Официальный сайт SLAED CMS
Журнал изменений
The parser decides escaping from a trust boundary instead of a stored flag, comments and private messages gain their own runtime, and every column an author writes into is wide enough for what the editor is allowed to produce. Three schema paths — fresh install, 6.2 upgrade and the deployed patch — now agree column for column and index for index.
Core changes:
- Trust boundary and tag policy (core/security.php, core/classes/parser.php):
filterTrustedTags() becomes the single authority over [usehtml] and [usephp]
- both tags are reserved for the super administrator and stripped for everyone else
- filterText() and filterHtml() pass the capability, comments and messages never do
the pipeline runs filterCode() before filterBbBlocks(), so a documented example is quoted rather than executed
- a breaks format renders a single newline as a line break without touching authored text
the persistent parse cache moves from the news module into the parser and keys on the configuration hash
- Comment runtime (core/classes/comment.php, core/user.php):
format, iphash and the materialised path are gone; branches resolve through recursive CTEs over pid, and tombstones survive by an upward keep walk
reqkey is BINARY(16) under a unique index, so a repeated submit returns the first row instead of writing a second
- the flood window reads (ip, time, id) and the edit mark reaches the rendered view
- Private messages (core/classes/privat.php):
four independent states replace one overloaded status column, with a single owner per side and no shared flag between inbox and outbox
- Room for embedded images (setup/sql/*.sql):
fifteen columns that hold an authored text widen to MEDIUMTEXT
- EMBEDMAX allows 65536 bytes of binary, which is 87384 characters of base64 before the
data URI prefix, while TEXT holds 65535
- with STRICT_TRANS_TABLES the overflow is a lost post, not a lost image
the summary columns stay TEXT on purpose, because a list query draws twenty of them onto one page and an image referenced by address still fits there in a few dozen characters
users.sig moves from VARCHAR(255) to TEXT: of 828 stored signatures the longest is exactly 255 and 95 sit above 240
- Deployment and tooling (setup/sql/update6_3_patch.sql, setup/sql/update6_3_text_once.sql):
one idempotent patch carries a deployed 6.3 to full parity, including the mail queue and the campaign columns it never received
a one-time text repair normalises stored break markup and entities in comment and message bodies, ordered so that authored text is never mistaken for writer output
the address policy, the upload rule reader and getImgText() stop guessing the module and take it explicitly
Benefits:
an author capability decides what may be written, and the escaping policy decides how it is rendered; neither can be reached through the other
- a repeated submit, a lost connection or a double click can no longer duplicate a comment
- a branch and its tombstones are answered by one query instead of a maintained path column
- an embedded image can no longer end a post with ERROR 1406
- fresh install, upgrade and patch produce the same 38 tables, 494 columns and 172 indexes
Technical notes:
- reqkey is stored as raw bytes; the query interpolator renders non-UTF-8 values as 0x hex
comment.body, privat.body and thirteen further columns are MEDIUMTEXT; intro columns, users.block and order.info stay TEXT
- update6_3_patch.sql and update6_3_text_once.sql are both idempotent and safe to re-run
deployment order: files, update6_3_patch.sql, update6_3_text_once.sql, then clear storage/cache/pages
- no breaking change for stored content: every existing body renders unchanged
Stage 2 of docs/PRIVAT-2026.md: private messages now store the source the author wrote and render it safely by format, replacing the write-time filterHtml() encoding and the admin-side getDecodedText() compensation. Closes steps 11 and 12 of the plan.
