Re: Testing rust 1.56.0

[Havard Eidnes <he%NetBSD.org@localhost>]

>> > Surely llvm should be updated rather than requiring
>> > bundled llvm simply because it's a newer version?
...
>> I feel that adding a rule that one can't do the 2nd thing is just going
>> to delay rust and delay other things, or lead to a rushed update and
>> breakage
>
> pkgsrc really should have consistent policies on bundled libraries.
> Most other package managers do.

I'm not on principle opposed to an update of llvm.  However,
forcing me to become the llvm package maintainer isn't going to
work for several reasons.  Those may be available time, trampling
in on other people's circles of responsibility (perceived or
real), and ability to verify that the update is "safe" (that goes
to "time" above, I expect), and possibly own abilities to deal
with fallout.

It will also potentically cause certain rust updates (as this
one) to be severely and needlessly delayed.

That said, I've started testing rust built without the rust-llvm
option.

>> But for now, I don't entirely understand what is going on.  The recent
>> discussion seems to be only about macOS, but it also seems that the
>> option is turned on for all systems with HAVE_LLVM in mk.conf, without a
>> bug reference.

That's probably because it is an "unsettled problem".

In my case, on NetBSD/amd64 current-ish, rust builds without the
rust-llvm option set, and I can build librsvg and firefox with
the resulting compiler.  I didn't have immediate success in the
"run" phase, will test more.

Regards,

- Håvard