Release v6.3.0 (2023-04-05)

Published 2023-04-05 08:13:29Z

• fylr_checksums.txt

• fylr_v6.3.0-fix1_windows_amd64.zip

• fylr_v6.3.0_darwin_amd64.tar.gz

• fylr_v6.3.0_darwin_arm64.tar.gz

• fylr_v6.3.0_linux_amd64.tar.gz

• fylr_v6.3.0_linux_arm.tar.gz

• fylr_v6.3.0_windows_amd64.zip

Highlights 🎉

Changes and improvements all over the place. Our Zoomer got a lot faster now (make sure to use our latest iiif service configuration, see below) and we added a re-sync button to produce new file versions easily.

Known Problems

Under Windows, you need to remove the plugin easydb-barcode-display-pdf-plugin, or else fylr cannot start. Fixed in the added download fylr_v6.3.0-fix1_windows_amd64.zip.

New

• /inspect/files got a new Re-sync done button. It can be used to re-sync done originals. That is helpful after the base config was changed to update all auto produced versions.

• Includes public OAUTH2 clients for integrators & our mobile App.

• New endpoint /api/event/stream allows to open a Websocket listening to fylr's event stream. Latest web frontend uses that too, so no more polling for events. With this change, updating the search after updating a record should be almost immediate.

• fylr.services.api.oauth2Server.allowHttpRedirects: This new settings allows to configure OAUTH2 in HTTP only environments. Before this was only supported for http://*.localhost. This is useful in intranet environments where it is hard to obtain HTTPS certificates regularly.

• Introducing chain loaded fylr.yml. Now it is possible to use multiple -c on the command line to pass in more than one fylr.yml. A new internal fylr.default.yml is a starting point which can be used for most configurations. It includes a fully configured execserver (for Linux & Mac OS systems). The new .yml format allows to replace, add and delete value from config files which were loaded first. Current fylr.yml setups continue to work, as the new syntax requires a top level property fylr+: to use the merge functionality. More information can be found in fylr.example.yml.

Improved

• Much faster tile production for IIIF and Zoomer output. The new code uses a BMP interim format and fast system memory mapping to achieve fast tile production, especially for big images. You need to add the convert command to the iiif service in your fylr.yml:

fylr:
  services:
    execserver:
      services:
        iiif:
          commands:
            convert:
              prog: convert

• Zoomer endpoint now also supports .png output.

• Support encrypted data in export plugin transport options. If a property ends in :secret it's value will be stored safely in fylr's database and only decrypted when presented to the transport plugin.

• Actually use password requirements as defined in the base config. Before we only stored the regexps, but never enforced them. Administrators can overwrite the policy and have the server ignore it.

• Added preview for pptx. Before we only had ppt support.

• No more standard cached are used on /api/db endpoint. This makes sure that linked object data gets merged from the current database rather than from the standard cache which follow the indexer.

• Fast indexer wake-ups in system which have the indexer running on the same fylr binary as the API service. This makes the indexer less laggy and together with the new event stream facility, the web frontend can show changes almost immediately.

Fixes

• Reading images width & height was buggy if the EXIF data contained different values. Fixed by relying solely on exiftool Composite information about the file size. This causes the Zoomer to behave erratically.

• Parallel access to WebDAV volumes has been fixed. rclone easily runs 8-10 parallel tasks to sync files, that caused our WebDAV endpoint to produce errors pretty quickly.

• Certmagic now works on Port 80 to pose the Let's encrypt challenge. With that issuing HTTPS certificates gets easier behind certain firewall configurations.

• /api/search/parse got more syntax fixed. This time a simple OR case failed to compile correctly.

• Fixed aggregations in hierarchies. If hierarchies were filtered by parent, in bigger lists the filter was filtering the wrong data and not returning any results.