Core changes:
- Safe rendering onto the format contract (core/classes/privat.php):
- format joins the field list and every row a read answers
addMessage() normalizes both fields and resolves the format once before the first attempt, so a retry rewrites nothing
- getBodyFormat() refuses an html editor and keeps the body as markdown source
- filterMessageText() keeps the trusted-html tokens, autolinker and censor,
drops only the write-time escape
- addMessageRow() stores format with the row
- Frontend and admin adapters (core/user.php, core/admin.php):
- detail view renders filterContent($body, true, ..., $view['format'])
- the send hands both fields over as submitted; the compose form refills unescaped
- getDecodedText() removed; the admin popover renders the same safe call
- Migration parity fix (setup/sql/table_update6_3.sql):
the bare MODIFY on _privat id/title/body becomes modcol(), so a table already on the final definition is not rewritten
- Documentation:
docs/VERSIONS.md: stage 2 entry with the deployment sequence, the mandatory tool run, the gate, and the maintenance-window requirement
- docs/PRIVAT-2026.md, docs/PARSER.md: stage 2 recorded
docs/CONTENT-CONTRACT-2026.md: handover note carrying the "without format" decision and a separately-tracked unauthenticated preview RCE finding
- Tests (tests/Support/privat_class_probe.php, tests/Unit/PrivatClassTest.php):
what a send stores per editor format, what the renderer answers, what the template does to the title, and what the writer still filters
Benefits:
No stored private-message value is trusted HTML; both frontend and admin render through Parser::filterContent() with safe = true
- The title is plain source escaped at the template boundary
- The upgrade no longer rebuilds a table that already carries the final schema
Technical notes:
Storage is source; format names plain or markdown; the '' state exists only inside the stage 2 maintenance window and the gate clears it before code is live
- tools/privat-migrate.php is mandatory and needs a closed site and a rehearsed restore
Verified: composer analyse, composer test (698 tests), php-cs-fixer --dry-run, 47 Privat tests, and a 48-check authenticated HTTPS route matrix, all passing
Steps 9 and 10 of docs/PRIVAT-2026.md still recorded themselves as uncommitted after cfa91cb8 shipped them. The progress table is what a session reads to decide where the work stands, so a stale evidence line there is a wrong answer to the next question asked of it.
Core changes:
- Progress table (docs/PRIVAT-2026.md):
rows 9 and 10 open with the commit that carries them instead of "Not committed"
- row 9 is the format column in all three update channels
- row 10 is tools/privat-migrate.php with its probe and its test
Benefits:
the plan reads the same on every machine that pulls it, which is what makes the migration resumable across sessions
Technical notes:
- documentation only, no code, schema or behavior touched
Stage 2 of docs/PRIVAT-2026.md opens: the message table gains the format column in all three update channels, and tools/privat-migrate.php is what computes the verdict per body and rewrites the stored source. No runtime code reads format yet, so this release changes nothing a visitor sees.
Core changes:
Format column in all three update channels (setup/sql/table.sql, setup/sql/table_update6_3.sql, setup/sql/update6_3_patch.sql):
format VARCHAR(20) NOT NULL DEFAULT '' as the last column of _privat
- the 6.2 channel appends it in Batch K, after the three state columns and
before the backfill, so an upgraded table keeps the column order of a fresh
one
- the patch channel gets its own section 6, because an installation that
already applied section 5 never re-runs it
the column ships empty on purpose and no statement backfills it: the verdict is per body, not per column
- Migration tool (tools/privat-migrate.php):
five modes and no sixth: report, classify, convert, title, sample
- classify writes format and the ledger storage/migrate/privat-format.json
and rewrites nothing, so the verdict is reviewed before any message changes
- convert and title are two independent ledger-driven passes that store the
value they replaced and rewrite nothing twice
- --db and --prefix run the whole thing against a restored copy first
the body classifier is the one tools/comment-migrate.php carries, because the writer is literally the same filterHtml()
the title has no column to keep its verdict in, so the ledger keeps it: a title was written by the html branch when its own row was
classify refuses a ledger whose rows a pass has already rewritten, counting the rows rather than reading a completion stamp, because an interrupted pass leaves finished rows behind and stamps nothing
Migration and tool tests (tests/Support/privat_probe.php, tests/Unit/PrivatMigrationTest.php, tests/Support/privat_format_probe.php, tests/Unit/PrivatFormatTest.php):
state E joins the probe: the table stage 1 shipped, with only format missing, which is the only shape the patch channel will ever meet in the field
the format probe boots the real core, so a fixture that names an editor is encoded by the writer an installation really runs
10 fixture messages cover both entity maps, both break signatures, the ampersand-last rule and an interrupted pass
Benefits:
the irreversible half of the content contract is rehearsable: dump, restore, classify, read the report, and only then convert
- an interrupted maintenance window resumes instead of reversing a body twice
- all three channels still end on one byte-identical table definition
Technical notes:
storage: format is '' only inside the stage 2 maintenance window, between the schema section and the gate; the gate is what makes it a state the release cannot end in
the tag map runs on the legacy branch alone, because the plain entity map turns an authored <a> back into a real tag that must stay text
- backward compatibility: unchanged, no runtime path reads or writes format yet
Stage 1 of docs/PRIVAT-2026.md. The procedural private-message subsystem becomes one Privat class, and the single status column both participants shared becomes four independent state columns, so what one of them saves or deletes no longer rewrites what the other one sees. Every frontend, block, profile and admin caller moves in the same change; no compatibility wrapper, schema detection or dual schema fallback is left behind.
