The other half of the sweep. ClientsController and its API twin checked
`abort_unless($client->isClient(), 404)` and nothing else -- a type
check, not a boundary, which is the phrase #1701 used about the group
membership routes for exactly the same reason.
Measured before the fix, with a client-scoped role holding the client
permissions:
GET /clients every client on the installation, name + email
GET /clients/{stranger} 200
PATCH /clients/{stranger} 302, name actually changed
DELETE /clients/{stranger} 302, client gone
The tell was one route over. ClientFilesController::index already draws
this line with StaffLibraryScope::canAssignClient and calls it "the same
boundary StaffLibraryScope enforces everywhere else in the library". Its
neighbours in the same family did not.
So the predicate is not new here. What is new is StaffLibraryScope::clients(),
the listing half of canAssignClient, so a screen narrows by the rule its
own buttons are guarded with instead of restating it -- restating it is
how this went wrong, and how the last four of these went wrong.
Eight actions take it: edit, update, destroy and the two-factor reset on
both surfaces, plus both listings. Answering 404 rather than 403, since a
client outside the roster should not be distinguishable from one that is
not there -- matching the isClient() guard already above it.
Account requests stay installation-wide on purpose: a self-registered
client who has not been approved belongs to nobody yet, so there is no
roster to narrow by and narrowing would empty the screen.
The published API document is unchanged -- both routes already documented
the 404 that the type check produced.
The sweep after #1685 turned up the same leak two widgets further down
the same controller. largestFiles() and expiredFiles() already take the
viewer -- to decide whether their rows get links -- but queried with a
bare File::query(), so a client-scoped staff member's dashboard named
files belonging to clients they hold nothing of.
The note above largestFiles() says a link that 403s is accepted rather
than adding per-row scope checks. That reasoning is about the link. A row
that should not be there at all is a different problem, and the name is
the part that leaks: "Q3 delinquent accounts" says plenty without ever
being downloadable. Scoping the query is also cheaper than the per-row
check that note declined -- StaffLibraryScope builds a scoped user's
query once per request.
Reachable in the default configuration, unlike the last few of these: the
Client Manager role ships client-scoped and holds view_statistics.
topClientsByStorage() goes with them; it names clients rather than files,
which is the thing MembershipRequest::approvableBy and ActivityLogScope
already exist to keep inside a roster.
counters() and transferSeries() stay installation-wide, and now say so.
A total carries no names -- "417 files" tells a scoped viewer nothing
about whose they are -- and if that ever stops being the line, both move
together.
One consequence worth stating rather than discovering: scopeVisibleToClient
ends in notExpired(), so a scoped viewer's expired-files widget now lists
only their own expired uploads, not a client's. Safe, and under-inclusive
-- telling them about a file auto-delete is about to take needs a library
query that keeps expired rows, which is a boundary to decide rather than
to invent inside a leak fix.
#1701 closed a real hole: group membership decides what a client reaches,
and through File::scopeVisibleToClient it decides what the staff member
holding that client reaches, so `edit_groups` alone was never a boundary.
The predicate it added asks whether everything shared with a group is
already inside the actor's library.
It asked by counting: pluck the group's assignment rows, count how many
of those ids the library query returns, and require the two to match. An
assignment row outlives the thing it points at — nothing clears them when
a file or folder is deleted — while files() and folders() exclude trashed
rows by construction. So one deleted file left a count that could never
balance again, and the group closed permanently: the scoped staff member
could no longer add their own client to it, or remove anybody from it,
with a 403 and nothing to explain it. Every group accumulates dead
assignments over time, so groups would have gone quiet one at a time.
Asked the other way round — is there anything live, shared with this
group, that is outside my library — the dead rows drop out by
construction, because the query starts from File/Folder rather than from
the assignment. That is also the truer question: a deleted file is not
reach, since nobody can reach it.
Three tests. A group stays usable after a file shared with it is deleted,
including removing a member; the same for a deleted folder assignment;
and the half that must not soften — a live file still out of reach is
still refused, deleted siblings or not.
#1685 fixed the dashboard rebuilding a log row by hand and dropping
`origin` from it. One layer down, the same method was skipping something
larger: it ran a bare ActivityLog::query(), so ActivityLogScope never
applied.
That scope exists for this exact case, and says so in its own docblock —
`view_actions_log` is not the whole answer for a client-scoped staff
member, because a log entry carries the subject's *name*. An unscoped log
reads out the name of every file in the installation, and who touched it,
to somebody who gets a 403 on the files themselves.
Measured before the fix, one client-scoped viewer with the permission:
/activity → []
/dashboard → Uploaded the file "Q3 delinquent accounts"
Same person, same permission, opposite answers. The activity page and the
download history both apply the scope; the dashboard was the one caller
that did not, which is the same shape of gap #1685 was about.
More reachable than it looks: the Client Manager system role ships with
`view_actions_log`, so this is the default configuration rather than
something an administrator has to build.
Two tests: a scoped viewer sees only the entry about a file in their
library, and an unscoped one still sees everything.
transferSeries() is left alone on purpose. It is unscoped too, but it
returns per-day counts with no names or subjects attached, which is a
different exposure and arguably not one at all.
#1698 moved the library boundary into FileCommentPolicy, where it
belongs, and said plainly what that cost: the moderation screen went
from 65 queries to 465 for a client-scoped moderator with five assigned
clients. Measured here, those numbers are exactly right.
The cost is not in asking. It is that StaffLibraryScope::files() rebuilds
its query every time, and building one runs four immediate lookups per
assigned client — the client's group ids, the same ids again inside
Folder::sharedFolderIds(), that method's own assignment lookup, and the
shared-folder get() in Folder::scopeVisibleToClient(). None of them
depend on the query being built. Gate resolves a fresh policy for every
check, so a listing paid for all of it once per row.
