Re: Do pretests reach end users?

[Phillip Lord]

Dmitry Alexandrov <dag@gnui.org> writes:

> Phillip Lord <phillip.lord@russet.org.uk> wrote:
>>> https://alpha.gnu.org/gnu/emacs/pretest/windows/emacs-27/emacs-27.0.91-x86_64-installer.exe
>>
>> Yes, I release this (in a rather erratic fashion) as well as Emacs-28 
>> snapshots
>
> My gratitude for that (though I never happened to use them).

No worries. I don't happen to use them either!


>> To be honest, I suspect GNU/Linux gets more people testing it,
>> because the users of that platform are more likely to test
>> things. It's also relatively easy to build there -- once you have it
>> set up "git pull;make -j" and run from in source rather than
>> installing and you are done.
>
> Sure.  Thatʼs exactly why I supposed, that hardly anyone use the
> pretests as @eliz@gnu.org explained them: people are either on
> whatever their distro ships or build from master.


I wouldn't be so sure about that actually. I build form the release
branch most of the time; I am currently using a Emacs-27 build as my
main one. I'll move to Emacs-28 at some point during the Emacs-27
release cycle.




>>>> So I don't see how all this could help making a release faster.
>>>
>>> It can help persuade people in charge (such as you :-) to take
>>> advantage of the control over GNU distributors they have — and
>>> shorten the release cycle.
>>
>> I think this is a secondary argument. It would be good to have
>> pre-releases (and snapshots) available to install because it is a
>> good thing in itself.
>
> Indeed.  There is zero hope for major distros to ships nightly
> snapshots, though.  So until Emacs cycle is completely revamped only
> pretests and release candidates are in question.

Over all, I think that this would not be a good thing. You want Emacs to
be relatively stable out there, and only people with an interest to
install the pre-release or snapshots. Ubuntu has PPAs for exactly this
purpose.

The counter argument is MELPA, of course. The statistics show that most
MELPA users are quite happy and don't both with MELPA stable, but move
along on the bleeding edge of what ever packages they are using.



>> If it shortens the release cycle that is a bonus.
>
> Shorting the release cycle is a _mean_ to achieve that — that was my point.
>
> To recap: maintainers are unanimously unwilling to package a piece of
> software labelled ‘pretest’, that should be downloaded from a server
> named ‘alpha’.  You can either (a) persuade all of them, that this is
> not what it means, but actually a pretty stable thing for the year
> 2020, or (b) simply stop calling it ‘pretest’: the final result is
> identical.


I can see this. If we just had a shorter release cycle (especially a
minor release cycle), then a less stable Emacs might be less a worry,
since a new release would be just around the corner. Emacs bugs can be
annoying, but they generally do not break other things.

Phil