANGES link → archive
  • final <NEW> git tag exists (not just -rcN)
  • javadoc → publish/<NEW>/api (both tracks); Antora branch published (4.x); main antora.yml bumped to the next SNAPSHOT
  • mvn clean install (never bare install); pages styled (CSS + sidebar)
  • svn status/svn add done, commit handed to RM (chunk api/ if it resets)
  • live site 200: https://tika.apache.org/<NEW>/index.html?cb=… — else re-kick

Troubleshooting

SymptomCauseFix
aggregate-no-fork → No source files for package org.apache.tikaruns against tika-parent; relative <sourcepath> can't resolveuse forking javadoc:aggregate after clean install -Pfast (step 7)
javadoc → package org.slf4j does not exist etc.aggregate without a prior build → empty classpathmvn clean install -Pfast first
site build: Error parsing '.../index.apt': line [N] missing '>'unescaped </> (e.g. <= 0, ->) in apt prosebackslash-escape (\<); bundled script does this
pages unstyled (no CSS/sidebar)incremental install left publish/css/ stalemvn clean install (never bare install)
site-plugin / Doxia error on mvn installmaven-site-plugin 3.4 on too-new a JDKbuild with Java 11
notable-changes bullet split on a version numberold numeric heuristic (removed)use the bundled script; re-run
contributors have bots/AI, or surname orderold behavior (fixed): filter + str.casefold sortuse bundled extract-tika-contribs.py; RM curates
commit E175012 timed out / E000104 Connection reset~55 MB api tree too big for one transactionhttp-timeout=1800; chunk the api/ (step 9). Size, not auth.
commit Authentication failed / 403genuinely bad/expired credentialsvn commit --username <you> to re-cache
svn: E155004 working copy lockedprior commit died mid-transactionsvn cleanup, retry
commit E155011 Directory ... is out of dateworking copy behind HEADsvn revert -R publish && svn update, re-run publish-docs.sh + mvn clean install, retry (fixed base only)
committed, site still old even with ?cb= (origin 404s/old)web-node svn up choked on the big commitre-kick: whitespace commit under publish/ (step 10); stuck ~30 min → #asfinfra
home-page CHANGES link 404s for the previous releaseit was svn rm'd from the live dist mirrorrepoint to archive.apache.org/dist/tika/<prev>/CHANGES-<prev>.txt (step 6)

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Ground rules for working in the Tika codebase — git policy, Maven wrapper/repo conventions, building and testing specific modules, code and test conventions, pre-commit checks. Load at session start for any Tika development task.

日本語の概要は準備中です。原文の説明を表示しています。

apache/tika4,0932026年10月10日 更新

Taking a multi-PR feature from "shape unknown" to merged without five review rounds per PR: spike until interfaces stop moving, write the contract, cut PRs along contract seams, one review per PR. Use when starting a feature that touches more than one lifecycle object or public interface, when a PR review keeps changing interfaces, or when splitting a large branch.

日本語の概要は準備中です。原文の説明を表示しています。

apache/tika4,0932026年10月10日 更新

Examine what a file claims about itself and what it actually contains — powered by Apache Tika. True content-based type detection (extensions lie), provenance claims (authors, dates, creating application), revision and tamper signals (PDF incremental updates, tracked changes, hidden slides, zip integrity), hidden and embedded content (attachments, macros), risk indicators (PDF JavaScript actions, encryption), and content digests. Evidence gathering, not verdicts: Tika reports what the file asserts and what parsing observed; it does not attribute authorship or validate signatures. Use for triaging suspicious files, provenance questions, e-discovery-style review, or "is this file what it claims to be."

日本語の概要は準備中です。原文の説明を表示しています。

apache/tika4,0932026年10月10日 更新

Turn almost any file into Markdown plus metadata — PDF, Office, HTML, email, archives, images, audio/video, 1000+ formats — powered by Apache Tika, either via the tika-app CLI (zero setup, one file) or a running tika-server (curl, warm process, many calls). Leads with rmeta (structured, embedded-item-aware output) as the default operation rather than flat concatenated text, since you can't tell whether a file has embedded content from its extension. Covers metadata-only triage, type and language detection, OCR, and output-size discipline for agent context. Use whenever a task involves reading the content of a file whose format you don't want to hand-parse.

日本語の概要は準備中です。原文の説明を表示しています。

apache/tika4,0932026年10月10日 更新

Run Apache Tika as a Docker container when you need guaranteed OCR (scanned PDFs, images) or geospatial raster support with zero local install — `apache/tika:<version>-full` bundles Tesseract, GDAL, ImageMagick, and fonts. Also covers the minimal image, port/volume/memory conventions, the path-identity mount gotcha, and how to confirm OCR actually ran rather than silently returning no text. Powered by Apache Tika. Use when a local `tika-app`/`tika-server` doesn't have Tesseract installed, or you want a disposable, self-contained parsing environment. Companion to the `file-to-markdown` skill, which covers the parsing calls themselves once a server is up.

日本語の概要は準備中です。原文の説明を表示しています。

apache/tika4,0932026年10月10日 更新

Working with tika-metadata-schema, the build-gated registry of Tika metadata keys — regeneration after Property changes, gate tests, naming conventions, post-rename sweeps. Use when adding or renaming metadata keys or when the schema gate fails.

日本語の概要は準備中です。原文の説明を表示しています。

apache/tika4,0932026年10月10日 更新

apache のスキルをすべて見る

このスキルの問題を報告する