Winit: New versions of `parking_lot` are failing builds for 1.24.1
The minimum compiler version that winit tests is 1.24.1, but parking_lot has been updated and the version we're pulling no longer supports that version of the compiler.
Should we bump the minimum version, or make parking_lot's version explicitly support 1.24.1?
Osspial
>All comments
Bumping to 1.26 would have the added benefit of letting us use impl Trait, though I'm only aware of one place where we'd actually use that. Having the flexibility is always nice, though.
1.26 is 4 months old now, but is still too new for Ubuntu: Amanieu/parking_lot#94. 1.24.1 is also the version used in Debian stable, though I assume most desktop users don't use that. Either way, it's likely that anyone developing with winit would be willing and able to use rustup, and they'd hopefully ship binaries of their application to end-users.
I imagine that pinning to an older version of parking_lot could be more likely to cause annoyances for people than moving to 1.26 would (I know for instance that servo is unhappy when multiple versions of a library are pulled in). I tried to evaluate which downstream packages would be affected, but I found that surprisingly few of them target a specific rustc version... the upcoming ggez version is unaffected, at least.
On our end, changing our minimum to 1.26 is the simplest thing to do (especially since it will inevitably become necessary in the long term), and it could easily be reverted if it somehow became a big issue for people.
francesca64
on 9 Sep 2018
Related issues
chemicstry
路
3Comments
francesca64
路
5Comments
JDTX
路
3Comments
chrisduerr
路
3Comments
swiftcoder
路
3Comments
Most helpful comment
Bumping to 1.26 would have the added benefit of letting us use
impl Trait, though I'm only aware of one place where we'd actually use that. Having the flexibility is always nice, though.1.26 is 4 months old now, but is still too new for Ubuntu: Amanieu/parking_lot#94. 1.24.1 is also the version used in Debian stable, though I assume most desktop users don't use that. Either way, it's likely that anyone developing with
winitwould be willing and able to userustup, and they'd hopefully ship binaries of their application to end-users.I imagine that pinning to an older version of
parking_lotcould be more likely to cause annoyances for people than moving to 1.26 would (I know for instance thatservois unhappy when multiple versions of a library are pulled in). I tried to evaluate which downstream packages would be affected, but I found that surprisingly few of them target a specificrustcversion... the upcoming ggez version is unaffected, at least.On our end, changing our minimum to 1.26 is the simplest thing to do (especially since it will inevitably become necessary in the long term), and it could easily be reverted if it somehow became a big issue for people.