Files
shopdb-flask/tools/export-github.sh
cproudlock d8a2bfeb29 Publish the branch you are on, not always main
export-github.sh syncs the WORKING TREE, not a branch, and then committed
whatever it found onto the publication repo's main and printed
"git push origin main".

So running it while checked out on a site branch published that site's work AS
THE PRODUCT, to the repository every other site consumes, and said nothing. The
branch model made that reachable in one wrong command, from a checkout that
looks completely normal.

It now reads the current branch, mirrors it in the publication repo (creating it
from main the first time, as in the dev repo), bundles that branch, and prints
work-PC instructions naming it. Publishing anything other than main also prints
a reminder that other sites read main and these commits are not there yet.

The publication checkout uses --force, which is safe only because that tree is a
mirror: the rsync has already replaced it, so there is nothing local to lose. It
would not be safe anywhere else.

Also cleans up the publication repo: origin pointed at a bundle in a scratchpad
from a session that ended long ago, and two orphaned remote-tracking refs
(github/main, wilm/review-fixes) survived the removal of their remotes. None of
it broke the export, which never pushes, but a `git fetch` in there failed with
a confusing path.

Scope: product
(cherry picked from commit 81df9607e7)
2026-08-30 11:12:09 -04:00

274 lines
12 KiB
Bash
Executable File

#!/bin/bash
# Export the working repo to the GitHub publication repo and emit a bundle.
#
# Pipeline: working repo (full history, internal refs) -> pruned/scrubbed
# tree -> commit in ~/projects/shopdb-flask-pub (the local mirror of what
# enterprise GitHub holds) -> full git bundle in /home/camp/pxe-images/ for
# transfer to the work PC, which pushes it to GitHub.
#
# Usage:
# tools/export-github.sh "Commit message for the publication commit"
# tools/export-github.sh --dist # also rebuild BOTH frontend dists
# # (/ops = dev, /shopdb = prod)
#
# This script lives in tools/, which is itself excluded from publication.
set -euo pipefail
WORK=/home/camp/projects/shopdb-flask
PUB=/home/camp/projects/shopdb-flask-pub
OUT=/home/camp/pxe-images/github
BUILD_DIST=0
MSG=""
for arg in "$@"; do
case "$arg" in
--dist) BUILD_DIST=1 ;;
*) MSG="$arg" ;;
esac
done
[ -n "$MSG" ] || { echo "usage: $0 [--dist] \"commit message\""; exit 1; }
[ -d "$PUB/.git" ] || { echo "publication repo missing at $PUB"; exit 1; }
# --- 0. which branch is being published -------------------------------------
#
# This syncs the WORKING TREE, not a branch, and it used to commit whatever it
# found onto the publication repo's main and print "git push origin main". So
# running it while checked out on a site branch published that site's work AS
# THE PRODUCT, to the repository every other site consumes, without saying a
# word about it.
#
# The publication repo now mirrors the branch you are actually on.
BRANCH="$(git -C "$WORK" rev-parse --abbrev-ref HEAD)"
if [ "$BRANCH" = "HEAD" ]; then
echo "detached HEAD in $WORK - check out a branch before exporting" >&2
exit 1
fi
echo "==> publishing branch: $BRANCH"
# --- 1. sync the tree (working -> pub), minus everything never published ---
#
# NOTE the deploy/windows/installer/bundle exclude below. That is ~100MB of build
# output holding a COPY of the whole application tree plus the wheels and vendor
# installers. It is gitignored, but this sync walks the WORKING TREE rather than
# git, so it came through anyway - and its copies of config.py and
# requirements.txt then tripped the scrub gate. Anything else generated into the
# working tree needs excluding here too, for the same reason.
# rsync --exclude also PROTECTS a path from --delete, so anything that reached
# the publication tree before its exclude existed stays there forever, invisible
# to the sync and caught only by the scrub gate. Purge the generated paths first
# so adding an exclude is enough on its own.
#
# docs/ does NOT publish - the prose becomes the WIKI instead. One exception,
# and it is not a documentation decision:
#
# docs/openapi.json is read AT RUNTIME by shopdb/core/api/docs.py, which
# serves the /api/docs page every published site ships. Excluding it meant
# that page loaded and then answered {"error":"openapi.json not generated"}
# on every site installed from the published repo. It is a product file that
# happens to live under docs/, not documentation.
#
# It also un-skips tests/test_openapi_spec.py, which had been reporting green
# by not running - see scripts/check-gates-ran.sh.
