Release v6.6.1 (2023-08-10)

Published 2023-08-10 07:16:37Z


  • If a fylr+.yml contains an empty command declaration in it could happen that the loader received a panic and fylr didn't start. An easy fix is to set the empty command to {} which avoids this problem. However, this fixes allows an empty config and avoids the panic.

  • Setting up the automatic environment for helper programs called by programs which are actually configured in failed for configurations where programs in were configured with relative or absolute path names (and not just with the program name). In v6.6.0 the FYLR_CMD_* environment variables can still be set manually to avoid this problem.

  • Fixed overwriting of the default set FYLR_CMD_* environment variables using the top level setting This bug didn't really show in most configurations because of the previous bug. Also allow to set an empty program. Can be used to disabled ffmpegthumbnailer.

  • Fixed setting environment on the level. This works for the first command only, if you want to set environment for commands which are started by the main service command, you need to set the environment on the fylr.servicees.execserver.env level.

  • /inspect/file re-sync has been fixed. If re-syncing multiple versions at once it could easily happen that unique index violation would occur. The error looks something like pq: duplicate key value violates unique constraint "file_u3"". The underlying problem has been addressed and re-syncing should work more reliably now.

  • If pdftotext returns an error while parsing text from PDF documents, this error is now stored as the full text information of the PDF. It is a common error that no text can be extracted for secured PDF. Before errors would have the metadata production fail and thus the whole produce chain would stop executing, not producing any thumbnails or other previews.

  • Adjusted timeouts for pdf and video recipes to 1 hour; was 15 and 30 minutes, resp.

  • /api/schema: When sending pre-defined column-ids for reverse_link and link types, they would be ignored, leading to sync problems with centrally stored data models.

Last updated