The built query is now memoised per user and handed back as a clone,
since every caller adds to it, and the scope is registered as `scoped`
rather than transient so the memo survives a request. Scoped rather than
a singleton on purpose: a long-lived queue worker keeps singletons
between jobs, and a library query built from one job's data has no
business answering the next one's question.
That is 465 queries down to 60 on the same page — below the 65 it cost
before #1698, because the memo also helps the callers that were already
asking repeatedly. FileVersions::sharedAudience(), which runs the same
helper twice per candidate while resolving notification recipients, gets
it for free.
So the answer to the question #1698 left open is neither of the two it
offered. can_delete stays a real question asked of the policy; nothing
restates the boundary; and the page is faster than it was before the
fix. Three tests: one user's query never answers another's, one caller's
constraints never follow the next, and the moderation screen does not
ask once per row.
The four routes that edit a group's membership -- add and remove, web
and API -- contain no authorization call of any kind. `can:edit_groups`
in front of them is the whole of it, and a permission is not a boundary.
The authorization sweep looked at these and let them stand, on the
grounds that groups are installation-wide by design: GroupsController
::index lists every group unfiltered, so list and single-object access
agree, and there is no listing/direct-access mismatch to fix. That is
true, and it is the answer to the question of who may *see* a group.
This is a different question: what a write to one *does*.
Joining a group hands the new member everything shared with it. When
that member is one of a client-scoped staff member's own clients,
File::scopeVisibleToClient hands the same content straight back to them
-- that scope is what StaffLibraryScope::files() is built out of. So the
one write turns a file they get a 403 on into a file in their library,
and the download that follows is a 200. ResolvesShareTargets draws that
line on the sharing path through canAssignGroup(); nobody drew it on the
membership path, and canAssignGroup() is *derived from membership*, so
whoever may edit the list also decides what the list entitles them to.
StaffLibraryScope::allowsGroupMembership answers it directly instead of
through the derived predicate, which is the wrong tool here twice over.
Membership asks about reach, so it checks reach: the client must be one
this staff member holds, and the group must not already reach past their
library -- no file assigned to it, and no folder shared with it, outside
StaffLibraryScope. A group nothing has been shared with passes trivially,
which matters, because canAssignGroup() would have said no to a group
that has no members yet and left a scoped staff member unable to put the
first client into one they had just created.
The same write has a second door. MembershipRequestsController::approve
joins a client to a group with identical consequences, under
`approve_groups_memberships_requests`, and deny() decides about somebody
else's client and emails them about it. Both go through the same
boundary, answering 404 to match the guard already above approve().
The queue and its sidebar badge are narrowed to the clients the viewer
holds, through one scope on the model that both read -- the rule the
comment badge in HandleInertiaRequests already states two branches down
("a client-scoped staff member is not shown a number they cannot act
on"), and the reason VisibleCommentScope owns its own pendingTotal()
rather than leaving the middleware to count for itself. Each row carries
the client's name and email, so an unnarrowed queue was also handing
those over for clients outside the roster. Unscoped staff still see every
pending request.
That narrowing is on the client, not on the group: whether a group is
reachable depends on what is shared with it, which is not a question to
ask row by row in a listing. A scoped viewer may therefore still be
shown a request they would be refused on -- one of their own clients
asking to join a group out of their reach. The names were the part that
leaked.
Unscoped staff are unaffected throughout -- both halves of the predicate
are true for them by construction. No seeded role reaches this: Client
Manager is the only client-scoped role that ships, and it holds no group
permissions, so a custom role is needed to get here at all.
The published API document gains a 403 on both member routes.
Regenerated with php artisan scramble:export; Scramble reads abort_unless
out of the method body but not out of a private helper, which is why the
guard is written out at each of the four call sites rather than shared.
approve() refuses a request that is not pending:
abort_unless($group !== null && $client !== null
&& $membershipRequest->status === MembershipRequest::STATUS_PENDING, 404);
deny(), one method below, checks nothing. Denying is not idempotent, so
repeating it is not a no-op:
- denied_at is stamped again, and that is what the client's re-request
cooldown counts from (MyGroupsController::inDenyCooldown). Repeating
the request keeps one client out of one group for as long as somebody
cares to keep asking, without a single new decision being made.
- a second GroupMembershipDenied entry goes into the activity log, for
a denial that did not happen.
- a second "your request was declined" mail goes to the client.
The queue lists only pending requests, so nothing on the screen offers
this; it takes asking for the route directly. It needs
approve_groups_memberships_requests, so it is not a stranger's move.
The guard is the same one, answering the same 404, placed where deny()
can reach it. deny() keeps tolerating a vanished group or client -- that
tolerance is deliberate and separate: the denied row persists for the
cooldown even when the group it named is gone, and index() already
filters those rows out with whereHas.
Not in this change: deny() writes the status, the log entry and the
notification without a shared transaction. approve() has exactly the same
shape, so fixing one alone would replace a symmetry with a difference,
and doing both means also deciding where the mail sits relative to the
commit -- which is the question #1691 answers for file bytes, and worth
answering on its own rather than inside a state-machine fix.
consumeRecoveryCode() reads the whole list, filters the used code out,
and writes the whole list back. Two requests that both read before
either writes each store their own copy, and the second write puts back
the code the first removed. So a spent code comes back, and the same
code offered twice is accepted twice -- while the method's first line
says "each code works exactly once".
Nobody gets in through this who was not already holding a valid code, so
it is a promise not being kept rather than a door standing open. The
promise is worth keeping anyway: it is the whole reason a printed sheet
of recovery codes can be crossed off, and it is what makes a code that
somebody watched being typed in stop working.