#
# docs/api-inventory.json stays OUT: nothing reads it at runtime, so it is a
# development gate and belongs with the wiki. That does mean
# test_api_inventory_parity cannot run on GitHub, which is why the gate guard
# separates the checks that must run everywhere from the ones that only run
# where the repo is complete.
# instance/plugins.json is one site's plugin CHOICES. It is excluded, so it was
# never published - but an old copy sat in the publication tree for months,
# protected from --delete by that same exclude, and it made the published repo's
# own test suite fail locally: 13 plugins enabled where the code expects 16.
# A publication tree that cannot pass its own tests is not one anybody can
# verify before shipping.
for stale in .pytest_cache .ruff_cache htmlcov .coverage instance/plugins.json \
deploy/windows/installer/bundle deploy/windows/installer/Output \
deploy/windows/installer/plugins.iss deploy/windows/installer/version.iss; do
rm -rf "${PUB:?}/$stale"
done
# .env is excluded by name, but a stray copy BESIDE it is not: a hand-made
# .env.bak-<date> was picked up by this sync, because it walks the WORKING TREE
# and .gitignore does not apply to it. It got no further than the scrub gate,
# which is what that gate is for, but a file holding live credentials should
# never reach the publication tree at all. Hence --exclude '.env.*', with
# --include '.env.example' BEFORE it: rsync takes the first matching rule, and
# .env.example does publish.
#
# Keep flags on their own lines. A comment inside this backslash-continued
# command is not a comment - bash hands the text to rsync as arguments, and it
# dies with a usage dump rather than anything that names the real problem.
rsync -a --delete \
--exclude '.git' \
--exclude '.gitea' \
--include '/docs/' \
--include '/docs/openapi.json' \
--exclude '/docs/*' \
--exclude '/tools' \
--exclude '/mcp' \
--exclude 'start-api.sh' \
--exclude 'start-ui.sh' \
--exclude 'CLAUDE.md' \
--exclude 'frontend/CLAUDE.md' \
--exclude 'tests/test_docs_contract.py' \
--exclude 'tests/test_plugins/test_geenforce_parity.py' \
--exclude 'tests/test_plugins/test_zabbix_live.py' \
--exclude 'venv' \
--exclude 'node_modules' \
--exclude 'deploy/windows/installer/bundle' \
--exclude 'deploy/windows/installer/Output' \
--exclude 'deploy/windows/installer/plugins.iss' \
--exclude 'deploy/windows/installer/version.iss' \
--exclude 'frontend/dist*' \
--exclude 'frontend/src/.plugins-staged' \
--exclude 'frontend/src/router/routes.gen.js' \
--exclude '/instance' \
--include '.env.example' \
--exclude '.env.*' \
--exclude '.env' \
--exclude '__pycache__' \
--exclude '.pytest_cache' \
--exclude '.ruff_cache' \
--exclude '.coverage' \
--exclude 'htmlcov' \
--exclude '*.pyc' \
--exclude 'scripts/site_imports/wjf/idmap.json' \
"$WORK/" "$PUB/"
# --- 1b. every bundled plugin must have survived the sync ---
#
# The excludes above are rsync patterns, and a pattern with no leading slash
# matches at ANY depth. 'tools' was meant to drop the repo-root tools/ dir (this
# script lives in it) and silently dropped plugins/tools/ as well, so a plugin
# was committed, exported, bundled and deployed without ever being in the
# payload - the failure only surfaced as a missing manifest.json on the server.
# The excludes are anchored now; this check is what makes a recurrence loud.
missing=""
for manifest in "$WORK"/plugins/*/manifest.json; do
[ -e "$manifest" ] || continue
plugin=$(basename "$(dirname "$manifest")")
[ -f "$PUB/plugins/$plugin/manifest.json" ] || missing="$missing $plugin"
done
if [ -n "$missing" ]; then
echo "ERROR: bundled plugin(s) missing from the publication tree:$missing" >&2
echo " an rsync --exclude above is matching them; anchor it with a leading slash." >&2
exit 1
fi
cd "$PUB"
# Mirror the branch. The rsync above has already overwritten the tree, so this
# has to happen with the working tree ignored - checkout would otherwise refuse
# or clobber. --force is safe here precisely because the tree is a mirror: it
# holds nothing that is not about to be replaced from $WORK.
if git show-ref --verify --quiet "refs/heads/$BRANCH"; then
git checkout --force "$BRANCH" >/dev/null
else
# A new site branch starts from main, the same way it does in the dev repo,
# so its history is the product's history plus whatever it adds.
echo "==> publication repo: creating branch $BRANCH from main"
git checkout --force -b "$BRANCH" main >/dev/null
fi
