Re: Rust on NetBSD 9

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

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

>> This is the same issue as with 1.85, which was discussed
>> previously (with the unhelpfully continually changing email
>> subjects), starting with:
>>
>> https://mail-index.netbsd.org/pkgsrc-users/2025/05/20/msg041603.html
>>
>> I also filed a PR about this, noting that, yes, using
>> rust-internal-llvm also allowed the build to succeed for
>> me. (That being PR pkg/59435.)
>
> I interpreted the end of that thread so that Greg thinks it's
> fine to force a given GCC version for a given platform and not
> necessarily do it by insisting that the code needs c++17 (if
> that's not really the case) as long as a comment is left
> explaining why this is done, and that there probably is an
> un-diagnosed bug lurking back there with the previous gcc which
> this is a workaround for.

Not sure how this relates to the current issues.

I am in general uncomfortable with any GCC_REQD setting.   Certainly,
one should set c++NN based on what the package needs.
Anything that is nonstraightforward needs a comment.

And then, bugs should be filed someplace somewhere if
 
  - the package actually uses c++NN but does not document it

  - the package fails with a gcc that our compiler.mk says is ok,
    which is perhaps a gcc bug, perhaps a bug in compiler.mk, and
    perhaps a bug in the upstream code that it has improper code

> But apparently nobody picked up the ball and carried it over the
> goal-line.  Want me to do this?  For NetBSD/x86_64 9.x only?
> (That's the only verified problem, for now; my build hosts are
> working on sparc64 and i386.)

What problem and what fix?  You've been writing about rust not building
with pkgsrc llvm 19, but I'm not clear on what that has to do with gcc.