The decision now comes from the row as it stands, re-read under a lock
inside the transaction that writes it -- the shape SendNotificationDigest
already uses to claim the rows it is about to delete. A conditional
update, as in PublicShareController's downloads_count and the delivered_at
claim in #1692, is the other precedent in the tree, but the column is
`encrypted:array`: there is nothing in it a database can compare, so the
comparison has to happen after decryption, under something that holds the
row while it does.
The lock is what makes it atomic against a request arriving at the same
moment. The re-read is what makes the decision right, and it is the half
a test can show: SQLite ignores lockForUpdate, so the accompanying tests
pin the re-read and say so rather than claiming to prove the locking.
config/database.php runs MySQL or Postgres in production, and both honour
it. Saving through the caller's own instance keeps that instance in step
with the row, so a caller cannot go on to decide from a list the database
no longer has.
Folder uses SoftDeletes. The `exists` rule runs against the table, so a
folder in the trash passes it -- while every resolution that follows goes
through Folder::query(), which honours the soft delete and finds nothing.
Ten rules across five controllers rely on that check, and each one reads
it as "this folder exists".
Two of them then wrote the id anyway. Api\FilesController::store()
resolves the folder, hands the null to Folder::uploadableBy(), is told
yes -- correctly, that is the rule for a root upload -- and passes
$validated['folder_id'] to the write. FilesController::store() is the
same shape once #1694 gives it the guard. FilesController::update() and
its API twin write it straight through with nothing in between.
The result is a live file inside a deleted folder, which is a state
nothing else in the application produces: FolderService::delete() deletes
every file in the subtree along with it. The row is reachable by id, in
search and over the API, and missing from the listing its uploader would
look in.
Rules::folderId() makes the check mean what its readers assume, once,
where the reasoning can be written down -- the same argument slug() makes
for itself one method above. Every site takes it, so the file cannot end
up with two spellings of the same rule and no way to tell which is the
safe one.
What changes, path by path:
- POST /files, POST /api/v1/files, PATCH /files/{file} and
PATCH /api/v1/files/{file} refuse a folder in the trash instead of
writing its id. This is the fix.
- POST /uploads used to accept it and quietly file the upload at the
root -- its guard and its write already agreed, on null. It now says
so instead, which is what the other upload paths do.
- files/{file}/move, files/bulk-edit, folders, folders/{folder}/move
and the portal's my-folders already refused, through
StaffLibraryScope::folders() or Folder::scopeVisibleToClient(), both
of which drop trashed rows. They still refuse; the answer is now 422
naming folder_id rather than a bare 404. Those two guards are asking
a different question -- "is this folder yours" -- and they keep
asking it.
No live folder id behaves differently anywhere, and the root (a null
folder_id) is untouched.
The published API document is unchanged: `exists` renders the same either
way. Regenerated with php artisan scramble:export and byte-identical.
Follow-up to #1687, which made a zip build report failure honestly. Four
things it passed near, none of them regressions it introduced.
A zip has never had a size limit — only a cap of 10,000 files, which
bounds nothing that costs anything. Ten thousand spreadsheets zip in
seconds; two hundred videos is an hour of stream-copying and an archive
that fills the disk. Bytes are what a build actually costs, so the new
Settings → Downloads screen caps the total size instead, at 2 GB out of
the box. It is a setting rather than a constant because the safe figure
depends on free disk, on whether sources live on a remote disk, and on
the plan a hosted tenant is on — the file count stays fixed, since it is
a foot-gun rail and not a knob anybody needs. The controller measures
the selection at request time and names both numbers when it refuses;
the job measures again, because it re-derives the selection at run time
and a folder can grow while the job waits in the queue.
Every shipped topology runs exactly one queue worker, and everything
shares the default queue, so raising the job timeout to an hour handed
any signed-in person an hour of everyone else's notification mail. There
is now one build in progress per requester and a named throttle bucket
on the endpoint, which had neither. A pending row older than an hour is
treated as abandoned rather than in progress, so a worker killed hard
enough to skip failed() cannot lock somebody out for good. Giving zip
builds their own queue is the structural fix and wants its own change:
it touches compose, supervisord and the systemd unit in INSTALL.md, and
an install that upgrades without changing its worker command would stop
building zips silently.
zip_downloads.requested_by cascades on delete, so removing a user takes
their rows with it and strands every archive they built — invisible to a
purge that walks rows, and to OrphanFileScanner, which skips zips/ on
purpose. The purge now also sweeps files in zips/ that no row explains,
after a day's grace so a build in progress is never taken out from under
itself.
Two smaller things while in here. A build that failed because every file
had already hit its download limit said only that nothing was available,
and dropped the skipped list — the same distinction the store guard goes
out of its way to draw at request time. And a failed close() now logs
libzip's reason, which the @ silencing had been discarding: "the disk is
full" and "the source vanished" are different problems for whoever has
to fix one, while the requester still sees a message with no server
paths in it.
update() validated preferences.*.type as ['required', 'string'], so any
string at all became a row in notification_preferences. Nothing reads it
afterwards: emailEnabledFor() looks preferences up by a key the registry
knows, so a row under an unknown key is invisible for good.
It is not a way into somebody else's settings — user_id comes from the
session, never the payload — which is why this is validation rather than
authorization. The cost is a table that quietly accumulates rows nobody
can see, explain, or remove through the interface.
edit() already knew the answer. It filters the registry down to the types
that can email at all, by either route, and renders exactly those as
toggles. That list is now derived once and used by both halves, so what
the screen offers and what it accepts back cannot drift apart.
Rejecting a registered-but-unmailable key (client_uploaded is the one in
tree) is deliberate rather than incidental: FilesServiceProvider explains
that it has no mail companion on purpose, so a preference row for it
could never change what anybody receives.
FileCommentPolicy::moderate() asked only whether somebody is staff and
holds moderate_comments. It never weighed the file the comment sits on,
and delete() returns true the moment moderate() does — so a client-scoped
moderator could delete any comment on the installation by naming its id.
