Closed RyanGlScott closed 3 years ago
Thanks for looking after that—I had overlooked the loss of SafeHaskell. I'll re-release with this today if I can.
Seems to be working; I'll release this as 2.6.4.1 now.
Do I understand right that network-uri-0.6.4.0
dropped Safe
annotation? (And wasn't even Trustworthy
)?
I think that is a violation of PVP (spirit) and that release should be blacklisted (with build-depends: base <0
) revision. Otherwise there a chances the issue will resurface.
I consider dropping Safe
a major breaking change, it does break stuff which rely on it.
Right, 2.6.4.0 lacked the Safe annotation, due to an oversight.
I'm thinking it may not be such a big deal, since the solver is very unlikely to choose 2.6.4.0 while there is a 2.6.4.1?
But I agree with the principle of forcing it not to be used. As far as I know, the best way to handle that is to mark 2.6.4.0 as "deprecated" in Hackage. It sounds like you're suggesting another way, that I'm not familiar with.
I'll wait to take action until I hear from you, @phadej.
Deprecated annotation on Hackage means "you can use it if you need to", where making revision with base <0
will say "never use it".
TL;DR, Deprecations of particular versions are not that useful, but they predate the revision mechanism.
I noticed that 4f8c4ffce8e16eff7a09adfbbbacc5bee29b585f uncommented the use of
{-# LANGUAGE Safe #-}
inNetwork.URI.Static
. I think this is (unintentionally) my fault, as I forgot to markLanguage.Haskell.TH.Syntax.Compat
asTrustworthy
in the initial release ofth-compat-0.1
, and as a result,Language.Haskell.TH.Syntax.Compat
was inferred to beUnsafe
due to the fact that it imports fromGHC.Exts
. I have since releasedth-compat-0.1.1
which marks that module asTrustworthy
, which allowsNetwork.URI.Static
to be marked asSafe
.