v0.4 Plans

[juntyr]

• add features for impls that need nightly features: sync_unsafe_cell, exclusive_wrapper, maybe cfg_target_has_atomic
• replace never_type feature with the never say never hack
  • impls must refer to the actual type, not a type alias
  • add a feature gate for ! impls
• remove usage of specialization feature
  • inhabited: Simplify the inhabitedness computation #11
  • typeset fixed-point computation: Extract the type layout graph without using the unstable+nightly specialization feature #17
  • extracting the discriminant for the typeset
    • perhaps we can just store the discriminant in the largest possible type since its type is already implicitly encoded in the repr and the typeset?
    • store the discriminant value in size-independent negabinary representation, since the discriminant type is already implicitly encoded in the repr and the typeset: Update discriminant value encoding #16
• check if decl_macro feature is still necessary
• check if generic_const_exprs feature can be eliminated or put behind a feature
• bump MSRV to 1.83 at least to use const_mut_refs
  • we can be forward compatible so there is no rush to upgrade the MSRV right now
• upgrade to syn v2 Upgrade to syn v2.0 #13
  • update test that was blocked

[juntyr]

The discriminant should be computed by the derive macro using syntax only - this is equivalent and doesn't use unstable or internal features

[juntyr]

I'm debating whether to keep the internal usage of specialization in v0.4 so that v0.4 can be released soon-ish (since the removal will likely take some more experimentation and time).

[juntyr]

I'm debating whether to keep the internal usage of specialization in v0.4 so that v0.4 can be released soon-ish (since the removal will likely take some more experimentation and time).

Spoke too soon :D #17