10.11 64bit - gcc14 not working

[Ryan Carsten Schmidt]

On Feb 24, 2025, at 02:56, Riccardo Mottola wrote:
> 
> Joshua Root wrote:
>> 
>> Either some port has +universal as a default variant, or doesn't support x86_64 and builds as i386, thus causing its dependencies to be universal.
> 
> That is what I suspect (or user error, I activated universal during some test manually).
> Such default would be different from the build server though, I so I guess more it is an issue or combination of settings on my side

Can you provide evidence for this claim? When MacPorts is running on the build machines, it has no idea it is running in a build machine. It is doing exactly what it would do on a user's machine. 


>> This will do the same but only for recursive dependents of openssl3:
>> port -v installed rdependentof:openssl3 | grep -F i386
>   openssl @3_21+universal requested_variants='+universal' platform='darwin 15' archs='i386 x86_64' date='2024-11-06T10:45:41+0100'
>   openssl @3_22+universal (active) requested_variants='+universal' platform='darwin 15' archs='i386 x86_64' date='2025-02-19T09:57:59+0100'
>   python310 @3.10.15_0+lto+optimizations+universal requested_variants='+universal' platform='darwin 15' archs='i386 x86_64' date='2024-11-06T10:46:17+0100'
>   python310 @3.10.16_0+lto+optimizations+universal (active) requested_variants='+universal' platform='darwin 15' archs='i386 x86_64' date='2025-01-27T01:31:31+0100'
> 
> that looks like the issue is python itself that is requested universal?

Yes, that output does show that you have requested the universal variant of python310. 

> how can try going back to x86_64 only for all packages, or at least, all packages that don't need to be universal?

If you do not actually want python310 installed with the universal variant, then install it without that variant.

There is no built-in command to reinstall all ports without the universal variant. You can use a shell command pipeline to get the list of ports that are installed with the universal variant and reinstall them without that variant. I've probably posted such a command pipeline to this list or the other one many years ago, although that probably predated the tracking of requested variants and thus probably will not correctly preserve them.