Two defects in the stage 0 bootstrap, both surfacing as "Access denied" on a
server where the operator was holding the correct password.
CREATE USER IF NOT EXISTS is a no-op on an existing user - it does NOT change
the password. Stage 0 generates a fresh password every run and overwrites
.dbpass with it unconditionally, so any path that re-runs the bootstrap over an
existing account left the handoff holding a password the server had never been
told. ALTER USER now follows each CREATE, so the stored password and the handoff
always agree.
The user was also only created for 'localhost' and '127.0.0.1'. On current
Windows, 'localhost' resolves to the IPv6 loopback FIRST, so an operator who
types localhost rather than 127.0.0.1 arrives as '<user>'@'::1' - an account
that did not exist - and MySQL answers "Access denied" naming a host they never
typed. The ::1 account is now created and granted alongside the other two.
Note the datadir guard means the first defect could not fire on a straightforward
re-run - stage 0 refuses a non-empty data directory before reaching the
bootstrap. It was still wrong, and reachable once the directory has been cleared
by hand, which is what the failure message tells operators to do.
Asked why the password box does not pre-fill from .dbpass. It should not - but
it should not have been demanding a password either.
.dbpass is the ACL'd handoff stage 0 writes when it creates the database itself,
and stage 2 already reads it automatically when no password is supplied. The
wizard, though, required a password whenever .env was absent, without checking
for the handoff. On a server where stage 0 had completed but stage 2 had not -
which is exactly what a partly-failed install leaves - the operator was blocked
on a secret the installer already had, and sent hunting for a generated password
they were never meant to handle.
Blank is now accepted when either .env or .dbpass is present, and the sign-in
page says so when it sees a handoff.
Deliberately NOT pre-filled into the password box, for two reasons. It is the
only copy of a generated password, so round-tripping it through a UI control and
back out through a temporary password file adds exposure for no benefit - stage 2
reads the file directly. And .dbpass belongs to the BUNDLED database; on the
existing-database page the operator is pointing at someone else's server, where
a locally generated password is simply the wrong answer.
Trimming the description brought the Username box back and left Password off the
bottom. CreateInputQueryPage stacks its fields below the description and neither
scrolls nor shrinks, so a field that does not fit is drawn past the surface and
simply never appears - no error, no scrollbar. Sizing the description against a
pixel budget that varies with DPI and font scaling is guesswork, and it had now
failed twice.
Connection details (host, port, database) and sign-in (username, password) are
now two pages of three and two fields. Both fit under any reasonable
description, at any scaling, without anyone having to estimate.
The upgrade hint about leaving the password blank moves to the sign-in page,
where the password field actually is. ShouldSkipPage hides both pages for the
bundled-database option, and the stage arguments read the values from their new
homes.
Reported from the Server 2019 test: the Existing database page rendered a
truncated "Username:" label and no input boxes at all below it, so there was no
way to enter connection details.
CreateInputQueryPage lays its fields out BELOW the description text. The
description had grown to include a five-line CREATE DATABASE / CREATE USER /
GRANT block, added so a DBA could be handed the exact SQL. With five fields
underneath, the last two fell past the bottom of the page surface, where they
are simply not drawn - no error, no scrollbar, just missing controls.
The description is back to three lines. The SQL moves to
docs/INSTALL-WINDOWS.md, which is where someone would look for it anyway and
where it can be copied without being retyped from a wizard page.
Wizard page descriptions are a fixed budget: anything long enough to be worth
reading twice belongs in the guide, not on the page.
A Server 2019 install reported all 96 payload files as simultaneously missing
and unexpected, with mangled names - wheels/heels/flask.whl,
python/ython/python-3.14.6-amd64.exe, mysqlclient/lient/mysql.exe. Exactly five
characters of each directory name survived, which is the difference between
ADMINI~1 and Administrator.
Inno extracts the bundle under C:\Users\ADMINI~1\AppData\Local\Temp\..., an 8.3
SHORT path. Resolve-Path kept that short form while Get-ChildItem returned the
long one, so the root was five characters shorter than the prefix being sliced
off every FullName, and every relative key came out wrong. The payload was
correct; the comparison was not - the verifier refused a perfectly good bundle.
The root now comes from Get-Item, which goes through the same provider as
Get-ChildItem so their path forms agree, and the prefix is checked with
StartsWith before being trimmed. If the two ever disagree again this throws
instead of inventing paths.
Verified against the real failure mode rather than assumed: running the check
through C:\SHOPDB~3\bundle in a Windows VM now passes.
Nothing on Linux or in a normally-pathed Windows directory could have caught
this - the short name only appears under a profile directory long enough to need
one, which is where Setup extracts.
Second failure from the Server 2019 test. The previous fix worked - msiexec went
from exit 1639 (ERROR_INVALID_COMMAND_LINE, which is why it printed its usage
dialog) to exit 1603 (ERROR_INSTALL_FAILURE), so the command line parses now and
the MSI itself is failing.
It failed in 1.1 seconds. An MSI that dies that fast has not begun installing;
it has failed a launch condition. MySQL 8.4 requires the Visual C++
redistributable and a bare Windows Server does not ship it - the same runtime
mysql.exe and mysqldump.exe import, which was visible when their DLL
dependencies were trimmed and went unnoticed.
