PyO3
Bump MSRV to 1.68 ?
I have had some pain #5142 due to needing const fn features (which were introduced in 1.83). I think I see a way to workaround, however it's been somewhat painful to debug.
Working with rust versions before 1.68 (the cargo sparse index) are particularly painful to infrequently test due to needing to do huge downloads of the full crates index.
Debian Trixie has just hit feature freeze with rust 1.85. It looks like RHEL and Alpine are on 1.86 / 1.87 too.
cryptography recently bumped to 1.74 in https://github.com/pyca/cryptography/pull/12919
I would like to propose we bump to 1.68 or perhaps follow cryptography in going to 1.74.
Is there a particular motivation for considering 1.74? There is nothing there that jumps out to me.
@reaperhulk can double check my memory, but I believe we ended up on 1.74 because it was a super clear threshold in our download data where we could drop up to it with minimal impact on users but 1.75 had big impact.
That's correct. I wanted 1.83, but the metrics didn't support it so we picked 1.74 as the highest we could do. There was no specific feature in it.
Is there a particular motivation for considering 1.74? There is nothing there that jumps out to me.
I proposed just to match cryptography. Seems unreasonable to go higher than that for now!
Just for future reference: the metrics for cryptography were that 95% of all sdist downloads for the last 30 days for cryptography versions 43+ (aka roughly the last year) were 1.74+. That dropped off significantly for 1.75, and was down to 75% at 1.83.
We'll be watching this closely going forward since we have a feature actively blocked by MSRV for us right now and the upcoming Trixie release should help move the needle a bit faster (at least temporarily).