# --- 2. re-apply the publication wording (idempotent) ---
# docs/ lives only in the wiki on the GitHub side.
# NOTE the `|| true` on each scrub grep below. grep exits 1 when it matches
# NOTHING, and under `set -euo pipefail` that failing pipeline aborts the whole
# script - silently, because grep prints nothing to say so. That is exactly what
# happened once every internal reference had been scrubbed: the gitea grep found
# no matches, the export died before committing or rebuilding the dists, and it
# exited without a single line of output. A scrub that finds nothing left to
# scrub is the SUCCESS case, so it must not be an error.
{ grep -rlZ 'docs/' --include='*.py' --include='*.md' --include='*.sh' \
--include='*.tmpl' --include='*.vue' --include='*.example' . 2>/dev/null || true; } |
while IFS= read -r -d '' f; do
case "$f" in ./CHANGELOG.md) continue ;; esac
sed -i -E \
-e 's/\[`?docs\/([A-Za-z0-9_-]+)\.md`?\]\((\.\.\/)*docs\/[A-Za-z0-9_-]+\.md\)/the \1 page in the project wiki/g' \
-e 's/docs\/proposals\/ge-enforce-plugin\.md/the ge-enforce-plugin proposal in the project wiki/g' \
-e 's/`docs\/([A-Za-z0-9_-]+)\.md`/the \1 wiki page/g' \
-e 's/docs\/([A-Za-z0-9_-]+)\.md/the \1 wiki page/g' "$f"
done
sed -i 's|<code>docs/GE-ENFORCE.md</code>|the GE-ENFORCE page in the project wiki|' \
frontend/src/views/geenforce/ManifestEditor.vue 2>/dev/null || true
# internal infra never named on GitHub; this repo's own URL maps to the
# real GitHub home, anything else degrades to a placeholder.
GITHUB_URL='https://github.com/ge-aero/shopdb-flask'
{ grep -rlZ 'gitea\.proudtech\.net' . 2>/dev/null || true; } | while IFS= read -r -d '' f; do
sed -i -e "s|https://gitea\.proudtech\.net/ge-aerospace/shopdb-flask|$GITHUB_URL|g" \
-e 's|gitea\.proudtech\.net|<git-host>|g' "$f"
done
{ grep -rlZi 'gitea' --exclude-dir=.git . 2>/dev/null || true; } | while IFS= read -r -d '' f; do
sed -i -e 's/the GE Aerospace Gitea/the internal GE Aerospace git server/g' \
-e 's/Gitea Actions CI/CI/g' \
-e 's/Gitea Actions/CI/g' "$f"
done
# CHANGELOG compare/release link definitions reference tags that do not
# exist on GitHub (history is squashed there) - drop them.
sed -i '/^\[[^]]*\]: .*\/\(compare\|releases\)\//d' CHANGELOG.md
# frontend/CLAUDE.md publishes under a neutral name
if [ -f "$WORK/frontend/CLAUDE.md" ]; then
cp "$WORK/frontend/CLAUDE.md" frontend/DEVELOPMENT-STANDARDS.md
fi
sed -i "s/rootpassword/changeme/g" shopdb/config.py 2>/dev/null || true
# --- 3. scrub gate: refuse to commit if anything internal leaks ---
# The site patterns are ADR-015's, and they belong here as much as the internal
# tooling names do: this repository is public and the product is multi-site, so
# one plant's server name or internal networks reaching it is both a disclosure
# and a lie to every other site. They were absent, and a bundled plugin shipped
# a production UNC path to GitHub for a fortnight before anyone noticed.
LEAKS=$(grep -rlIiE 'claude|anthropic|fable 5|gitea|proudtech|home/camp|rootpassword|tsgwp00525|wjs\.geaerospace\.net|10\.134\.48\.|10\.48\.249\.' \
--exclude-dir=.git . || true)
if [ -n "$LEAKS" ]; then
echo "SCRUB GATE FAILED - internal references in:"; echo "$LEAKS"; exit 1
fi
# --- 4. commit (no-op safe) ---
git add -A
if git diff --cached --quiet; then
echo "no changes vs publication tree - nothing to export"
else
git commit -m "$MSG"
fi
git log --oneline -3
# --- 5. bundle for transfer (full bundle: stateless, fetch takes only new) ---
# Carries the branch that was published, not a hardcoded main.
git bundle create "$OUT/shopdb-flask-pub.bundle" HEAD "$BRANCH" --tags
git bundle verify "$OUT/shopdb-flask-pub.bundle" >/dev/null
echo "bundle: $OUT/shopdb-flask-pub.bundle"
# --- 6. optional per-instance dists (both bases: /ops = dev, /shopdb = prod) ---
# Two instances run on the box; each needs its own base-path build. Always
# rebuild BOTH so prod never ships a stale frontend.
if [ "$BUILD_DIST" = 1 ]; then
cd "$WORK/frontend"
for base in ops shopdb; do
VITE_BASE_PATH=/$base/ npm run build --silent
rm -rf "$OUT/frontend-dist-subpath-$base"
cp -r dist "$OUT/frontend-dist-subpath-$base"
echo "dist: $OUT/frontend-dist-subpath-$base/ (base /$base/)"
done
fi
cat <<EOF
On the work PC (in the shopdb-flask-pub clone):
git checkout $BRANCH
git fetch <path-to>\shopdb-flask-pub.bundle $BRANCH
git merge --ff-only FETCH_HEAD
git push origin $BRANCH
EOF
if [ "$BRANCH" != "main" ]; then
cat <<EOF
NOTE: this published $BRANCH, not main. Other sites read main, so nothing here
reaches them until these commits are promoted:
tools/wj-status.sh
tools/promote-to-main.sh <sha> ...
EOF
fi