Three call sites already knew this and wrote the boundary out by hand,
each with its own abort_unless($library->allowsFile(...), 403) after the
gate. The two that did not are FileCommentsController::destroy(), web and
API: both bind a comment directly, so nothing earlier in the request
establishes that the viewer may see its file.
The intent was documented in three places and enforced in none of them by
the policy — StaffLibraryScope says "the policies consult allowsFile() so
direct access respects the same boundary", VisibleCommentScope says "a
moderation screen is not a way around the visibility model". Put the rule
where those docblocks already say it lives.
moderate() now takes the comment when there is one. Named against the
class it still answers the coarser "does this user moderate at all",
which is what the queue's gate and the affordances ask. Membership is
tested by file id, so a file soft-deleted out from under its comments is
not in a scoped moderator's library either.
The author branch of delete() is deliberately untouched: deleting your
own words inside the edit window is not moderation, and a client is not
client-scoped in StaffLibraryScope's sense.
Approving through the API now derives its 403 from Gate::authorize rather
than the removed abort_unless, so the committed OpenAPI document gains
the shared AuthorizationException ref in place of an inline "An error"
schema — the shape nine of the other twelve documented 403s already use.
I said they could not, in the guide, the Zapier page and the changelog.
That was carried over from the limitation of polling a list, where a
deleted row stops being returned and nothing marks the moment it went. It
was never true of the activity endpoint: file.deleted is recorded like
any other action, so ?action[]=file.deleted works today.
Checked with a test rather than the enum, which turned up the shape a
caller needs: a deletion entry has no subject, because the row is gone by
the time the entry is written, so the name is snapshotted into
context.name instead. Reading subject.name there gets you null.
Every list in /api/v1 answers "what is there now". Nothing answered
"what happened", and for the two events people most want to act on there
was nowhere to look at all.
Sharing a file writes an assignment row and never touches the file, so no
amount of polling /files?updated_since= will ever show a share. A
download is recorded only in the activity log. So the most requested
automations for a file-sharing product — tell me when a client gets a
file, tell me when they open it — were not possible to build.
GET /api/v1/activity is one feed rather than one endpoint per event,
because the log already records every one of them and a caller filtering
by action gets whatever the application grows later without waiting for
us to expose it.
It reuses what already exists: view_actions_log is the permission the
activity screen uses, and ActivityLogScope narrows the rows the same way,
so a staff member limited to their assigned clients cannot read the whole
installation's log through a token when the screen would not show it.
Two deliberate limits. Class names never reach the wire — subject.type is
a stable public string, or moving a model between namespaces would be a
breaking change to a frozen contract. And no ip_address, though the
column exists and the screen shows it: a person looking at a log has
decided to look, where an integration streams every row to somebody
else's servers by default.
PollingQuery grew an optional column so it can walk a table that is
appended to rather than edited. The parameter stays updated_since
everywhere, because on an append-only log the two timestamps are the same
thing and one shape learned once is worth more than a second name.
BuildZipDownloadJob deferred every write to ZipArchive::close() but then
marked the row STATUS_READY regardless of the result:
- close() returns false when a source file was deleted between addFile()
and close() (a concurrent staff delete runs FileDiskCleanup at once) or
the disk filled up; the row went ready over an archive libzip never
wrote, and the download controller X-Accel-served a path that isn't there.
- An archive that ended up with no entries (every selected file removed or
its allowance spent before the queued job ran) is written as no file at
all by libzip, yet close() still returns true — again marked ready.
Check both the close() return and the added-entry count, and fail the row
(deleting any partial archive) when either says nothing was written.
The job also had no $tries/$timeout/failed(): a build of up to MAX_FILES
sources runs past the worker's default 60s timeout, and the kill skips the
catch, stranding the row as PENDING while the frontend polls forever. Give
it room, run it once, and add a failed() backstop that fails a row still
pending (leaving an already-resolved one alone).
Finally, purge leftover zips/{id}.zip* by row id: a killed build leaves a
partial archive and libzip temp file with no path recorded, so the path
field alone never cleaned them up.
The Recent activity widget rebuilt each log entry inline instead of going
through ActivityPresenter, and the inline copy dropped `origin`. On the
frontend "System" and "Anonymous" are both actor_name null and only
`origin` tells them apart, so every actorless entry -- public and
share-link downloads, anonymous comments -- rendered as "System ...".
Present the entries through the shared ActivityPresenter, the same
sentence-ready shape the activity page and detail panels already use, so
the dashboard cannot drift from them again.
The orphan scanner skips derived artifacts by path prefix, but the list
was a hard-coded ['thumbnails/', 'zips/'] that never learned about
'previews/'. ImageRendition::Preview caches under previews/ (and
previews/external/) on the local files disk, so every cached preview was
reported as an orphan: offered for import on the orphans screen, and
deleted by the purge command once past the grace period. An imported
preview also became a File row pointing at a path the rendition cache
owns -- destroyed the moment its source file was deleted or the cache
was flushed.
Derive the rendition prefixes from ImageRendition::cases() rather than
repeating them, so a future rendition can't be forgotten here the way
previews were; 'zips/' (the download-bundle job's) stays as it was.
An actor with no personal access token has always meant a browser, and
for as long as a session and a Sanctum token were the only two ways to
authenticate, that was true. It stops being true the moment anything else
can, and the failure is silent: the action gets recorded as a person
clicking, in the one table whose whole purpose is answering "did I do
that, or did something acting for me?"
Nothing misreports today — every call site that passes an explicit actor
is a browser request, an API request whose actor carries the token, or a
console command with no actor at all. This closes the trap before the AI
connector in cloud-modules walks into it.
