[Nix-dev] Regarding the unstable channel

[Joel Moberg]

On Mon, Dec 14, 2015 at 2:53 PM, Eric Sagnes <eric.sagnes at gmail.com> wrote:

> Hi,
>
> I am using the unstable channel and recently encoutered some issues to
> update
> the system with `nixos-rebuild`, to list a few:
>
> - flashplayer source not found (PR #11675)
> - udevil not building (PR #11702)
> - signing-party source not found (PR #11703)
>
> All these issues have in common that the binary substitute was not
> available
> and they were triggered during source building.
> As I understand it, a channel is updated when *all* the packages have been
> built and the binary cache is available [1].
>

I think this sentence is tricky:

The NixOS channels are updated automatically from NixOS’s Git repository
after certain tests have passed and all packages have been built.

I believe this is the case, someone who knows more please correct if it's
wrong:
Binaries will be published to the cache channel when a set of test passes.
Hydra will checkout a certain git revision, try to build as many packages
as possible, then run tests and when all is green copy binaries (if needed)
to the channel.



> I totally understand that issues are expected when using the something
> labelled
> unstable, but it might confuse some less technical users, or worse give a
> bad
> impression to people that are trying NixOS.
>
> This problem has no severe consequences as the system stays in a stable
> state, but it is a little unpleasing when trying to update the system with
> a
> a local dev branch rebased on the current unstable and get update failures.
> (maybe this is more a development workflow problem?)
>
> I am not familiar with hydra so I don't know what is possible to improve
> the situation.
> But for example, could it be possible to update the unstable channel when
> none
> or only a few jobs fail?
>

If all packages were tested on unstable it probably would mean that updates
of unstable would come less regularly and I think that is unwanted. But if
the packages you needed are well maintained then it's probably no harm
adding a test for them.


> Also, I have no idea if it is technically possible, but having an automatic
> weekly digest on the ML with some hydra evaluations statistics and a list
> of
> "newly failing jobs" could be a way to bring attention on potential
> issues.


Hydra have this functionality, if you are listed as a maintainer to some
package you will get notified when Hydra fail or succeed to build it.


>
> Cheers,
> Eric
>
> [1] http://nixos.org/nixos/manual/index.html#sec-upgrading
> --
> Eric Sagnes
> サニエ エリック
> _______________________________________________
> nix-dev mailing list
> nix-dev at lists.science.uu.nl
> http://lists.science.uu.nl/mailman/listinfo/nix-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.science.uu.nl/pipermail/nix-dev/attachments/20151214/4b5c8f2f/attachment-0001.html