fluxengine
fluxengine copied to clipboard
davidgiven
Better testing needed
It's far too easy to break things. The existing encode/decode tests check for consistency between the encoders and decoders, but that's not really enough. We really need a corpus of disk fluxes that the unit tests can decode and compare against goldens. Sadly, I don't think github would be happy with me checking in about 50 ~50MB files.
i have a few captures we can use unfortunately yeah, file size is limited to 50-100mb per item which is about 3-5 revs of SCP... 50mb is the complaint-limit, 100mb is the hard limit. it should definitely be in a different repo tho, to avoid folks accidentally forking a gig.
im also happy to store my home grown fluxes on adafruit's CDN or archive.org and use github LFS to point to the files https://docs.github.com/en/repositories/working-with-files/managing-large-files/about-git-large-file-storage
I've made a repo, https://github.com/davidgiven/fluxengine-testdata, containing some test images. I'll see how that goes. I can always delete it...
If anyone has any flux files of known good blank disks (to avoid copyright issues), I'd love to have them.
what about disk images with non-copyrighted data (public domain text, images) - might be a better way to verify we read the sectors right?
"Solved" fluxmaps (or those generated by fluxengine) should compress a lot harder due to not having any noise. If they don't, maybe that's something we should fix.