Stage 0 now installs VC_redist.x64.exe from the bundle before touching MySQL,
skipping it when vcruntime140.dll is already present, and fails with a sentence
naming the requirement if the redistributable is absent from the bundle
altogether. vcredist\ is an optional locked payload.
msiexec also gets /l*v now. A bare 1603 names neither the failing action nor the
reason, and it is the most common MySQL install failure - diagnosing this one
took a launch-condition inference rather than a log. The MSI log lands beside
the installer's own in ProgramData, so the next failure is readable instead of
guessed at.
Reported from a Windows Server 2019 test: a "Windows Installer" dialog listing
every msiexec /Option appeared, then the wizard reported that the bundled MySQL
database could not be installed. That dialog is msiexec's usage help - it prints
it when the command line does not parse - so the install never started.
Cause: $MysqlRoot defaulted to 'C:\Program Files\MySQL\MySQL Server 8.4', which
contains spaces. Invoke-Native wraps any argument containing whitespace in
quotes, producing "INSTALLDIR=C:\Program Files\...". msiexec takes public
properties as PROPERTY=value and expects the VALUE quoted -
INSTALLDIR="C:\Program Files\..." - so it rejected the line, printed usage, and
exited non-zero.
This file already carried the rule, next to the Python target: "Never put a
space in a path this installer controls." I broke it setting the 8.4 path.
Two fixes, because one of them alone leaves the trap in place:
- $MysqlRoot is now C:\MySQL84, space-free like C:\Python314. The MySQL client
search paths in the installer, the preflight and the operator console all look
there first, keeping backups working against the bundled server.
- Invoke-Native now quotes PROPERTY=value correctly, so passing a spaced path
explicitly no longer produces an unparseable command line.
tests/test_installer_defaults.py fails if an installer-controlled path default
ever contains a space again.
The preflight page began refusing to continue while any check was failing, which
is right for something the operator must go and fix. HttpPlatformHandler was
marked FAIL when absent - so on a server without it the wizard stopped dead,
telling the operator the server was not ready, over a module the bundle carries
and stage 4 installs a few pages later. The only way forward was to go and
install by hand the exact thing the installer was about to install.
It is now INFO: reported, not blocking, matching how URL Rewrite is already
handled. Nothing the installer SUPPLIES may block the wizard, and
tests/test_installer_defaults.py now fails if that rule is broken again.
The site-port conflict check is downgraded from FAIL to WARN for the same class
of reason: it runs before the operator reaches the Address page, so it tests the
DEFAULT port rather than the one they intend to use, and blocking refuses an
install over a conflict the very next page lets them resolve.
Genuine blockers are unchanged - no IIS, no WebAdministration, wrong Windows
edition or architecture, no disk, and the MySQL 5.6 index flags. Those the
operator really does have to fix first.
Adopting a site means getting its asset register in. The HTTP import API suits a
site with a source system and someone to script against it; a sister site with a
spreadsheet and no developer needs something else, and that is the common case.
FOREIGN KEYS TAKE NAMES. This is the whole design. A CSV row has to say where an
asset is, and the database stores locationid, an integer. Requiring the number
means importing locations, reading back the generated ids and pasting them into
the asset sheet - a workflow nobody finishes. Every foreign key here accepts
either a numeric id or the referenced row's name:
assetnumber,assettypeid,statusid,locationid
CMM-01,Measuring Tool,Active,Gage Lab
The column keeps its database name, per CONTRIBUTING.md; the value is whatever
the operator actually knows. Names resolve across files in one run, so
assets.csv can reference a location that only exists because locations.csv was
read moments earlier. A name that does not resolve is reported with its line,
column and value, not as a foreign key violation from three layers down.
Dry run is the default, and writes go into the transaction either way - the
rollback is what makes it a dry run. Skipping the writes instead made every
cross-file reference fail, which is the one thing a folder-wide check exists to
verify. Validation covers every row before anything is written, so a typo on
line 400 cannot leave 399 rows imported. Files are matched on a natural key, so
correcting a spreadsheet and re-running updates rather than duplicates.
TEMPLATES ARE GENERATED, NOT MAINTAINED. "flask csv templates" builds them from
the live schema, annotated with required/optional and which file each foreign
key refers to. The prompt for this was a hand-written template set that had
invented columns on seven of eleven tables and named a table that does not
exist, while looking entirely plausible - and described an import mechanism
(a Data Import page, a flask import-csv command) that had never existed. A test
fails the build if a generated template ever offers a column the schema lacks.
User accounts are deliberately not importable: passwords do not belong in a
spreadsheet in either direction.
Verified end to end against MySQL 5.6 - a folder dry run catching one bad
reference, the fix, the commit, and a re-run reporting updates rather than
inserts. 16 tests.
The bundled-database option could not actually be built. Stage 0 looks for
mysql\mysql-8.0.x-winx64.msi, and Oracle no longer publishes a standalone server
MSI for 8.0 - every 8.0.x returns 404. What remains for 8.0 is the MySQL
Installer bundle, which is an installer-manager: 'msiexec /i INSTALLDIR=' would
install THAT rather than a database, and stage 0 would then fail on a missing
mysqld.exe.
MySQL 8.0 also reached end of life in April 2026, so bundling it would have put
an unsupported database on every new site.