Core changes:
- Storage model in all three update channels (setup/sql/):
status becomes viewed, and saved, delin and delout join it as own columns
- the backfill reads status through the new runifcol() guard before rencol()
consumes it, so the step disappears by itself once the rename has happened
- modcol() forces viewed onto TINYINT UNSIGNED NOT NULL DEFAULT 0, which a bare
rename cannot do: it keeps the BOOLEAN a column storing 2 was declared as
the single-column keys uidin, uidout and status give way to five composites
- in_box, in_new and out_box serve the mailboxes, their counts and their quotas
- out_new and flood were added on measured plans, not on expectation: the
outgoing unread badge was a full table scan and the send interval a filesort
update6_3_patch.sql gains section 5, so an installation already running 6.3 is not left behind by a release whose runtime code reads the new columns
- the stale addidx on status is deleted: the same file created and then dropped it
- The subsystem (core/classes/privat.php, core/system.php):
final class Privat owns every read and write of the table, built beside Comment
- every mailbox predicate is written once, so a list, its counter and its quota
can never disagree, and no caller restates a state column
- every mutation is transactional and authorizes itself through the detail
predicate of the side that asked; one foreign id refuses the whole batch
addMessage() answers ['id' => int, 'error' => string] from a closed code set
- the id is getSqlLastId(), not the newest row of the table, which under a
concurrent send put a foreign message id in the notification link
- both accounts are locked in one statement ascending by id, and the interval
and both quotas are read behind that lock
- the recipient is resolved before the transaction opens, because the first
plain read of a transaction fixes the snapshot every later one answers from
- a deadlock or a lock timeout is retried once, and only for a send that owns
its transaction
Mailboxes and routes (core/user.php, index.php, modules/account/index.php, blocks/user_info.php):
reads are GET routes that carry no token, mutations are POST with the token in the body; save and delete no longer travel as credentials in an address
- opening a message is what marks it read, so that one is POST and CSRF checked
one mutation route serves a row action and a bulk action alike; inbox, saved and outbox gained read, unread, save and delete over a selection
- only the opening mailbox is rendered, the other tabs are fetched on click
the detail view resolves the counterpart, so the outbox shows the sender the recipient's profile instead of their own
- the session counter cache is gone: both sidebar badges are counted per render
Administrator panel and account deletion (core/admin.php, admin/modules/ privat.php, admin/index.php, modules/account/admin/index.php, core/system.php):
the admin list reads the class, pages from the same numbers its rows came from and renders the five derived states as labelled chips
the duplicate go=5 delete route is deleted rather than re-secured; one POST handler remains behind checkAdminPost('privat') and the existing isAdmin(true)
both paths that delete a user row now clean the mailboxes in the same transaction, and the counterpart keeps a readable copy of an anonymous account
- the private-message branch of ad_status() is gone with the column it rendered
- Templates and language (templates/lite/, lang/):
table, table-row, content-list, block-content and inline-badge gained the check column, the bulk slot and the state flags; theme.css owns their styling
- six new PRIVAT constants in all six locales, none of them a scoped duplicate
- Verification (tests/):
PrivatMigrationTest drives both upgrade channels through states A to D in a disposable schema and holds every one of them to the fresh definition
PrivatClassTest holds the predicates, the limits, the independence of the two sides, the bounded bulk actions and two real concurrent sends
- SchemaUpdateValidationTest reads modcol declarations and both channels
Benefits:
- one owner of the table, one place a mailbox rule can be wrong
sender and recipient states are independent, which is what the old model could not express at all
- concurrent sends cannot both take the last free place of a mailbox
fresh install, 6.2 upgrade and 6.3 patch end on the same table, and each is safe to run twice
Technical notes:
the schema section and the runtime code of this release are deployed together: either half alone answers an SQL error on every private-message page
behavior changes, each intended: a sender may delete a message already read, a saved message stays in the sender's outbox, one participant's delete leaves the other's copy, read and unread are two actions, and a send is refused when the recipient's saved folder is full and not only their inbox
the notification preference is psmail, the setting the profile form has always offered and which had no effect while the code read the forum preference fsmail
- docs/VERSIONS.md carries the dated entry with the channel table
- docs/PRIVAT-2026.md stays: stage 2, the content contract, is not started
- stored titles and bodies keep their current encoding contract until then
Six defects found by re-verifying the finished upload migration, plus the consolidation they exposed: three copies of the SQL statement splitter became one, and the copy the module installer used was corrupting every script it ran.
