Registrator.jl
Registrator.jl copied to clipboard
Observations from testing
Hi, thanks for making this. I tried it out with FastClosures and got several very helpful error messages (https://github.com/c42f/FastClosures.jl/issues/12) before coming up with the following:
https://github.com/nkottary/General/pull/33/files
However I wondered whether this should also be a failure or a big warning: it changes the git tree hash of an existing version. (Obviously this will be required to fix things in some limited cases, but it seems dangerous to allow by default.)
Creating a new version, I then generated the following:
https://github.com/nkottary/General/pull/34/files
which looks better.
Now if this wasn't a test, I'd have liked to cancel issue 33 but I can't because I don't have permissions on nkottary/General. It might be nice if there was a JuliaRegistrator command to cancel a registration for cleanup purposes.
I also tried to clarify the docs a bit: https://github.com/JuliaRegistrator/juliaregistrator.github.io/pull/1
Cheers!
However I wondered whether this should also be a failure or a big warning: it changes the git tree hash of an existing version. (Obviously this will be required to fix things in some limited cases, but it seems dangerous to allow by default.)
I believe I originally had a check for this. It was quite strict and did not allow retagging old versions or tagging versions that skipped something in semver ordering. Perhaps when that was relaxed it was relaxed too much? Do we really want to relax that requirement? What was the problem with forcing version numbers to be in a reasonable order?