8.4 LTS still ships the standalone MSI (129MB, which is what the '125MB' note in
stage 0 was written against) and is supported into 2032. Defaults follow it:
install root MySQL Server 8.4, service MySQL84. The operator console still looks
for an 8.0 install path as a fallback, for sites already running one.
Also bundles mysqlclient\ - mysql.exe and mysqldump.exe with the two OpenSSL
DLLs they actually import, 20MB rather than the 51MB of debug and auth-plugin
libraries the archive ships. Stage 2 stages it onto the server, so a site whose
database is on ANOTHER host can still take the pre-upgrade backup that every
upgrade depends on. That was the gap the preflight had started warning about.
Bundle is now 221MB.
The GitHub backend job failed on test_there_are_docs_to_check, correctly. docs/
is stripped from the published repository - it lives in the wiki on that side -
so on the mirror the glob matched nothing and the guard fired exactly as
designed.
An absent docs/ and a glob that silently matches nothing in a tree that HAS docs
are different conditions, and the test conflated them. The module now skips when
the directory is not there at all, and the guard still fails when it is there and
empty. Verified all three ways: 7 pass here, 7 skip in a docs-less checkout, and
the guard still fails against a docs/ containing no markdown.
The gate has to ship rather than be excluded from publication, because the
published tree is where the GitHub CI that would catch a regression runs.
rsync --exclude also PROTECTS a path from --delete, so anything that reached the
publication tree before its exclude existed stayed there permanently - invisible
to the sync and surfacing only as a scrub-gate failure. That cost two rounds of
'add the exclude, still fails' on the installer bundle and again on
.pytest_cache. The generated paths are now purged before the sync, so adding an
exclude is sufficient on its own.
tests/test_docs_publishable.py assembles its search terms from fragments: a file
containing the literal strings the scrub greps for tripped that scrub on itself.
Excluding the file from publication would have removed the check from the
repository it protects.
docs/ is excluded from the code bundle and its scrub gate, because it goes to
the GitHub wiki instead - via a generator that has no gate at all. So the one
part of the repository written in prose, by people, about internal
infrastructure, was the one part nothing checked.
What was reaching a public wiki: the internal git server's URL and hostname,
.gitea workflow paths, developer home directories in the GE-Enforce cutover
reference, and a dev database root password inside a copy-pasteable command in
the import guide.
All replaced with neutral equivalents. tests/test_docs_publishable.py is now the
gate, at the source, in CI - a wiki page cannot be un-published, so catching this
after the fact is not good enough.
PROJECT-REVIEW.md also referred to internal tooling by name throughout; those
references are generalised. It remains an internal candid assessment of this
project that is nonetheless published, which is worth a separate decision.
The sync walks the working tree rather than git, so deploy/windows/installer/
bundle came through despite being gitignored - about 100MB of build output
containing a copy of the whole application tree, the wheels and the vendor
installers. Its copies of config.py and requirements.txt then tripped the scrub
gate, which is the only reason it was noticed.
Excluded along with the installer's other generated files. Note that rsync
--exclude also protects a path from --delete, so a copy already in the
publication tree has to be removed by hand once.
The '# via' annotations recorded the absolute path of the temporary file the
lockfile was compiled from, which is meaningless to anyone else and does not
belong in a published artifact. They now read 'requirements.in', which is where
these requirements actually come from.
Pins and hashes are unchanged - verified by a hash-checked dry-run install.
CLIENT IP / SPOOFABILITY. docs/geenforce-api-cutover.md claimed that removing the
IIS rewrite rule made the allowlist fail closed and that it does NOT become
spoofable. The opposite is true. IIS never sets X-Forwarded-For on its own; the
rule is the only thing that does. Remove it and IIS still forwards whatever
X-Forwarded-For the CALLER sent, waitress trusts it because it arrives from
127.0.0.1, and remote_addr becomes attacker-controlled - so a token-less caller
can fetch manifests from anywhere on the network. The document and the
_trusted_client_ip docstring now say so, waitress runs with
--trusted-proxy-count=1, and stage 5 checks the rule is actually live rather than
assuming it. The wizard question is rephrased to something an operator can verify
with their network team instead of guessing at.
NON-ASCII. The style gate only ever checked .py/.vue/.js/.ts, so documentation
accumulated em-dashes, arrows and box-drawing characters against this repo's own
convention - including in files added this week. Cleaned, and the gate now uses
INCLUDES_ALL so Markdown, JSON and YAML are covered.
PLUGIN DEFAULTS. The wizard pre-ticked measuringtools and printedparts, both of
which ship default_enabled=false, so every site taking the defaults installed and
enabled them against their manifests. Inno has no JSON parser so the list must be
hardcoded, but tests/test_installer_defaults.py now fails when it drifts.
UPGRADES. The payload copy merges, so a plugin dropped from a site's profile kept
its code forever - which defeats a lean build and leaves core's optional-import
guards succeeding for a plugin the site no longer has. Stale plugin directories
are now deregistered and removed before the copy.
add-plugin used 'plugin install', which for the five default_enabled=false
plugins left them installed but DISABLED - and printed a green success line
anyway. It now goes through apply-profile, and the success line is gated on the
exit code. Invoke-Flask records its own exit status, because $LASTEXITCODE keeps
a stale value when flask.exe is missing and no native command runs.