Core changes:
- Remote address policy (core/classes/upload.php):
IPv6 became an allowlist of the prefixes IANA delegated to a regional registry
- 2000::/3 is far from fully delegated, so reserved space is refused by absence
- the deny-list let fec0::/10, 3ffe::/16 and everything outside 2000::/3 through
a refusal answers its own code and names the address in error_file.log
- a list behind the registry no longer looks like a host that is simply down
- Admin files module (modules/files/admin/index.php):
a delete with no valid token no longer removes the file and its rows
- save() dispatched posttype=delete outside its own token guard
- delete() disabled its check whenever a row id was given
- every handler authorizes through checkAdminPost('files') and the dial posts
- an update matching no row counts as a failed write and compensates the file
- SQL statement splitting (core/admin.php, admin/modules/*, setup/index.php):
getSqlbatch(), getSqlclean() and getSqlinfo() moved to core/admin.php
- the Inquiry tab, the module installer and setup/index.php share one splitter
- setup/index.php loads it by name and the guard of that file admits SETUP_FILE
stripslashes() over SQL is gone from both places that ran it
- it turned an escaped empty default into an unparsable column definition
the module installer fills every placeholder, not only {prefix}
- a table.sql declaring {engine} and {collate} could never install
- a comment in front of a statement is no longer part of the statement
- Smaller fixes (core/classes/parser.php, modules/files/index.php):
- a thumbnail that could not be produced renders the full size image
- a guest submitting a file binds 0 into the uid column instead of an empty string
- Verification (tests/, tools/, admin/info/):
- tools/ipv6-registry-check.php compares the shipped prefixes with the IANA CSV
- DatabaseBatchTest holds the splitter, the placeholders and the installer contract
- the uploads help page describes current behaviour and carries no changelog
Benefits:
- one address policy, one splitter, one place to correct each of them
- a refused address and a refused script are both diagnosable from the log
- the module installer and the system installer run scripts the same way
Technical notes:
behavior changes: the files admin panel is POST-only, a remote address outside a delegated prefix is refused, and the install report names statement types
- a prefix IANA delegates after this release is refused until it joins ALLOWSIX
- docs/UPLOAD-2026.md is removed: the migration it planned is finished
Batches 8, 9 and 10 of docs/UPLOAD-2026.md: admin upload moves onto the Upload class and the procedural pipeline is deleted, the 2026 format set goes live across every configuration and render path, and the final sweep proves nothing legacy survived. Two audits of that sweep are folded in, each of which found real defects rather than style.
Core changes:
- Admin upload and settings (admin/modules/uploads.php, config/uploads.php):
uploadsave() publishes local first and remote second through the class
- every limit comes from the all record, nothing is hardcoded at the call site
- the all record carries 104857600 bytes and 1600x1600 to match
configsave() validates extensions against Upload::getSupportedTypes()
- an unsupported extension is dropped, reported and never stored as enabled
- both extension fields show the supported list as a hint
- upload(), check_file() and check_size() are gone with their last caller
- The 2026 format set (config/uploads.php, config/filetype.php, config/files.php):
the canonical 21 formats replace the old 26
- ten formats withdrawn, each on its own evidence, two spellings corrected
- modules gain only webp and avif; the full set reaches typ and the all record
config/filetype.php becomes five family templates over 21 keys
- audio renders through audio, video through video, pdf through object
- the dead screens class is gone and the lightbox binds to sl-attach
- Upload class hardening (core/classes/upload.php):
images must decode, not merely parse a header
- a truncated png passed getimagesize() and was published before this
- a missing decoder answers unsupported instead of passing the file through
- the remote write callback rejects a short write instead of counting it whole
- the address policy blocks every IPv6 range IANA marks as not globally reachable
- the class owned naming grammar is matched at its fixed salt length
- Configuration writer (core/system.php, setup/index.php):
a value whose line would pass 180 characters is written as a concatenation
- the split is deterministic, so an unchanged save round trips byte for byte
- the installer copy follows, or a fresh install would write a different file
- Verification (tests/, tools/upload-route-check.php):
UploadIntegrationTest drives the resolver and the accessor against real config
- all fourteen module records survive resolve and serialize byte for byte
- the five adapters are held to the ordering rule of the plan
tools/upload-route-check.php walks 47 rows of the route matrix over real HTTP
- credentials come from the environment, TLS verification is on by default
- every scenario compares tree and database around itself and cleans up
Benefits:
- one publication primitive, one rule parser, one type policy, one lock location
- every publishing flow validates by content rather than by extension alone
- the acceptance of the plan is reproducible from the repository, not by hand
Technical notes:
- ext-gd joins ext-curl and ext-fileinfo in the composer require block
behavior changes: content stored with a withdrawn format renders as a plain link, a guest no longer sees historical files in the editor panel, and a build without a GD decoder refuses that image format instead of accepting it
nine other configuration files still hold over-long lines and will be rewritten in the wrapped format the first time each is saved from its own settings tab
stored reference formats are unchanged: avatar filenames, project relative _files.url values and the editor JSON shape all keep their existing contracts
setContentActive() still runs before the checked UPDATE of the admin files handler; that ordering predates this migration and is left to its own task
The last three findings of the backup and scheduler plan, plus the comment rule applied to the code that work produced. The plan document is removed now that every batch of it lives in the code.
