Re: rust, rust-bin, bulk builds

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

Tobias Nygren <tnn%NetBSD.org@localhost> writes:

> Isn't this just a matter of pulling up all ld.elf_so changes to -9?
> Does pkgsrc have to bend over backwards to work around a src bug?

[details in other msg]

I can see your point, but my perception of reality is that we have had a
lot of pain with rust building over the last few months.  If we can end
up in a situation where rust simply builds every time somebody tries, in
all sane configurations, and we believe that this will continue across
updates, that would be great.

> I have now also realized that RUST_TYPE defaults to bin, even on
> -current. I object to this. Why does it not use the same logic as
> MAKE_JOBS_SAFE?

I interpret your comment as:

  you don't object to the existence of rust-bin

  on NetBSD-current, you think rust should default to lang/rust.

> rust-bin is not without it's own set of issues and headaches. For
> example it requires a COMPAT_80 kernel, pulls in the whole of
> emulators/compat80 and still it does not work correctly on HEAD-llvm:
>
> => Checking for missing run-time search paths in rust-1.42.0nb1
> ERROR: bin/cargo: missing library: libgcc_s.so.1
> ERROR: bin/cargo-clippy: missing library: libgcc_s.so.1
> ...

I guess there is 9-llvm too to check.

> If I wanted pkgbin instead of pkgsrc I would just install Debian
> instead.

:-) I have noticed how at least some people using some Other Systems
perceive that building things from source is too hard to actually do.