Re: [m-users.] Installing MMC over MMC

Admittedly so, Julien.

When this happens (it does not too often), then there is no way of 
building mmc until a fixed ROTD is released by the team. As the Docker 
algo fetches fresh ROTDs on a rolling basis as they are released, if a 
crash occurs in some GitHub action run, it will automatically be fixed 
within a couple of days. This is what the backlog of GH actions shows 
anyway.

So that empirically, released images/packages are always coherent as 
only successful builds get released by the GH scripts (then to DockerHub).

F.

Le 12/11/2022 à 14:13, Julien Fischer a écrit :
>
>
> On Sat, 12 Nov 2022, fabrice nicol wrote:
>
>> Yes it works, this is the algo used by the everyday Docker build 
>> released on Github.
>>
>> It contains two builds: one fresh ROTD as a bootrap and a repo 
>> version built at HEAD with the bootstrap ROTD build.
>>
>> https://github.com/fabnicol/gentoo4mercury/releases
>>
>> (note: there are Ubuntu and Alpine releases too, but these have had 
>> distribution-related issues for a few days. Nothing related with 
>> Mercury source code).
>
> That seems problematic in the case where we delete or rename a standard
> library module between the ROTD and the repo head; you may end up with
> interface files hanging about for a library module that no longer
> exists.
>
> (Voker's use case is fine since he is just reinstalling the same version
> and will necessarily have the same set of library modules.)
>
> Julien.
_______________________________________________
users mailing list
[email protected]
https://lists.mercurylang.org/listinfo/users