ActivityOrigin is a closed enum, so core has to publish both the case and
the hook before a package can use either. ResolvingActivityOrigin is
asked only in the ambiguous case: a request carrying a token is the API
and a request with nobody signed in is public or system, and neither is
in any doubt, so neither is offered — one package must not be able to
quietly relabel how every integration's actions are attributed.
The person stays the actor. They authorised it, and a log naming the
assistant instead would lose the only fact that matters. What the
connector was called goes in api_token_name, beside a null token id,
because that column means a row in personal_access_tokens and this is not
one.
The new origin is kept out of the activity filter unless the edition can
actually produce it. A filter option that can only ever return nothing is
a feature dangled at an edition that does not have it, which is the one
thing the edition boundary exists not to do.
Both fixes are follow-ups to the mail providers @denkfabrik-li added in #1679.
A refresh token is good for exactly one use — Microsoft and Google both
retire it as they issue the next one. Two queue workers finding the same
expired access token would therefore both spend it, and the loser gets
invalid_grant back. That is the same answer a revoked grant gives, so a
healthy connection would be marked broken, painted red on the settings
page and mailed to every admin. Refreshes now hold a per-connection lock
and whoever waits re-reads the row, which normally means finding a token
the winner already stored and not refreshing at all.
The Connect button read the provider dropdown, but the flow it starts
uses the saved provider. On an installation with both vendors registered,
switching without saving would open the wrong consent screen. The
dropdown now counts as an unsaved change like any other field, which also
gives it the right "save first" hint for free.
Both sides added a .gitignore rule in the same place: this branch's
exception for docs/email-oauth.md, and main's block for the local dev
TLS material. Keep both.
#1680 fixed the redirect rendered from an exception and said plainly
what it did not cover: EnsureSetupIsComplete, EnsureAccountIsActive and
EnforceTwoFactor answer before HandleInertiaRequests is ever entered, so
a response they return never unwinds through Inertia's 302 to 303
upgrade either. Same 405, reached a different way — an account
deactivated while its owner was part-way through a form, or one being
made to enrol in two-factor.
The rule now lives in one place rather than four. Three copies of "if
the method is PUT, PATCH or DELETE" is how the fourth caller gets it
wrong, and WriteSafeRedirect can carry the explanation of why 303 —
which is worth more than the three lines it replaces, because nothing
about a bare setStatusCode call says what a browser does with a 302.
PUT /timezone is the route the setup test uses: it is one of only two
writes a guest can reach and the only one that middleware does not
exempt, so the case is real rather than defensive. All three new tests
were run against the unfixed middleware and fail there.
Extends the work of @denkfabrik-li, who found the gap and wrote it down.
A write that meets an expired session lands on the login page
A browser follows a 302 by replaying the request method, POST aside, so
a widget save whose session had expired replayed as PUT /login — and
/login takes GET and POST only. The 405 that came back hid the one
thing the person needed to be told, which was to sign in again.
Inertia already upgrades 302 to 303 for writes, but a redirect rendered
during exception handling never travels back through the middleware
stack, so the guest redirect after AuthenticationException was never
reached. bootstrap/app.php now does it for those.
Fixes#1673, reported by @mstewart14. Found, diagnosed and fixed by
@denkfabrik-li, who also wrote down the part this does not cover: the
direct redirects from EnsureSetupIsComplete, EnsureAccountIsActive and
EnforceTwoFactor sit outside HandleInertiaRequests and can still produce
the same 405 on a write.
Two bugs a green suite could not find, both from pointing the
application at a real Google Cloud Storage bucket.
The adapter attaches a legacy per-object ACL to every write, and a
bucket with uniform bucket-level access — which our own setup
instructions require, and which Google recommends — refuses it:
"Cannot insert legacy ACL for an object when uniform bucket-level access
is enabled". So the default configuration could not write to the
recommended bucket. The library ships
UniformBucketLevelAccessVisibility for exactly this, and nothing is
lost by never setting an ACL: every object here is private and every
read is a signed URL.
The second is worse and was never about Google. Both file disks are
configured 'throw' => false, so a refused write returns false rather
than raising, and LocalPartStore ignored the return. The upload reported
success, the File row was written, and the bytes were nowhere — the
listing showed a file whose download could never work. An expired S3
credential did the same thing. It now checks, and the controller already
turns that into a validation error rather than a 500, so the person
uploading is told.
Verified against a live bucket with a key scoped to
roles/storage.objectAdmin: the probe lists, writes land, reads
round-trip byte for byte, and a signed URL comes back 200 carrying
"Informe año.pdf" intact through both the ASCII and RFC 8187 forms of
Content-Disposition.
A redirect born in exception handling - the guest redirect after an
expired login, above all - never travels back through the middleware
stack, so Inertia's usual 302-to-303 upgrade cannot reach it. Browsers
follow a 302 by replaying the request method on the redirect target
(only POST is downgraded to GET), so a widget save whose session just
died replays as PUT /login and fails with a 405 that hides the real
"please sign in again" (#1673).
Repeat the upgrade in the exception pipeline: any 302 answered to a
PUT, PATCH or DELETE becomes a 303. Reads keep their 302, POST needs
nothing - browsers already downgrade it.
External storage meant S3 and nothing else, which is an odd hole for a
product whose users are as likely to be standing on Google Cloud as on
AWS — and paying to move bytes between two clouds to use this. The
Storage screen now asks which provider first, and the answer decides
which fields it shows, which it validates, and which driver the
files_external disk resolves to.
One disk, not two. files.disk is a stored column, so a third disk name
would fragment the data model and make every $file->disk consumer know
three names instead of two; the driver is swapped instead. A service
account key gets its own encrypted column rather than sharing `secret`,
because the two are validated, labelled and displayed differently and
one column meaning two things is how that goes wrong later.
Three things do not work by simply adding the adapter, and all three
fail quietly:
Laravel's temporaryUrl() looks for getTemporaryUrl() on the adapter,
while League's GCS adapter names it temporaryUrl(), so without the
registered callback every download and preview is a 500.
