Re: [gentoo-dev] [PATCH 1/2] profiles.desc: Reduce the status of most exp profiles to dev

[Ulrich Mueller]

>>>>> On Thu, 11 Jan 2018, Michał Górny wrote:

>> So you're *promoting* the ones considered to be broken from "exp"
>> to "dev"?

> Please point me to one bit of documentation that says that 'dev' is
> better than 'exp' because I haven't been able to find any. Well,
> except the fact that PMS lists 'stable' and 'dev' as example
> statuses, and doesn't list 'exp' at all.

Initially, "exp" profiles were introduced here:
https://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/profiles/profiles.desc?revision=1.149&view=markup
with the commit message "*Drop* all Prefix profiles to experimental
state for the time being" (my emphasis). From this it seems to be
clear that "exp" was intended to be less stable than "dev".

Later, there was this council decision:
https://projects.gentoo.org/council/meeting-logs/20140225-summary.txt

- Vote: Minor archs with inconsistent stable keywording should be
  marked "exp".
  Accepted unanimously.

Subsequently, these arches were dropped from "dev" to "exp" status:
https://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/profiles/profiles.desc?revision=1.238&view=markup

> That said, repoman currently treats failures in exp as errors, while
> in dev as warnings. This makes me believe 'exp' was considered
> higher.

> Furthermore, the switch for -e is boolean, while -d is unary.

How can syntax of an option in one of our tools be an argument here?
Also, isn't that simply inconsistent? If anything, they should both
changed to be consistent, e.g., changed into simple switches -d and -e
without option argument.

> So we can enable testing exp by default without having to change
> usage of repoman.

So repoman will become even slower, by about a factor of about three
(assuming that time is linear with the number of profiles)? That's not
acceptable, because it will impede on maintainers' workflow.

Ulrich