Re: rust-bin fails on netbsd-9 i386

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

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

>> => Checking for missing run-time search paths in rust-bin-1.70.0
>> ERROR: lib/librustc_driver-e6814d6038053cfe.so: missing library: libstdc++.so.7
>> ERROR: lib/librustc_driver-e6814d6038053cfe.so: missing library: libLLVM-15.so
>> *** Error code 1
>>
>> Stop.
>> make[1]: stopped in /nfs/pkgsrc-current/lang/rust-bin
>> *** Error code 1
>>
>> Stop.
>> make: stopped in /usr/pkgsrc/lang/rust-bin
>>
>> my pkgsrc mk.conf is pretty vanilla w.r.t this.   I have std++.so.8 in
>> compat80, but no .7.
>
> Hmm, this appears to be a result of
>
> GCC_REQD+=      7
>
> in the rust package, and a bootstrap built on netbsd-8.  Thus,
> gcc7 from pkgsrc is used, this can be seen in "readelf" output
> from e.g. cargo:
>
>  0x0000000f (RPATH)                      Library rpath: [/usr/pkg/gcc7/i486--netbsdelf/lib/.:/usr/pkg/lib]
>
> gcc7 from pkgsrc uses libstdc++.so.7.24.
>
> Normally as I recall I have a habit of using the cross-built bits
> (from a NetBSD/amd64 10.0-ish system), which doesn't end up with
> a dynamic dependency on LLVM or gcc from pkgsrc.  I'll admit that
> I can't off-hand recall why I did it that way this time around...

How are we going to to fix it?   We should fix it and pullup to the
branch before applying any other rust changes.

Is it feasible to have a spearate bootstrap for 8 and for 9/10, and
would the built-on-9 version be ok?