The two SDKs spell the signing options differently, and an unrecognised
one is dropped in silence — the symptom is a download named after the
storage key, not an exception. GoogleCloudStorageDriver translates, so
callers keep speaking one vocabulary, and the test asserts on the URL's
contents rather than on "a redirect happened", which is what would let
it regress.
That callback is also re-bound to the FilesystemAdapter before it runs,
so the translation is captured before registering rather than called as
$this->
`provider` is validated with 'sometimes', not 'required': absent means
S3, which is what every payload written before this choice meant, and
stops a browser holding a stale bundle from failing to save on a field
it cannot see.
Verified in a browser as well as in tests — which is how the null
provider on an unmigrated row was found, since the suite migrates and
never sees that state.
Two routes still assumed every file sits on local disk, which stopped
being true the moment external storage was switched on. A share link
answered with X-Accel-Redirect whatever the file's disk said, pointing
nginx at a path it has nothing behind; a public listing built a
thumbnail from Storage::disk('files')->path(), which for an externally
stored file is a path nobody ever wrote. Both fail only for installs
using S3, and only on those two routes, so the same file downloading
correctly from the file manager made the share link look like the
broken thing rather than where the file lives.
Neither is a new rule. FileDownloadController and
FileThumbnailController already did it right, which is the actual
finding: the knowledge was sitting in a private method on one class and
inline in another, so the next caller could not inherit it and did not.
Both are now objects with one job.
StoredFileResponse replaces InlineFileResponse and grows an
attachment() alongside inline(), since the two differ only by
disposition. LocalSourceFile takes a closure rather than returning a
path: the version that returned one also left the caller to unlink it,
and both of those are exactly the mistakes made here.
The regression tests fail against the previous controllers — checked in
both directions rather than assumed.
It is used by both OAuth mail test files, which --parallel runs in
separate processes — exactly the situation tests/Helpers.php exists
for, as its own header explains. CI caught what a whole-suite serial
run hides.
Same delegated shape as the Microsoft 365 provider, through the same
broker interface: the admin registers an OAuth client in Google Cloud
Console, connects the Google account the installation should send as,
and outgoing email goes through the Gmail API's messages.send as that
account.
The shared authorization-code machinery (exchange, refresh, token
storage, id_token account detection, RFC 6749 failure telling a dead
grant from a transient one) moves into an abstract OAuthCodeFlowBroker;
the two vendor brokers keep only their endpoints, scopes and consent
URL parameters. Google's quirks live where they belong: offline access
with a forced consent screen (the only way Google issues a refresh
token), and a refresh response that never re-sends one — the store
keeps what it has.
The settings screen needed no changes: the dropdown, the credential
form and the connect flow all derive from the provider enum.
Adds "Microsoft 365 (OAuth)" to the Email settings provider dropdown.
Selecting it swaps the SMTP form for an app registration (client id,
secret, optional tenant) and a "Connect mailbox" flow: the admin signs
into the mailbox the installation should send as, and outgoing email
goes through Graph sendMail as that mailbox — no password, no app
password, no SMTP AUTH, which Microsoft is winding down.
Delegated flow on purpose: it needs no admin consent and works for
work/school and personal accounts alike. Its one weakness — a grant
can die silently behind a password reset or a Conditional Access
change — is answered by a daily scheduled refresh that keeps the
token alive and, on a dead grant, warns the settings admins once
in-app and on the settings page instead of letting mail stop quietly.
Tokens and the client secret live encrypted in their own row and are
read fresh at send time, never through the boot-config cache. The
stored SMTP transport survives a provider switch untouched.
publicListingFile() and publicListingImageFile() were defined in
PublicGroupsTest.php and used from PublicFilePreviewTest.php too. Pest
declares a test file's functions as ordinary globals, so that works only
once the defining file has been loaded — which under --parallel depends
on how the runner happens to distribute files across processes. Adding
any unrelated test file anywhere in the suite reshuffles that and takes
PublicFilePreviewTest.php down with "Call to undefined function", and
running it on its own with --filter never worked at all.
tests/Helpers.php exists for exactly this and its docblock describes this
failure; these two had just been missed. publicPageProps() stays where it
is, since only one file uses it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The value was read with env() inside the withMiddleware closure in
bootstrap/app.php. That closure runs when the HTTP kernel is resolved,
which is before the dotenv bootstrapper reads .env — so on every web
request env() returned null for anything set in .env, and the proxy was
never trusted. It worked when the value came from a real environment
variable, which is why the Docker compose path was fine and the manual
install described in INSTALL.md, where we tell people to put it in .env,
was not. Artisan bootstraps in the other order, so a check from the
command line reported the setting as working the whole time.
Behind a TLS-terminating proxy the consequence is not subtle. Laravel
falls back to the connecting address and the plain scheme, builds every
link and redirect with http:// while the browser is on https://, and
marks the session cookie non-secure. The browser then declines to send
that cookie to what it reads as a different, less secure origin, the
session arrives empty, and the first write fails with a 419 that reads as
"your session expired" — most often on the create-your-admin form, which
is the first thing a new install submits. Afterwards each redirect leaves
and re-enters over the wrong scheme, which is the random bounce back to
the login screen people report as flakiness.
Moved to config/trustedproxy.php, the key the framework's TrustProxies
middleware already falls back to on its own. Config files load after
dotenv, so the value is there whether it comes from .env or from the
environment.
This was also the only env() read outside config/, which means
config:cache is no longer dangerous on this application.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A download is not served by PHP. PHP authorizes it and hands the web
server the path with X-Accel-Redirect, so the web server has to open a
file PHP wrote. Where those are different users — cPanel and Plesk
commonly arrange it that way — it cannot: uploads land 0600 inside a 0700
directory, and traversing 0700 means being its owner. Nothing else on the
site shows a symptom. Uploading works, the library lists everything, and
only downloads fail, as ERR_INVALID_RESPONSE in the browser and
`open() ... failed (13: Permission denied)` in the web server's log.
