Clarity on rust failures (NetBSD)

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

(This is all only about NetBSD.)

I've heard some people say that rust builds reliably now, and some say
that it doesn't.  I don't find many open PRs.  I would like to get to a
situation where we jointly understand the situation and have PRs or at
least comments in the Makefile about what's not ok, so that we can
document workarounds or actually fix things.

Things I think we all agree on (but tell me if not):

  Rust needs entropy.  (Documented in /quarterly as a required workaround.)

  There was a serious linker bug fixed in NetBSD currrent in April and
  pulled up to 9 in May.  This has not been pulled up to NetBSD 8.

  People building rust (or anything) on current should be using current
  current.  (We don't support old current at all.)

  People buidling rust on 9 should use netbsd-9 after the pullup.

  lang/rust semi-reliably fails to build on NetBSD-9 (pre pullup) with
  the included MKAE_JOBS_SAFE=no.  With MAKE_JOBS=1, it seems to work.
  No PR, but a lang/rust/Makefile comment.  We agree that this is not
  understood, but it's also an older version and thus water under the
  bridge.  If it applies to pkgsrc-current rust, then it matters.

Things we do not agree on:

  Whether lang/rust builds reliably on NetBSD-current, either with
  MAKE_JOBS=1 or elevated.

  Whether lang/rust builds reliably on NetBSD-9 (post pullup), either
  with MAKE_JOBS=1 or elevated.

  Whether lang/rust builds reliably on NetBSD-9 (pre pullup) with the
  included MAKE_JOBS_SAFE=no.

  Whether lang/rust builds reliably on NetBSD-8 with the included MAKE_JOBS_SAFE=no.

Reliable in all cases is meant to mean in all environments, including
Xen, when accomodating randomness as /quarterly says, and typing 'make
package' once results in a package, with no need to do anything.

I throw out, without conviction and for contradiction, that the above
four situations all build reliably.  If you have evidence or rationale
to the contrary, please speak up and let's get a PR entered.

(I will say that I am not agitating to move away from rust-bin for amd64
for NetBSD 8 and 9 for this branch.  We've had a lot of trouble and I
think it's more important to have a good build this time.  This is about
understanding where we are.)