Re: $ORIGIN in RPATH suppression

[pin <voidpin%protonmail.com@localhost>]

> It could be as ugly as a package that installs scripts that does the
> cross builds and patches up the rpaths, so the builds aren't done under
> wrappers, but still the recipe is checked in.

Now, we are talking :)
That's actually a smart "work around" because, if cross-compilation is feaseble, we can move forward.

> Or the wrappers listen to some environment variable and with that set
> to some magic token (which the bootstrap pkg could set) avoid the
> RPATH sanitization.
> 
> Martin

Slightly different "work around".

> I wonder if we should have a policy that such pkgs should only go in
> pkgsrc (instead of wip) if the required rust version is at least a year
> old or they are special developement pkgs (like ${something}-git where
> we also have a stable/sane ${something} pkg).
> 
> And we should file upstream bug reports if this happens and stops pkgs
> from being updated.

One year?! If we are talking Rust, I'd be happy with things building with a compiler that's 3 months old.

> It's certainly an upstream bug to require super-recent versions of
> dependencies, unless there is a really good reason and avoiding such
> would be significantly difficult. It's also a bug to require a specific
> rust compiler version, rather than a language specification. But
> upstreams that use rust tend to think this is ok, and while I'm
> philosophically in agreement that filing bugs makes sense, it seems
> unlikely to be productive.
> 
> (My general rule of thumb is that a program released on D should be ok
> with any version of a dependency that was current at any point from D-2y
> to D. And that exceptions should be relatively rare and have a good
> reason.)

Two years is asking too much. But, honestly I have experienced quite ok upstream feedback.
Some have been positive and reverted to an older version of the compiler.

For example sysutils/broot reverted back to 1.76 when we had issues a few months back.
Now, they recently moved to 1.79 and the reason is explained in the changelog,
"rustc minimal version changed from 1.76 to 1.79, which allows better performing image rendering"

Unfortunately, not all are as responsive.

> I have not gotten the impression that people who work on updating rust
> feel that they are being hassled by things pending in wip.

Far from hassel anyone. I'm happy running the most recent Rust compiler.
But, as mentioned, I would like to focus on a solution, and now we are at least talking "solutions".

Thanks!