CHARSET. The utf8mb4 compiler hook lived inline in migrations/env.py, so it
covered the CORE chain only: plugin baselines inherited the server default, which
on a latin1 server means two charsets in one database. It is now
shopdb/utils/mysql_charset.py, imported by both, and preflight reports the
database's default charset.
BACKUP HONESTY. The dump was described as 'all of your asset data'. Uploaded
branding and floor-map images live in instance\ on disk, not in the database, so
a restore from the .sql alone comes back with no map. backup now archives
instance\ alongside it and says both are needed.
VERSIONING. AppVersion was hardcoded at 0.9.0 while the product, the frontend and
the newest tag said 0.7.0 - and 0.9.0 collides with a retired contract version.
Both builders now generate version.iss from shopdb/__init__.py.
Smaller: rollback overwrites .env before deleting it, as uninstall already did;
appcmd unlocks are scoped to this site's location rather than server-wide, with
the wide unlock as a fallback; DEVELOPMENT-SETUP says Python 3.14; the README
plugin list gains printedparts; prune-schema --force is documented as
first-provisioning-only; HTTPS is documented as not-the-default with the steps to
add it; the DBA SQL is on the wizard's database page; the features page says
unticking does not remove an installed feature; and the installer README states
that bundle-lock cannot vouch for the exe itself - that needs signing or an
out-of-band hash, neither of which is wired up.
Nine fixes from a review of the installer against its actual audience: DT leads
at sister sites who are not Windows, IIS or Python specialists and who will lean
on an AI assistant to get through it.
TRUTHFULNESS. The preflight was advisory - an operator read 'IIS is not
installed', pressed Next, answered five more pages and the install died partway
through with Python already on the box. The results page now blocks while
anything is failing, repaints on every run instead of latching after the first,
and offers 'Check again' so a fixed problem does not mean starting over. On
failure the wizard said 'Nothing was left running', which is false in every path
because the stages run with -OnFailure never: it now says the server is
part-configured, that re-running is safe, and how to remove it. The final page no
longer reads 'ShopDB-Flask is ready' after a failed install.
SECRETS. The generated MySQL root password went to Write-Host in a process the
wizard runs hidden - so nobody saw it - and stdout is forwarded into the setup
log operators are told to send to support, so it was permanently recorded for
everyone who did not need it. It now goes to an ACL'd file. Database dumps, which
contain every user password hash, landed in a ProgramData directory readable by
every user on the box; the directory is now locked at creation.
UPGRADES ON REMOTE-DATABASE SITES. mysqldump was looked for only under local
MySQL install paths, so a site whose database is on another host silently skipped
every pre-upgrade backup - after stage 2 had already stopped the pool and
replaced the tree. Find-MysqlTool now prefers a client shipped in the bundle,
stage 2 stages it onto the server, preflight reports when it is missing, and
mysqlclient\ is an optional locked payload.
UNINSTALL. A subpath install is an IIS Application, not a site; removing only the
site left the application pointing at a deleted directory, so the parent site -
at West Jefferson, the live classic ASP - served 503 on that path forever while
Add/Remove Programs reported success. Uninstall now reads MOUNT_PATH and removes
the application. The firewall rule was created as "$SiteName $SitePort" and
removed as the literal 'ShopDB-Flask 8090', which matches nothing.
DAY-2 TOOLING. Every shortcut now passes -AppRoot and -SitePort, and the console
forwards them through its own elevation and 32-bit relaunches instead of
discarding them - a non-default directory or port made it report a healthy site
as broken, from a shortcut the installer wrote. 'Open ShopDB-Flask' resolved to a
hardcoded localhost:8090 that was wrong for every subpath install; it now asks
the console, which reads the address the installer recorded, and no longer
demands administrator to open a browser.
SMOKE TEST. The parent-site port lookup filtered for an http binding and
defaulted to 80, so an https-only parent site failed a working install with a red
dialog.
DOCS AND /api/docs. The installer was invisible: nothing in docs/, README.md or
CLAUDE.md mentioned it, so a DT lead or their assistant landed on the manual IIS
runbook and hand-built the very server the installer then refuses to upgrade.
docs/INSTALL-WINDOWS.md and docs/OPERATE-WINDOWS.md are now the canonical route,
the two manual runbooks are bannered as reference-only, README and CLAUDE.md
route by target, and llms.txt tells an assistant which document to follow and to
ask for 'check -Json' before diagnosing. Both ship on the server, along with
openapi.json and llms.txt - without those the self-hosted /api/docs was broken on
every installed box, which matters most to the sites least able to debug it.
Stage 5 now checks it actually serves.
shopdb-admin.ps1 gains 'check -Json': one structured, secret-free block covering
version, publishing method, IIS state, HTTP reachability, database, Python
version, plugins and errors. That is the cheapest useful answer to 'the operator
will ask an LLM' - it works with no infrastructure, which a install-time MCP
server could not.
Stage 4 decided whether the IIS objects it was about to reconcile were its own by
testing for .installed-version. Stage 2 writes that file, and stage 2 always runs
first in a '-Stage all' install - so by the time the guard looked, the stamp it
had just written made every server look like one this installer built, including
the hand-built ones the guard exists to protect.
Stage 2 now records whether a stamp was present BEFORE it writes its own, and
stage 4 reads that observation. Running stage 4 alone still tests the file, which
is correct there: no stage 2 has run to disturb it.