Core changes:
- Crash reconciliation reports its own failure (system.php, scheduler.php):
updateSchedulerCrash() returns null when the repaired state cannot be stored
- unlock() shows the state directory instead of its success message
- addSchedulerRun() refuses to start rather than running an unrecorded job
- Two regression tests drive it through a read-only state file
- POST-only moderation (comments.php):
The six write handlers read their payload from the body alone
- typ is read once, so a valid typ=0 is no longer swallowed by ?: and
replaced by a crafted query value
- status, modul, search and chng lost their query fallbacks as well
- A source-level test asserts no handler reaches for get or req
- Isolated mail probe (mail_probe.php):
The disposable schema is registered for cleanup before any table is copied, so a failure during setup cannot leave it behind
- The class docblock states what the probe really works against
- Comment placement (system.php, helpers.php, comments.php, probes, tests):
- Comments moved out of function bodies into the headers above them
- Two functions that had no header comment received one
- Overlong comment lines shortened instead of wrapped
Benefits:
- An operator is never told a crashed job was cleared when it was not
- The admin boundary cannot be steered by the address bar
- The test suite no longer depends on the site mail queue
Technical notes:
- No storage or configuration format changes
Backward compatible: the removed query fallbacks were unreachable from the rendered forms, which have posted since the boundary was introduced
Batches 6 and 7 of docs/UPLOAD-2026.md. Every remaining visitor-facing publisher now reaches the filesystem through Upload::addUploadedFile() after its own method, token and business checks have passed, so a preview, a refused form or a failed row write can no longer leave a file nobody references.
Core changes:
- Account avatar (modules/account/index.php):
saveavatar() gates the whole operation behind POST and the account token
- the old line validated the token only when op arrived in the POST body,
so GET ?op=saveavatar&avatar=<preset> wrote users.avatar with no token
- the preset branch now sits inside that gate and still calls no upload
the upload publishes through the class and stores the returned filename
- the rule is built from the $conf['users'] avatar keys, maxquota 0
- adirectory loses exactly one leading uploads/ before it is passed on
- a failed profile write deletes the exact path this request published
- Avatar preset grid (templates/lite/fragments/table-row.html, theme.css):
- the is_avatar_link branch gained a form/button variant keyed on action
sl-avatar-link moved from the anchor onto the form and got the two rules that strip the button chrome; line-height: 0 keeps the row height intact
- Frontend file module (modules/files/index.php):
send() calls the class only when the save passed every check
- a file counts as submitted when its error is not UPLOAD_ERR_NO_FILE
- a refused file no longer falls back to the typed external URL
- the gate also reads $conf['files']['upload'], which the old code ignored
- the row write is checked and compensated against the returned path
- Admin file module (modules/files/admin/index.php):
save() publishes straight into the selected path field, else the configured one, so the publish-then-rename branch no longer runs for a new upload
- relocation of an already stored file with no new upload is unchanged
- preview and delete publish nothing at all now
both row writes are checked; a failure deletes the published path and logs a stranded file through Logger::addFile()
- Plan record (docs/UPLOAD-2026.md):
last completed batch, the decisions of batches 6 and 7, and what the route tests actually proved on the stand
Benefits:
- no upload path can publish before authorization, CSRF and validation pass
- stored size and path come from the class instead of a probed request path
- one publication primitive, lock-guarded and collision-free, for five flows
- the admin file handler writes each file exactly once
Technical notes:
stored reference formats are unchanged: users.avatar keeps the filename only and _files.url keeps the project-relative uploads/<dir>/<file> shape
Breaking: avatar presets are POST-only now, so a bookmarked GET link that set a preset stops working
Breaking: an operator publishing through the frontend file form no longer gets a name without an owner suffix; the flow stores the user id or 0
config/files.php still lists gzip, which the type map does not know, so a .gzip upload is refused until batch 9 corrects the spelling to gz
admin upload and the format switch are untouched; upload(), check_file() and check_size() stay until batch 8 removes their last caller
The procedural upload pipeline is being replaced by a single Upload class that owns validation, transfer, naming, quota and atomic publication for every publication flow. This lands batches 1 to 5 of docs/UPLOAD-2026.md: one configuration parser, the image pipeline the new formats need, the class itself with its local and remote halves, and the Toast UI editor moved onto it.
