tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Rust 1.46.0 update



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

> The NetBSD/powerpc 8.0 bootstrap does not work on 9.0, due to
> some C++ ABI markings which are not portable across those two
> (I've muttered here(?) about that earlier).  It doesn't look like
> this is a universal problem, though.  Therefore I have a local
> 9.0 bootstrap kit for powerpc.  However, the tarballs produced by
> the rust build are not tagged with OS version, so it is mildly
> annoying to have to rename the kits after they've been built.
>
> I'll admit that I've not been as thorough as I perhaps should
> have been in testing the other 8.0 bootstrap kits on 9.0 (wasn't
> there some fallout on i386?).

The issue is that while generally binaries linked for 8 against *system
libs* work fine on 9, and current, with compat80, that bootstrap kits on
8 linked against openssl, which is from pkgsrc, because 8 base openssl
is old, and that package is generally not installed on 9.

>> Are you able to a construct a rust binary package that could be used to
>> e.g. build rsvg on earmv7hf-el?   Something that could be part of
>> rust-bin?  (I am assuming upstream does not build for NetBSD/earmv7.)
>
> The rust-bin package only supports official rust binaries.  I'm
> not up for changing that.  That said, the cross-built bootstrap
> kits appear to contain most of what a binary distribution would
> contain, so it is not entirely inconceivable that it could be
> used.

I don't care much about the officialness, and don't see the harm of
rust-bin having a TNF-built binary on systems where upstream doesn't
publish one.  To me the big deal is that it's wrapping a binary.

But, I also don't care if we have a rust-bin2 or something, that is
selected for systems that can't build from source and for which we have
a TNF-built binary, that's ok.

Attachment: signature.asc
Description: PGP signature



Home | Main Index | Thread Index | Old Index