FILES_WEB_SERVER_READABLE writes uploads 0644/0755 instead. Opt-in and
spread into the disk configuration rather than switched by a ternary, so
an install that does not set it keeps byte-for-byte the configuration it
had: the relaxed modes are readable by every account on the machine,
which is the wrong trade wherever the web server and PHP are one user, as
in the image and on most self-administered servers.
The two halves are not enforced alike, which is the part worth knowing.
`visibility` has Flysystem chmod each file after writing it, so 0644
holds under any umask. A directory is created by mkdir(), which masks its
mode argument, so 0755 is a ceiling: a pool at umask 0077 still produces
0700 and still cannot be traversed. That cannot be fixed from config, so
INSTALL.md carries it — how to tell the two users apart, the one-time
chmod for files already on disk, and the pool setting for the umask.
FilePermissionsTest asserts all three modes, umask cases included, since
the asymmetry is invisible from the configuration.
Reported by @denkfabrik-li (#1668), who diagnosed it and verified the
remedy on the affected host.
The two things staff most often want to know about a file — who
downloaded it, who looked at it — were answerable only by reading the
whole activity log past everything else that had happened to it, or by
going back to the library list for the details panel.
The file's own page now has a Downloads & previews tab: the twenty most
recent times it was taken or looked at, each with who did it and the
address it went to, over a running count of both. Below them, two
buttons open the file's full history already filtered — one to every
download, one to every preview — so the narrow question is one click
and the whole log is still one click further.
Which filter value stands for "every download" is decided server-side
and travels with the payload, because it is a fact about the log's
vocabulary: downloads are three actions and share a group, previews are
one action and are filtered by name. The history page now also keeps
whatever filter it was sent with visible in its dropdown even at a
count of zero, so a button cannot land somebody on an empty table above
a select that has gone blank.
ProjectSend prints the update instructions for the way this server was
installed, and it knew two answers where it needed three: anything inside
a container was handed `docker compose pull && docker compose up -d`. On
the Compose stack that builds from a checkout there is no image behind
those containers, so `pull` skips every ProjectSend service and `up -d`
then finds them all current — the update reports success, changes
nothing, and the dashboard goes on offering the same release. Reported by
@mueller7382, who stayed on 2.0.0 that way while 2.1.0 was out (#1661).
Those installations are now their own kind, told to `git pull` and
rebuild, with the two steps a checkout needs that an image does not: its
dependencies and its compiled frontend live outside git, so a release
that moved either leaves them stale.
Two signals decide it, in that order. The published image now declares
itself with PROJECTSEND_IMAGE, which is the only evidence an operator
bind-mounting over /var/www/html can neither hide nor forge; failing that
— images published before this — a working tree in the install directory,
which the image never has and the repository's own stack always does.
getenv() rather than env(), because a cached configuration makes env()
outside a config file return null, and the answer would flip silently on
exactly the installs most likely to have cached it.
The stale-code banner keeps treating both container kinds alike: what
clears it is recreating the container, whichever way its image was built.
The changelog also credits the reporter of #1663, which was missed when
that entry was written.
v1 could preview four kinds of file in a modal — images, video, audio and
PDF. v2 previewed only images, and not by decision: preview shipped as part
of the image *thumbnail* work (1c68aa1), so "previewable" quietly became a
synonym for "GD can decode it". FileThumbnailController::preview() gated on
ThumbnailGenerator::SUPPORTED_MIME_TYPES, the frontend mirrored the same
four types, and the dialog was a hardcoded <img>.
Rather than widen that list — it drives pathFor(), extensionFor(),
generate() and FileDiskCleanup, and a video reaching getimagesize() is a
500 — this separates the two questions. PreviewKind now answers "may these
bytes be served inline, and what element renders them?", while
ThumbnailGenerator keeps answering the narrower "can this app decode it
itself?", which is what renditions, the cache and the watermark hook
actually depend on. Image delegates to it so the two cannot drift.
The allowlist stays a security boundary: mime_type is sniffed from the
bytes, so text/html and image/svg+xml remain excluded, and PreviewKind is
deliberately narrower than "formats a browser might cope with" — no
quicktime, avi or matroska, because an embedded player for those shows a
black rectangle. Those still download exactly as before.
docs/security-audit-2026-08-05.md finding 1 recorded that adding
application/pdf "should be a conscious decision". This is that decision,
and three things were measured rather than assumed:
- An <iframe sandbox> cannot be used. Chrome refuses to run its PDF viewer
in a sandboxed frame at all (ERR_BLOCKED_BY_CLIENT, with or without
allow-same-origin) — the attribute removes the feature, it does not
harden it.
- nginx's `Content-Security-Policy: sandbox; default-src 'none'` on
/protected-files/ does work (a <video> frame lands in an opaque origin),
but Chrome exempts its PDF viewer from it, so it is not what protects
the PDF case.
- What does is the allowlist plus the browser's own PDF sandbox, where PDF
JavaScript has no DOM and no cookies.
Range requests were verified end to end: 206 with a correct Content-Range,
a byte-perfect file reassembled from three ranges, and a real browser
seeking to 10s of a 20s clip. nginx drops the upstream Content-Length on
the X-Accel path, so there is no collision.
Two settings, both defaulting on so no installation loses what it has:
clients_can_preview_files and public_listing_preview_enabled. Staff are
never gated. The anonymous side needed a route of its own — there was no
public preview endpoint — with its own throttle bucket, since a bare
throttle: shares one counter across that whole block.
A preview now logs at most one FilePreviewed per viewer per file per five
minutes: a <video> turns one deliberate act into a long tail of Range
requests, and a row each would bury the log.
