Skip to content

2.1.0

Latest

Choose a tag to compare

@dilux-release dilux-release released this 08 Oct 19:27
745d187
  • Enable Offloading looks for files added since the last sync first: it says how many there are and uploads them before offloading is turned on. A file added between the end of the sync and enabling offloading used to stay off the cloud, a broken image once offloading was on. A synced library also offers Scan and Complete Sync on Sync & Offloading › Sync.
  • Opening Sync & Offloading › Sync more than 90 seconds after a sync finished no longer resets the plugin to "configured", which turned offloading off without a word.
  • Complete Sync uploads the files added to uploads/ since the last sync; it used to report success and leave them out.
  • Deleting an attachment while offloading is on removes its copies left on this server too, not only the objects in the cloud.
  • Deactivating the plugin before the first sync keeps its state; it used to come back as "synced".
  • A site whose server path has a folder named cache, temp or tmp above uploads/ syncs its files; every file used to be skipped. Those folders inside uploads/ are still left out.
  • Cloud Provider › Credentials shows the answer of Test Connection; it was written but stayed hidden.
  • Clear List empties the failed files list on Sync & Offloading › Sync; it used to say so and leave the list as it was.
  • A browser tab that lost the sync to another tab stops when the sync finishes, instead of asking the server again and again.
  • The size of a file that is only in the cloud is its real size (the Media Library shows it for files without a stored size); it used to be 0.
  • The DiluxOne Offload menu sits at the end of the admin menu, after Settings, where WordPress places a plugin's menu by default.
  • A sync or download request keeps to its time limit when the server's clock is corrected while it runs. Before, a clock set back made the limit look untouched, and the request kept starting uploads past it.
  • A delete or an upload the storage service asks to slow down (a 429; Google allows one change a second to the same file) is tried again after a second. Before, deleting an image right after editing it could leave its files stored, and billed, in the cloud.
  • Uninstalling cancels the S3 multipart uploads a sync left unfinished, so the bucket stops billing parts nobody can complete any more.
  • Before the first sync, Sync & Offloading › Sync checks this site's folder in the container or bucket. If something is already there (an earlier install, a staging copy) it says how many files and how much, and lets you continue with them, empty that folder (typing the container's or bucket's name) or connect another one. Emptying never touches anything outside this site's folder.
  • A sync whose uploads go through ends a pause the connection health had recorded (a key that failed and was fixed since), instead of showing "Paused" for up to five more minutes.
  • An upload, a delete or a check that meets a temporary error from the storage service (a 500 or 503, a dropped connection) is tried up to three times in all, instead of failing at once. Backblaze B2 answers that way to about one upload in a hundred, which could leave an image without one of its thumbnails.
  • The initial sync uses every parallel upload from the start: a library that begins with large files used to send them one at a time, and now sends them alongside small ones, each upload starting as soon as another finishes.
  • A large file's parts (above 10 MB) now go up in parallel, like the other files, instead of one after another while the rest of the sync waited. A part that meets a temporary error from the storage service is sent again instead of failing the file. A very large file no longer has to go up within one request of the sync: the next one takes it up where it was left, sending only the parts the storage service does not have yet.
  • The sync and Disconnect from Cloud reuse their connections to the storage service from one group of files to the next, instead of opening a new one, with its secure handshake, for every file.
  • Status › Health names the storage service in use (Cloudflare R2, Amazon S3, …) and no longer shows a card that repeated the plugin's state. Screens that speak of the storage say "container or bucket", and a hostname too long for its card (Cloudflare R2's public one) is shortened, with the whole name on hover.
  • The Sync, Offloading and Disconnect screens and their windows are built on WordPress' own buttons, notices and cards, and take the accent of your admin colour scheme. Offloading's bar shows, in one full bar, the files only in the cloud and, striped, the ones that still have a copy on this server (the disk Delete Local Files can free). While a sync has files pending, the ones that already failed show as a red part of its bar. Messages these screens showed in English only are now translated.
  • Six more services in the S3-compatible provider's Service list: Hetzner Object Storage, Akamai (Linode) Object Storage, Vultr Object Storage, Scaleway Object Storage, OVHcloud Object Storage and IDrive e2. Each fills in the endpoint and, where the service has one, the public URL, and says where its keys come from and how to make the bucket readable.
  • A file too large for the storage service's limit on parts goes up in larger parts instead of failing: past about 5 GB on Scaleway (1,000 parts), past about 48 GB elsewhere.
  • Settings › Serving › New uploads: browser caching, on by default, stores Cache-Control: public, max-age=604800 (one week, editable, or off) with every new upload, so browsers and CDNs keep your media instead of asking the storage again. And a storage class for new uploads: Standard, or the cheaper infrequent-access class (Azure's Cool tier, STANDARD_IA on Amazon S3 and Cloudflare R2), which charges per read and bills a minimum of 30 days. Files already in the cloud keep what they were stored with.
  • The notices after saving settings, the provider or new credentials are translated.
  • Settings › Transfers › Folders the initial sync leaves out: one folder per line under wp-content/uploads/ (backups, caches); the scan skips them and lists them on the Sync screen. Files uploaded through the site while offloading is on still go to the cloud.
  • An e-mail to the site's administrator address when uploads pause (the storage failed three times in a row) and one more when they resume, never one per failure. On by default; Settings › Logging turns it off.
  • Tools › Site Health has a test for the connection to your storage, critical with the reason when uploads are paused, and an Info section with the provider, the service, the state and the version (never a key).

