From: | Brian Cosgrove <cosgroveb(at)gmail(dot)com> |
---|---|
To: | Smolkin Grigory <smallkeen(at)gmail(dot)com> |
Cc: | Bradford Boyle <bradford(dot)d(dot)boyle(at)gmail(dot)com>, Christoph Berg <myon(at)debian(dot)org>, pgsql-pkg-debian(at)postgresql(dot)org |
Subject: | Re: Packages for pg_doorman |
Date: | 2025-03-16 11:30:29 |
Message-ID: | CAGKCzNCUsK4voXbKB2P8PHAQrTUJr43o-+-Qav9MLFPq7wh1sQ@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-pkg-debian |
For such a new piece of infrastructure is there really a compelling reason
to support distributions older than Debian bookworm (released 3 years ago)?
Do you have any current use-cases for something older?
On Sun, Mar 16, 2025 at 4:17 AM Smolkin Grigory <smallkeen(at)gmail(dot)com> wrote:
> > Can we use cargo vendor?
> > https://doc.rust-lang.org/cargo/commands/cargo-vendor.html
> >My understanding is that Debian policy >does not permit downloading
> >during the build [0].
>
> >Would it be acceptable to install the compiler from https://rustup.rs/
> during package building?
> >> My understanding is that Debian policy does not permit downloading
> during the build [0].
>
> Ok, I suppose we have to live with distribution compiler. Though, it may
> make some old systems without rust 1.70.0 unsupportable.
>
> >> According to this post [1] on the debian-rust mailing list, "vendoring
> should be avoided unless there is a very good reason not to".
> >> The Debian Rust packaging team has a pretty comprehensive guide [2] and
> it looks like they have tooling to help assist with the packaging process.
>
> I think we have a very good reason here - a lot of distributions of
> different versions with different libversions. Even making it just to be
> "compilable" would require tremendous effort, supporting it will be
> unmanageable.
>
> I see no good reason not to use vendoring in case of "multiple
> distribution"-build process.
> 1. It's a tarboll artifact, released with main project, under the same
> release tag.
> 2. It is reproducible and has public hash, so it can be verified.
> 3. It allows local build, nothing has to be downloaded.
> 4. Consistent binary behavior across multiple distributions.
>
--
Brian Cosgrove
From | Date | Subject | |
---|---|---|---|
Next Message | Smolkin Grigory | 2025-03-16 13:23:26 | Re: Packages for pg_doorman |
Previous Message | Smolkin Grigory | 2025-03-16 07:32:39 | Re: Packages for pg_doorman |