Also fixes a layout bug the tests could never catch. A portal file row was
flex justify-between with three children — name, comment trigger, download
— so the middle one settled wherever the name happened to end and the
comment icon sat at a different place on every row. The name block now
takes the slack and every action lives in one trailing group, with the
comment trigger in a fixed-width slot so the icons form a column. And
because half the previewable files have no thumbnail to click — a PDF, an
mp3 and an mp4 all render as a generic icon — every row gains an explicit
PreviewAction beside DownloadAction, matching whatever style that theme
gives its download control.
The grouped filter had a second member, "All previews", built on a
public-preview action that does not exist: previewing is recorded one
way today, so its own option already answers "who previewed this?" in
full. Static analysis caught the reference; the group would have been
unreachable even if it had compiled, since a group with a single
present member is deliberately not offered.
Previews get a group here the day a second way to preview a file is
recorded separately, and the test now pins the single-member case on a
file whose log holds one flavour of download.
The installation-wide download history listed every download newest
first and offered nothing else, so "did that client ever actually
download the contract?" meant paging through everything that had
happened since.
It now filters by file name, by who downloaded it, and by date range,
in the same toolbar every other list uses: the query string carries the
filters, so a narrowed view is a link somebody can be sent.
Both names are matched against what the entry snapshotted rather than
through a join, so a file or an account deleted since is still findable
by the name it went out under — often exactly what this page is being
asked. The filters narrow the viewer's already-scoped query rather than
replacing it, so a client-scoped staffer cannot search their way to a
download of a file outside their library.
A file's history was only reachable from the library list, through the
details panel's Activity tab — so anyone who arrived at the file from a
link, a search or a notification had to go back and find the row they
came from to ask what had happened to it.
The file's own page now carries an Activity tab of its own, next to
General and Sharing: the twenty most recent entries, fetched only if the
tab is opened, and a link to the full history. It is behind the same
view_actions_log permission as everywhere else.
That full history is now filterable, which is the point of sending
somebody to it. The action list is built from the file's own log rather
than from the eighty-odd actions the software can record — all but a
handful of which can never apply to a file — and each option carries its
count. Downloads are three separate actions on purpose (a signed-in
recipient, a public link, the public group listing), so "All downloads"
asks that question once instead of three times; the group only appears
when the file's log actually holds more than one of its members.
Narrowing by who acted and by date range works the same as it does on
the main activity log, the reader's own calendar day included.
Delete a folder called Test and you could never have a folder called Test
again. The deletion worked, the folder left the screen, and the name went
with it — permanently, with an error that named a collision against a row
the interface will not show you and offered nothing to do about it.
Files and groups had it too. All three carry a unique index on slug and
all three soft-delete, so the trashed row sat in the index holding a name
nothing could reach. A public one failed outright at the validator, which
checks the table and therefore sees rows the screen does not. A private
one failed more quietly: the derived slug stepped around the trashed row
into report-2, then report-3, once per deletion, climbing forever.
The reservation was deliberate — a trashed row's slug was kept so that
restoring it could not land on somebody else's URL. But nothing in this
application restores anything. There is no restore() call, no route, no
screen; File's own comment says as much. Soft deletes are here so rows can
outlive their delete for foreign keys, the activity log and the erasure
grace period, never so they can come back. The slug was being held for a
page that could not return, and route binding already 404s the trashed row
in the meantime.
So deleting now hands the slug back, and the database is what makes that a
rewrite rather than a gentler lookup: teaching the collision checks to skip
trashed rows would leave two rows holding "report", which the unique index
rejects whatever the application thinks. The slug moves to report__deleted-42
instead. Underscores are the whole trick — Str::slug() turns them into
hyphens and Rules::slug() refuses them outright, so no derived slug and no
hand-typed one can ever land on a vacated one. That is a guarantee about
the character class rather than a hope about collisions.
The format lives in VacatedSlug rather than on the trait because the
migration needs it too and a trait constant cannot be reached through the
trait's own name — the first version of this was a fatal error waiting for
whoever ran migrations. The migration matters as much as the hook: without
it the fix only helps installations that have never deleted anything, and
every name already buried stays buried.
The collision checks still count trashed rows. It costs nothing and keeps
them honest about what the index will accept if a row is ever soft-deleted
by something that bypasses model events.
previous_file_id had this same bug and was fixed this same way, in
File::detachOnDelete — a trashed row holding its predecessor's unique slot
so the chain could never be re-linked. This is that fix, for the other four
unique indexes' worth of the same mistake. users.email is the one left, and
is deliberately not in here: an email address is a login identity rather
than a URL handle, and freeing it silently is the wrong answer.
Fixes#1645
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Failed queue jobs and read notifications both grow with use, and neither
ever shrank on its own. The failed-jobs list waited for somebody to press
"Delete all failed" — a fine tool for a backlog you are looking at, and
the only thing that ever emptied it. Notifications had nothing at all: one
row per recipient per event, kept for the life of the installation, on
what is easily the fastest-growing table here.
Both now have a retention window, set together on the Scheduler screen
under Housekeeping, and a nightly purge that honours it. Thirty days for
failed jobs and ninety for read notifications, and zero means keep
everything — the explicit choice somebody makes when a failure is evidence
rather than debris.
Unread notifications are never deleted, whatever their age. A notification
nobody has looked at is the one row in that table still doing its job, and
somebody back from four months away should find their news rather than a
clean slate. The activity log is untouched by any of this: it is an audit
trail, and it is never pruned.
Two things came out of building it. The API request log purge has been
running nightly since it shipped without ever appearing on the Scheduler
screen — so a failure of it was invisible on the screen that exists to
make failures visible — and there is now a test asserting the screen's
list and the schedule are the same list, because they had already drifted
once and would again.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Ignacio Nelson