Re: rust on 32-bit arm

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

Havard Eidnes <he%NetBSD.org@localhost> writes:

> I finally upgraded my test armv7hf VM from 9.0_STABLE to
> 9.4_STABLE, and wouldn't you know, the build of dua-cli using the
> cross-built rust compiler 1.83.0 now works where it immediately
> crashed earlier.
>
> I have a suspcion that the fixes to ld.elf_so related to thread
> local storage may be a reason for this, but that is only a
> suspcion and has not been verified.

That's great news.  While supporting the netbsd-9 branch is good, I
don't think it makes sense to expend much effort on 9.0.  Upgrades along
branches are easy and in my experience very safe, and indeed bugs are
fixed.

> I'm going to give my armv6 VM a similar treatment, and suspect
> it'll also start working again.
>
> Self-hosting the rust compiler on the 32-bit arm variants still
> appears out of reach, due to bloat in the rust compiler.

(Probably the following is related; I hadn't gotten to writing about it.)

I recently had an issue on

  2024Q$
  netbsd-9 (recent), earmv7hf-el, RPI3

where when building some package it tried to install from lang/rust176,
even though rust-bin-1.81.0 was installed.

This is due to lang/rust/platform.mk, which I have modified to comment
out the exception:

  .if ${MACHINE_PLATFORM:MNetBSD-*-earm*} && ${OPSYS_VERSION} < 100000
  #RUST_DIR?=     ../../lang/rust176
  .endif
  RUST_DIR?=      ../../lang/rust

Probably I had an older rust-bin and pkgrr happily replaced it, and I
was unaware that it wasn't supposed to work.   And it did because my
netbsd-9 is up to date.

I'm not sure how we should handle this.  I lean to just documenting that
if you are running pkgsrc on 9 you should be at least at 9.4 and not
worrying further.