[Nix-dev] [Nix-commits] SVN commit: nix - r30127 -innixos/trunk/modules:configinstaller/cd-dvdinstaller/generations-dirinstaller/toolsinstaller/tools/nixos-deploy-networkmiscsecurity services/miscservices/monito...

[Michael Raskin]

<20111030151958.AAF87CC149 at mx4.tudelft.nl>
<E1RKXTS-0003SS-00.7c6f434c-mail-ru at smtp14.mail.ru>
<E1RKXk5-0000wm-00.7c6f434c-mail-ru at smtp8.mail.ru>
<4EAD7529.6030300 at shealevy.com>)
Mime-Version: 1.0
Content-type: text/plain; charset="UTF-8"

>Shea: I feel your pain. That's one of the reasons why I forked - cause
>people revert patches (which feel randomly to authors) without even
>trying to discuss an issue enough so that its impossible to fix the
>issue causing the revert. This may lead to depression. Happened to me 
>trying to implement the "num-cores" patch.

I guess that to stop the insanity like "no reaction is no OK" we do need
to make allowances for "heavy problems merit revert". Having to reboot a
computer is not something we want to impose on too many people... 

>Anyway - why do people feel that strong about trunk? Is it because they
>have important systems which must run all day ? Such as "spawning
>emergency systems on amazon"? Then it should be them maintaining a
>"stable" branch - because that's the only way to ensure that their
>systems keep working - them depending on "trunk" would be insane anyway.

Two things.

1) Trunk is what Hydra builds and we don't want to get all the working
binary packages garbage-collected. 

2) Many people think that building trunk is safe and you only need to
start thinking after the build stops (succesfully or not). Me too, to be
honest. Maybe we need a vote to see if it is a common idea. If I 
perceive it to be a common idea and I see that building trunk crashes my
computer, I perceive making trunk safe as a net-positive action.

>And such a "stable" branch can be defined "stable" by "all test cases
>pass". That would make lot's of sense - if something breaks another test
>case should eventually be added.
>
>BTW: Should we revert all patches on the git kernel tree because BTRFS
>broke (causing the system to hang and become slower ?).

BtrFS is a feature you can avoid with rmmod. nix-instantiate-ing NixOS
trunk is a prerequisite to using any part of NixOS.