Found by review, not by test - the guard has no coverage, because exercising it
needs a live IIS.
An air-gapped site cannot be scanned from anywhere else, so when a CVE lands the
only way to answer 'is that component here, and at what version' was to RDP in
and go looking. The frontend was the real blind spot: nothing recorded which
version of leaflet, dompurify, jspdf or html2canvas ends up inside the compiled
SPA.
scripts/generate_sbom.py emits CycloneDX 1.6 covering both ecosystems - every pin
in requirements.txt with the sha256 the installer enforces, and every package in
package-lock.json. Build-only npm packages are marked scope 'excluded' rather
than dropped, so 'not here' stays distinguishable from 'not looked for'.
Dependency edges are real: uv's '# via' comments give the Python graph and
package-lock gives the npm one.
Hand-rolled rather than cyclonedx-py plus cyclonedx-npm because both inputs are
already pinned and committed - this is a format translation, not a scan - and
because the build box may be a work PC with nothing but Python and Node. It is
deterministic by construction: same inputs, byte-identical output, so
regenerating does not churn.
Staged into the application tree by both builders, so it installs onto the
server with the app. shopdb-admin.ps1 verify reports it and searches it by
component name, which is the question actually being asked.
Packages appearing at several depths in package-lock (node_modules/vite and
node_modules/vitest/node_modules/vite) are merged, and a copy reachable outside
the dev tree makes the component count as shipped. Emitting both produced
duplicate bom-refs, which CycloneDX forbids and scanners reject; getting the dev
merge backwards would have hidden a shipped package from a CVE search.
Not covered by bundle-lock.json on purpose: its provenance is git, not the
third-party payload.
Two guards for a server deployed by hand, which the West Jefferson production
box is.
An existing venv is reused, which is right for a repair or an upgrade of an
install this made, and wrong when the venv belongs to a different Python. The
wheelhouse is tagged for one minor version, so pip finds no candidate for the
compiled packages and dies partway through - after Python has been installed and
the application tree replaced. The two versions are now compared up front and
the run stops with both numbers and what to do about it.
Switching deployment method removes the other method's IIS artifact. That is
correct when this installer owns both and dangerous when it does not: a wrong
-MountAlias would call Remove-WebApplication on a live mount with no prompt and
no error, and the first sign would be the site returning 404. It now refuses
unless a version stamp shows this installer made the install, or -AdoptExisting
is passed, and the refusal lists exactly what it would have removed.
The lock records what IS in the wheelhouse, not what the application NEEDS, so
an incomplete wheelhouse was locked, blessed and shipped - and only failed on an
air-gapped server.
That is not hypothetical. Assembling the wheelhouse anywhere other than Windows
silently omits colorama, a win32-only dependency of click, because pip evaluates
environment markers against the machine doing the downloading rather than the
machine being targeted. The bundle built here was short exactly that one wheel.
Both verifiers now cross-check wheels/ against the staged requirements.txt,
ignoring markers, since a requirement guarded by sys_platform == 'win32' is
precisely the one that must be present. Names are normalised to PEP 427 wheel
form, so mysql-connector-python matches mysql_connector_python.
bundle-lock.json is the first real lock: 42 files, cp314/win_amd64 - 39 wheels,
Python 3.14.6, HttpPlatformHandler 1.2 and URL Rewrite. MySQL is absent and
optional; a site choosing the bundled-database option adds it and re-locks.
The naming gate now skips the installer's build output. It contains a staged
copy of the application plus a second SPA build under dist-subpath, which
--exclude-dir=dist does not match, so a staged bundle failed the gate on
vendored minified JS nobody in this repository wrote.
IIS does not set X-Forwarded-For on its own and HttpPlatformHandler connects
from loopback, so without a rewrite rule every client reads as 127.0.0.1. The
GE-Enforce IP allowlist, the dashboard visitor-location lookup and per-host
login rate limiting all stop working, silently. The rule needed URL Rewrite,
which the installer told operators to download - from an air-gapped server.
URL Rewrite now ships in the bundle, and the wizard asks which case applies,
because the two answers are mutually exclusive. Directly exposed: install it and
set X-Forwarded-For from REMOTE_ADDR, which is what stops a client spoofing its
own. Behind a proxy: leave the rule off, since REMOTE_ADDR is the proxy and
applying it would discard the real client IP.
The rule is enabled by deleting two explicit marker lines rather than by a regex
over the surrounding comment, so editing that prose cannot silently disable it.
An existing web.config is no longer overwritten. It is the one file on a server
that legitimately carries hand-edits, and replacing it reverted them without a
word - on a server where the X-Forwarded-For rule had been enabled by hand, that
alone would have turned the GE-Enforce IP allowlist off. The installer reports
what it found instead.
pip now runs with --require-hashes and --only-binary=:all:. Hash-checking is
requested explicitly rather than inferred from the lockfile, so shipping an
unhashed requirements.txt fails loudly instead of quietly dropping the check.
shopdb-admin.ps1 gains a verify command: which bundle this server was installed
from, and whether the installed packages still match what shipped.
The .iss states its compiler floor. WizardStyle uses the built-in windows11
custom style, which needs Inno Setup 6.6.0; older compilers now fail with that
sentence rather than 'WizardStyle is invalid'.
