Hi Stewart,

stewart mackenzie <[email protected]> writes:

> On Tue, Aug 30, 2016 at 8:02 AM, Shea Levy <[email protected]> wrote:
>> globin missed the fact that the naming convention is messing up the
>> lowPrio logic, and your original PR had nothing to do with that. If rust
>> were named properly, your fix would be wrong.
>
> What is the exact naming scheme that rust should adopt?
>

That is what the other ml thread is about, to decide if indeed it should
be changed. If so I can change it, but essentially the -beta/-unstable
identifier should come *after* the version number/date, not before.

>
> I will make a pull request to fix this.
>
>> Even in this case where
>> your fix is harmless (which globin reasonably missed), it doesn't
>> actually fix your issue and in the event that the rust naming is fixed
>> it would be harmful then.
>
> Secondly, I want to test this harmful aspect. What is the expected
> harmful behaviour if I follow the steps to reproduce i.e.: correctly
> name each rust and remove lowprio?

Sure, check out this patch: http://sprunge.us/fieF

With that applied to nixpkgs, nix-env -f /path/to/nixpkgs -i rustc
--dry-run will choose rustc-1.11.0. However, if you remove the "lowPrio"
From the definition of rustcUnstable, then it will choose
rustc-1.13.0-master-g308824a.

It turns out due to the nested nature of rustUnstable and rustBeta in
nixpkgs, the lowPrio doesn't matter, as those packages aren't exposed to
nix-env by name at all. So I will remove that.

~Shea

Attachment: signature.asc
Description: PGP signature

_______________________________________________
nix-dev mailing list
[email protected]
http://lists.science.uu.nl/mailman/listinfo/nix-dev

Reply via email to