Core changes:
- Upload rule resolver (core/system.php, admin/modules/uploads.php, core/admin.php, core/classes/parser.php, core/helpers.php, plugins/editors/toastui/driver.php):
getEditorUploadData() renamed to getUploadRuleData() and widened to all twelve named fields
- every consumer reads named keys, so no call site splits the configuration string any more
- setUploadRuleData() assembles the same twelve fields and configsave() writes through it
- getUploadModuleList() derives the settings tab module list from the records themselves
- checkEditorUploadAccess() takes the resolved rule array instead of raw field indexes
- Image pipeline (core/system.php, core/classes/parser.php, templates/admin/assets/css/base.css, templates/lite/assets/css/base.css):
create_img_gd() renamed to getImageThumb() and its WBMP/IMAGETYPE_SWF branch removed
- IMAGETYPE_WEBP and IMAGETYPE_AVIF added behind function_exists(), so an older GD build degrades
- the image lists and the data-URI pattern accept webp and avif
- .sl-attach and its alignment modifiers added to both themes, built from existing spacing tokens
- Upload class (core/classes/upload.php, core/system.php):
new class with addUploadedFile(), addUploadedFiles(), addRemoteFile() and deleteStoredFile()
- the configured allowlist plus an explicit extension to MIME map; executable and web-active formats are refused whatever configuration says
- transfer into .upload-<hex>.part, then the destination lock, the stale partial sweep, the quota recheck, collision retry and one rename
- remote fetch over ext-curl only: address policy, redirect revalidation, pinned address, connect and total timeouts, header and streaming byte limits
- getUploadService() builds the single instance over UPLOADS_DIR and LOGS_DIR/uploads
- Editor adapter (core/system.php, index.php):
addEditorUpload() is a thin adapter over addUploadedFiles()
- the result codes map onto the existing language constants, so no new constant is introduced
- the owner is the site user id, no suffix for a privileged moderator and the guest value otherwise
- getEditorFileJson() answers a guest with an empty list instead of every file of the shared guest suffix
- the generic go=4 default answers 400 with the editor JSON instead of reaching the old upload()
- Tests and dependencies (tests/Unit/UploadContractTest.php, tests/Support/upload_probe.php, tests/Unit/ImageThumbTest.php, tests/Support/image_probe.php, composer.json):
- the contract suite covers every local result code, the naming grammar, the lock, quota boundaries, the partial sweep and every remote rejection path without opening a socket
- ext-fileinfo and ext-curl added to require, because the class fails closed without them
Benefits:
- one validation, naming and publication path instead of a rule copy per call site
- uploads are verified by content rather than by extension alone
- remote fetch fails closed for private, reserved, redirected and rebound targets
- concurrent publications to one directory can neither overwrite each other nor oversubscribe the quota
Technical notes:
- stored reference formats are unchanged: users.avatar filenames, project relative _files.url values and the editor JSON shape
- lock files live in storage/logs/uploads, an existing ignored tree; no new directory for user files
- behavior change: a guest no longer sees previously uploaded editor files and receives only the files of the current upload response
- the source path of an upload must stay exactly as the SAPI reported it, or move_uploaded_file() refuses it on Windows
- account avatar, both file modules and admin upload still run the procedural pipeline; batches 6 to 10 of docs/UPLOAD-2026.md migrate them and delete it