The bundle carries ~40 wheels, a Python installer and two MSIs. All of them run
as SYSTEM on the target server, and nothing verified any of them. A missing
wheelhouse printed MISSING and the script still exited 0, so an empty bundle
compiled into a shippable installer and the failure surfaced on an air-gapped
server with no way to fix it.
bundle-lock.json now records that payload exactly - sha256 and byte size per
file - and verification is set equality: a missing file, an unexpected extra
file, or changed content all fail. Both builders check it and refuse to produce
an unverified bundle; the lock ships inside the bundle and shopdb-install.ps1
re-checks it on the server before running any of it.
This is deliberately a layer above requirements.txt hashes. pip lists every
artifact of a pinned version (cffi 2.1.0 alone has 100 hashes), so it proves a
wheel is genuine, not that it is the wheel this bundle was built and tested
with; it ignores extra files in the wheelhouse; and it covers none of the
executables.
refresh-bundle-lock.ps1 regenerates the lock but refuses to overwrite one until
the operator has seen the diff, because the commit is the review - it is the
only place a change to what runs as SYSTEM becomes visible to a human.
build-installer.ps1 is the whole build natively on Windows, so a work PC needs
no Bash. It shares the plugin closure resolver with build-site.sh.
Both builders now copy the installer scripts from the repository. They were
copied from a downloads folder, so the logic that shipped was not the logic that
was committed and the build worked on exactly one machine.
Two verifiers exist because PowerShell is the only thing guaranteed present on
the target server, while the Linux builder should not need pwsh.
tests/test_bundle_lock.py runs both against the same fixtures and fails if they
disagree.
Two fixes to the lean-site build.
The closure resolution moves out of an inline heredoc into
scripts/resolve_plugin_closure.py. The Windows builder needs the same answer,
and a PowerShell reimplementation would have been a second copy of the rules,
free to drift and produce a bundle whose plugin set did not match its profile.
The backend staging step copied all of deploy/ into the output tree. The Windows
installer stages its bundle at deploy/windows/installer/bundle, so that copy
recursed into its own destination and cp aborted with 'cannot copy a directory
into itself' - the documented build could not complete. Only
deploy/windows/web.config is read at install time, so only that is staged; the
rest of deploy/ is installer source and does not belong on an application
server.
test_plugins_only_import_contract_surface has been failing on main since
9a2d0cc: the employee name resolver imported shopdb.core.models directly.
shopdb.api already exports User (contract 0.13.0), so this is the same object
reached the way ADR-001 requires.
Both files are recompiled with --universal --generate-hashes, preserving every
pinned version. Three things change.
Hashes put pip into hash-checking mode, so a wheel whose sha256 is not listed is
refused rather than installed. The offline Windows install previously took
whatever file in the wheelhouse satisfied the version pin.
--universal means one lockfile serves Linux (dev, Docker, CI) and the Windows
wheelhouse. The Linux-only resolve had silently omitted colorama, a win32-only
dependency of click; in hash-checking mode a missing entry is a hard error, so
that omission would have broken every Windows install.
requirements-dev.txt is now compiled with -c requirements.txt, pinning shared
dependencies to the versions production runs. The two had been compiled at
different times and drifted: CI tested against alembic 1.18.5 while sites
installed 1.18.4.
Hashes pin the version and prove the artifact is one upstream published. They do
not pin WHICH artifact of that version is used, and they say nothing about extra
files in the wheelhouse - bundle-lock.json covers both.
Vite compiles the mount path into the bundle, so it cannot be chosen at install
time from a single build - a page served under /shopdb would load and then
request its assets from /assets/, and render nothing.
build-site.sh now produces both:
frontend-dist base / - the app on its own IIS site
frontend-dist-subpath base /<alias> - an IIS Application under an existing
site, e.g. http://<server-fqdn>/shopdb/
SUBPATH_ALIAS (default 'shopdb') is fixed per bundle and written into the staged
build as .alias, so the three places that must agree - the IIS application alias,
MOUNT_PATH in .env, and this compiled base - cannot drift apart. The installer
checks that marker and refuses rather than serving a page that cannot load.
The subpath build runs FIRST and is held in a temp dir: the root build has to be
last so frontend/dist is left in the state a developer expects, and the copy into
$OUT has to happen after the staging step that does rm -rf "$OUT".
Roughly 2500 lines of tested installer had been living in ~/Downloads and an
untracked folder - nothing was under version control.
It goes here rather than in a repo of its own because it depends on application
internals: the `flask plugin` verbs, site-profile.json, MOUNT_PATH, and the
plugin registry. Versioned separately it would drift out of step with the thing
it installs.
Contents: the read-only preflight, the staged installer (bundled MySQL, runtime,
schema, IIS, verify, uninstall), the operator console, the Inno Setup wizard, the
bundle builder and the artwork generator.
bundle/ and Output/ are ignored - regenerable, and ~220MB. plugins.iss is ignored
because build-installer.sh generates it from the staged payload. The artwork IS
committed so a Windows build box does not need Python and cairosvg.
Verified end to end on Windows Server 2025 against a bundled MySQL 8.0 and an
existing MySQL 5.6: fresh install, upgrade with backup and rollback, re-run
idempotency, uninstall, and both deployment methods including switching between
them. Not yet verified: a hypervisor-level air-gapped run, and any load from a
real browser (every HTTP check so far used curl, which sends no Origin header).
