Otherwise this leaves it enabled if the processing failed, which then
makes other specs fail where they expect the image not being processed
(for example still have the initial set dimensions, instead of the one
read from image after processing).
This fix was heavily inspired by Mastodon's fix for GHSA-9928-3cp5-93fm.
So, thank you Cure53 for finding this issue, thank you Mozilla for
paying Cure53 to look into it, and thanks for Mastodon for fixing it.
Some imagemagick-versions (I tested Ubuntu 22.04 and debian bullseye)
always loose exif data when converting from jpg to webp. So this made
our CI fail now, but even if it wasn't failing before, some pods always
had and have versions which might loose the information anyway. So
having a setting to keep exif information is kinda pointless, if we
can't guarantee that the information isn't lost. Also, diaspora isn't a
photo sharing platform and we don't display exif information anywhere,
so I think we should just always strip exif data (which was already the
default before), as we don't need them.
* Dropped all references to Resque
* Moved all jobs under app/workers since that's the Sidekiq convention
* Renamed Jobs module to Worker to match new location
* Adapted all jobs to Sidekiq
* Replaced all enqueue calls with perform_async
* Dropped Resque hacks from specs and features, replaced with
sidekig/testing in RSpec and sidekig/testing/inline in Cucumber
* Updated scripts to start a Sidekiq server
* Inline Sidekiq sinatra app
* Let Sidekiq create the actual Redis instance
* Workaround already initialized constant warnings in service models
* Resolved ToDo in one job definition by creating proper exception clases
for some errors in receiving posts
* Added sidekiq section to configuration to make it completly
configurable to the user
* Add Sidekiq middleware for clean backtraces
* Delay HttpMulti retry to give offline pods a chance to come back up
* Do not retry on GUID already taken and alike errors
* Be graceful about deleted posts in GatherOEmbedData