What was merged since 2.0.0

✨ Features

  • sync Enable Offloading uploads files added since the sync first; Synced offers Scan (#49)
  • settings folders the sync leaves out, an e-mail when uploads pause, Site Health (#42)
  • settings browser caching and a storage class for new uploads (#41)
  • s3 presets for Hetzner, Akamai, Vultr, Scaleway, OVHcloud and IDrive e2 (#39)
  • admin the screens on WordPress' own styles, and the Offloading and Sync bars redrawn (#38)
  • sync one pool of files and parts, and a large file resumes in the next request (#30)
  • sync before the first sync, say what the target already holds and offer to empty it (#22)

🐛 Fixes

  • providers a request the service throttles (429) is retried after a second (#50)
  • twelve bugs found by tests covering every class, hook, screen action and storage behaviour (#48)
  • security align with the newer wordpress.org pre-review: nonces, menu, escaping, extract (#47)
  • sync time budgets are measured on the monotonic clock, which a clock change cannot set back (#46)
  • uninstall cancel unfinished multipart uploads, and make pre-pr fails on i18n warnings (#43)
  • admin the target callout changes the bucket, Health names the service, long hosts shortened (#37)
  • providers send a request again when the storage service answers with a transient error (#27)
  • sync uploads that go through end a recorded pause (#21)

📝 Documentation

  • rules every failure path of a provider call cleans up (#26)
  • readme a map of the repository: the plugin, where to get it, how people and agents contribute (#31)
  • plan 2.1.0 folds in 2.2.0 and 2.3.0; the plan before the code (#20)

✅ Tests

  • real Google's stale reads of correct objects settle by stored MD5 (#60)
  • fix the flaky tests the CI history shows, and review-local recovers a broken checkout (#44)

📦 Build and dependencies

  • make plugin-check fails when the check itself does not run (#57)
  • deps-dev Bump proxy-addr from 2.0.7 to 2.0.8 (#56)
  • deps-dev Bump the dev-dependencies group across 1 directory with 2 updates (#24)
  • deps-dev Bump brace-expansion from 2.1.4 to 2.1.7 (#33)

👷 CI

  • adopt the organisation's v5 setup (#54)
  • review the paid review waits for every deterministic check, real-storage suites included (#40)
  • tests Dependabot pull requests skip the keyed real-storage suites, like forks (#32)
  • tests run the real-storage journeys against Backblaze B2 as well (#23)

🔧 Maintenance

  • release 2.1.0 (#61)
  • policy the commented example names Sonnet 5.5, the organisation's reviewer (#45)
  • dev make pre-pr runs every local check; an AI-first README and AGENTS.md guide you (#29)
  • dev make pre-pr runs the checks and the local review before a pull request (#28)
  • release main records 2.0.0; the versioning policy and the new roadmap are written down (#19)