The step was called "Starter Data" and offered to "seed the data a new site
needs", but seed_starter inserts eight vendor rows and nothing else: no assets,
locations, departments or statuses. An operator ran it, saw every dashboard
count stay at zero, and reasonably concluded the seed was broken.
Rename the step to Reference Data, describe what is actually seeded, and say
outright that no assets are created and that an empty dashboard is expected
here. Assets arrive later via the import API.
build-site.sh staged only shopdb/, the chosen plugins/ and frontend-dist, so the
output could be imported but not run or migrated. The Windows installer had to
assemble wsgi.py, requirements.txt, migrations/ and deploy/ separately, which
meant it could assemble a payload whose plugin set did not match the profile the
tree was staged from.
Stage those runtime files, and copy the profile in as site-profile.json so the
set is self-describing: `flask plugin apply-profile` at provisioning reads the
same profile the tree was built from, so installed plugins and shipped plugin
code cannot drift.
frontend-dist keeps its name; CI reads that path (ci.yml:79).
The closing hint now spells out `prune-schema --yes --force`. ADR-014's prose
says lean provisioning "uses --force", but --force alone only permits dropping
non-empty tables; without --yes the command is a dry run that prints a preview
and exits, so following the ADR literally silently skips the prune.
A fresh install landed on the anonymous dashboard instead of prompting to create
the first admin, so an operator had no way to discover /setup.
The router now asks /api/setup/needs-admin before rendering any unauthenticated
route and redirects to /login?firstrun=1 while no user exists. The result is
cached in a composable so it costs one request per session, and the lookup fails
open (a backend that cannot answer must not lock the login screen). Racing it
against a 4s timeout keeps a slow or hung backend from blocking the first paint.
Login.vue clears the flag after creating the admin so the gate stops firing
without a reload.
REQ-D: restore waitress and tzdata to requirements.in. They existed ONLY in the
generated requirements.txt (hand-added in bf9e60e), so the next
`uv pip compile` would have silently removed the WSGI server and the IANA
timezone database from every Windows install.
REQ-E: split production and development requirements. requirements.txt was
installing pytest, pytest-cov, pytest-flask, coverage, iniconfig and pluggy onto
production servers. Verified on a real Windows Server box before this change.
CI, scripts/test-external-plugin.sh and the dev docs now use requirements-dev.txt.
REQ-F: standardise on Python 3.14. The repo declared four different versions
(Dockerfile 3.12, DEPLOY-WINDOWS-IIS 3.12, INSTALL-WINDOWS-IIS 3.13, CI 3.13,
plus README, web.config and PLUGIN-EXTERNAL-REPO). 3.14 is in active bugfix
support until ~Apr 2027 and supported to Oct 2030; 3.13 entered security-only in
Apr 2026. All four compiled dependencies publish win_amd64 wheels for 3.14
(cryptography via an abi3 wheel), verified by building an offline wheelhouse and
installing it on Windows Server 2025.
REQ-G: state MySQL 8.0 as the standard for new installs; 5.7+/5.6 remain
supported on an existing server.
Lockfiles regenerated with uv pip compile. Production deps 44 -> 38.
Each marker bound both a hover tooltip and a click popup. On the map page the
click handler routes to the detail page, so the popup opened and the
navigation discarded it in the same tick - it was never visible. Where it did
render (the map editor and the picker forms) its 'View Details' link only
served to pull the user off an unsaved form.
Keep hover as a glance and leave the click to the consumer. Removes the popup
markup, its styles, and the two now-unused detail-route helpers.
The ring only has to separate the mark from the surface; at 2px it read as
part of the mark. Markers go to 1.25, legend dots to 1.5px (they are larger,
so the same visual weight needs slightly more), and the PDF marker and swatch
strokes drop to match.
Two independent ways a restart leaves the SPA unable to navigate, both of
which look identical to a user - a click that does nothing.
1. The router awaits loadEnabledPlugins() to gate plugin routes. An app-pool
restart leaves that request hanging (IIS queues it while the worker starts)
and axios sets no timeout, so the navigation never resolves. Worse, the
promise is cached, so every later navigation awaited the same dead request
and stayed frozen long after the backend recovered. Bound the wait and fail
open on expiry, and drop the cached promise when an attempt times out or
fails so the next navigation retries. The setup-state probe in the guard
gets the same bound (it already fails open, defaulting to "complete").
2. A deploy replaces the content-hashed chunk files, so a tab open across it
asks for chunks that no longer exist and the dynamic import rejects with
nothing handling it. Reload once on a chunk-load error, via router.onError
and Vite's preloadError, guarded by a sessionStorage flag against a reload
loop and cleared on the next successful navigation.
Also stop index.html being cached: it names the hashed chunks, so a stale copy
points at files the deploy already deleted. It now revalidates while the
hashed assets under assets/ cache for a year.
Keying the ring off the fill singled out the light colors: on the map the
orange network-device marker took a black ring while its neighbours kept white
ones, and the legend dot for the same type wore a surface-colored border, so
the key did not match the markers.
Ring by the SURFACE instead - dark on the white blueprint, light on the dark
one - which is uniform within a theme and still works for every fill, since a
fill that resembles the ring is by definition far from the background. Legend
swatches take the same ring, the theme watcher redraws the markers (their ring
now depends on it), and the PDF uses the light-surface ring throughout because
it prints on white.
