Closed bartelink closed 4 months ago
Mmh, I'm not sure about this one. If I remember correctly we made the move to FSharp.Core 8 because @SchlenkR had other build issues with the dependency on 5. But you're right, the 14.0 release was bad because of the implicit dependency. That was my fuckup.
But... it builds fine?
This scheme/policy is used in lots of places, e.g. Argu, github/jet
Ultimately, apps will typically have an implicit ref to the latest FSharp.Core, and are in a position to control that, and aside from not inducing risk by depending on a known broken/blacklisted thing like e.g. Newtonsoft < 13.0.3, it's just not something that a library should be dictating
Right now, people in the same boat as was (actually, a colleague) can fall victim to this. The remedies (aside from hoping they hit this via google) are one of:
@bartelink I only had a look at the issue; not at your PR. You basically did the work I finished right now. Please excuse me for not merging your PR; I would have done it if I had seen it before.
Thank you so much for your PR and all the information. This is so valuable, since such topics usually cost nerves and energy. Great stuff - 🙏❤️
Thanks @SchlenkR
One straggler piece is that the comments don't align with the PackageRef now, will make a PR
This is not critical - you've already done all the important stuff, and it's much appreciated...
FsHttp package versions 13-14.4.0 omitted to state their dependency on
FSharp.Core
This means that anyone who is using a
FSharp.Core
prior to8.0.100
(which v14.4.1
correctly declares its dependence on) winds up (if they use a tool that depends onnuspec
dependency declarations (such aspaket
, but also Rider etc):14.4.0
(technically correct, not apaket
bug)net6.0
runtime) getting surprised at runtime with a:This PR rolls all the way back to only demanding V5, like the older versions did
resolves #183