Re: rust - volunteers sought...

[Greg Troxel <gdt%lexort.com@localhost>]

pin <voidpin%protonmail.com@localhost> writes:

>> It is perfectly feasible to install the older version on amd64 and test
>> programs with it.
>
> Bear with me and try to understand my system.
> For me Rust is a priority, so I'm always on the latest possible version.
>
> Currently, it's not possible to have different versions installed.

pkg_delete rust; pkg_add rust-old
test
pkg_delete rust; pkg_add rust-current

> And, although I start to see the point of versioning Rust, I'm still
> not convinced this is a solution or a source to further problems.

Not trying to nitpick language but I want to be clear on goals.

Solution would be fixing rust culture and that's intractable.  We're
talking about the best strategy for pkgsrc to cope with a messy world.
That world is changing over time.

I see the options as:

  single version, hold updates until they work on a set of core
  platforms (what we do now)

  versioning

  single version, declare that earmv7hf is not longer a viable platform


so if we're ok with "hold updates" for now, that's ok   I see the third
option as not really ok, if versioning would help it along even if it
turns out to be 2nd class.

> With that out of the way, I always check upstream MSRV. Unfortunately,
> not everyone defines that variable in Cargo.toml.  So, what I usually
> do is, push the update, keep an eye on bulk builds and revert the
> update when needed.

There are not necessarily regular bulk builds for the platforms of
interest.

But, for our current world of single version, if you have installed the
version in pkgsrc, then it's fine to commit to pkgsrc if you built with
that version.

> I know you might disagree with this but, it's not that bad and usually
> there aren't that many failures when we have Rust at latest "-1".

I don't disagree with that.  But I thought that was our normal state.

> Now, if we are talking about Rust "-2", the risk increases but, I've
> found it to be very similar to "-1".  "-3" starts to be messy and
> quite a few things won't build with it.

So where are we now?  To me the big question is:

  if we don't have a rust update before the branch, is that ok?

if that's ok, then we can talk about it on 7/1.

If that's no ok, we have a bigger problem.