| 1 |
On Thu, Mar 26, 2009 at 03:28, Jeremy Olexa <darkside@g.o> wrote: |
| 2 |
> Its not a bad idea, per-say. I already keep /etc in git and I'm not |
| 3 |
> convinced that you can make this a whole project. It would probably take (a |
| 4 |
> guess) 2 hours to add a hook to portage that runs something after emerge |
| 5 |
> exits. Similar to /etc/portage/postsync.d/ |
| 6 |
|
| 7 |
Yes I did that once, it sucked. So while it is possible to hack this |
| 8 |
together in an afternoon properly doing it will take more time. For |
| 9 |
one thing which I think is significant to make confsink a reliable and |
| 10 |
pleasant experience to the user I do not want to use hooks. |
| 11 |
|
| 12 |
Having portage create the ._cfsxxx files only to remove them later is |
| 13 |
just unclean. It also makes it impossible to check if the relevant |
| 14 |
branches on the livesystem are clean and to commit to the vanilla |
| 15 |
branch at the right time. It would actually have to be reconstructed |
| 16 |
from the merged files along with the ._cfsxxx files. |
| 17 |
|
| 18 |
Also just using some post hook it is not possible to check that all |
| 19 |
relevant branches are clean at a reasonable time (after building the |
| 20 |
list of packages to be merged but before doing so). There are many |
| 21 |
similar problems to be expected. |
| 22 |
|
| 23 |
Generally I want it to be done right. Something that can actually be |
| 24 |
integrated into portage proper. Doing so will require making it an |
| 25 |
optional feature I am afraid. |
| 26 |
|
| 27 |
Well, as I have pointed out I would like to design an interface that |
| 28 |
is carefully designed and pleases both the expert and casual user. |
| 29 |
Just throwing together some script might not be that much work. But |
| 30 |
getting the interface right is some work. Modeling it after git of |
| 31 |
course as an obvious choice as well as makes this easier. |
| 32 |
|
| 33 |
Not only the frontend needs careful design but also the backend isn't |
| 34 |
as easily done right as it might seam. Just some random thing that has |
| 35 |
to be taken care of: How to merge packages/branches in different |
| 36 |
slots? |
| 37 |
|
| 38 |
But I do understand your concerns and indeed if one stops here it |
| 39 |
might not be enough to fill the summer with code. The additional |
| 40 |
non-essential things that I suggested are not just time fillers. To an |
| 41 |
Emacs user an Emacs mode always is an improvement over anything else |
| 42 |
for example :-) |
| 43 |
|
| 44 |
Any of the things mentioned might not take long to to implement on |
| 45 |
it's own, but I believe they will sum up. |
| 46 |
|
| 47 |
> Then after that happens, what's next? Another etc-tool? |
| 48 |
|
| 49 |
Yes, but a nice one :-) Well actually more like etc-update + equery + ... |
| 50 |
|
| 51 |
More about this tomorrow, I really have to go to bed now. |
| 52 |
|
| 53 |
-- Jonas |
Archives