Markers sit on two very different surfaces - the light blueprint on white and
the dark blueprint over the near-black navy card - but the palette only ever
suited one. The grey "no subtype" default sat at 1.88:1 against white and the
orange asset-type step at 2.16:1, so both effectively disappeared on the light
blueprint. The fixed white ring made it worse: on white it added nothing.
Re-step the asset-type colors to versions of the SAME hues that clear 3:1
against both surfaces, replace the grey default with a neutral that clears
5.4:1 / 3.7:1, and derive the ring from the fill's luminance (light fill ->
dark ring, dark fill -> light ring) so every marker keeps a hard edge on
either background. The ring also rescues a washed-out color a user picks by
hand for a subtype, which no palette change can reach. The PDF export applies
the same rule to its markers and legend swatches, keeping print in parity.
Colors were chosen against a contrast/CVD validator rather than by eye. Note
that five simultaneous hues cannot all stay distinguishable under color-blind
simulation - past roughly five subtypes on screen, the legend and the hover
tooltip carry identity.
Adds computed contrast assertions so a future palette edit cannot
reintroduce a washed-out step.
exportPdf passed the raw map_blueprint_light setting value, which is a
root-relative /api path. Under /ops or /shopdb that resolves to the server
root and 404s, so the export died with "Failed to load blueprint image".
The on-screen map was unaffected because it goes through blueprintUrlFor,
which applies withBase - use that here too.
Same file, so this also carries the subtype auto-palette replacement that
goes with the marker-legibility change in the next commit.
Every asset list shows a Type column (and printers a Model, machines and
network a Vendor), but the search filters only looked at the asset number,
name, serial and hostname. Searching a type returned zero rows: 'Part Washer'
on machines, 'Standard' on PCs, 'Thermal' on printers.
Extend the search on machines, computers, printers, network devices,
measuring tools and the unified asset list to cover the type name plus the
vendor/model where the list shows them. Joins are outer joins so an asset
missing a type or vendor still matches on its own fields; the core list uses
a correlated EXISTS instead, since its type-name filter already joins
AssetType.
MapView carried its own copy of the per-type subtype-id lookup and it
never gained a Measuring Tool branch, so selecting any measuring-tool
subtype filtered out all assets. Marker coloring and the PDF export were
unaffected because both already used the shared getSubtypeId helper.
Point the filter at that shared helper and delete the duplicate copy in
ShopFloorMap too, so one definition serves filter, coloring and export.
Adds a table-driven spec covering every subtype-carrying asset type.
The old kiosk kept relaunching the dead URL from an HKLM Run value the 32-bit
Inno installer wrote - WOW64-redirected into SOFTWARE\Wow6432Node, which 64-bit
tooling (and the earlier purge) never saw. Broaden the sweep to both registry
views, every loaded user hive, Run/RunOnce/Policies-Explorer-Run, matching by
legacy name AND by any value pointing at the old URLs, plus every per-user and
common Startup folder.
The white-on-login was the old Dashboard/Lobby installer's leftover autostart
relaunching the dead old URL (404 -> white), not a network race - so the
wait-for-URL launcher solved the wrong problem. Go back to the plain direct
Edge kiosk shortcut and clean up any stale launcher file. The real fix (the
legacy HKLM Run-key + old .lnk purge) stays; it just has to be published.
At auto-login the Startup shortcut fired before the network was up, so Edge
--kiosk navigated to nothing and sat on a blank white page with no retry.
Point the shortcut at a hidden VBS launcher (wscript, no console flash) that
polls the kiosk URL until it responds (up to ~3 min) and only then launches
Edge fullscreen, so the first paint is the real page. Falls through to launch
anyway after the timeout so a display is never left dark.
Notification start/end times displayed and stored wrong by the tz offset
(a 2:34 PM entry showed 6:34 PM). Two stacked bugs: to_dict emitted stored
UTC as naive ISO (no offset) so the browser read it as local, and the form
filled the datetime-local input from toISOString() (UTC).
Fix and generalize to a configurable site timezone (multi-site):
- New setting site_timezone (default America/New_York), public, editable in
Settings > Site > Localization (common-zone dropdown).
- Backend tags datetimes UTC (_utc_iso); parse normalizes to naive UTC
(_parse_utc); daily-reset expiry uses the site zone (_next_site_time);
calendar allDay events key off the site-local day (_site_date).
- Shared frontend util datetime.js (Intl-based, DST-safe) converts between a
UTC instant and a site-zone wall clock. Notification form, list, and
calendar all render/enter in the site zone.
The MCP server README names Claude Desktop / Claude Code (it is an MCP
server for those clients), which trips the scrub gate. It is a standalone
local tool distributed via pxe-images/mcp/ + setup-mcp.cmd, not part of the
shipped product, so exclude it from the public repo like docs/ and tools/.
The old LobbyDisplay/Dashboard Inno installers planted an HKLM
...\CurrentVersion\Run value (plus a Startup .lnk). The dispatcher already
swept the stale .lnk/.url launchers but never the Run value, so a display
with our new ShopDB Kiosk.lnk still relaunched the old kiosk URL at logon
(the Run key beats the Startup shortcut). Remove the two legacy Run values
and kill any running old-URL Edge so the display self-heals to the resolved
target on the next enforce cycle.
