| 1 |
On 7 August 2012 19:52, Michał Górny <mgorny@g.o> wrote: |
| 2 |
> Hello, |
| 3 |
> |
| 4 |
> Right now, every time a bigger bunch of stuff (installed by various |
| 5 |
> packages) needs to be moved around the filesystem, we have a lot of |
| 6 |
> work to handle it somehow. And finally, users end up having to either |
| 7 |
> rebuild a lot of packages to get the files in the new locations, or |
| 8 |
> we do ugly things to move those files for them. |
| 9 |
> |
| 10 |
> I believe we should consider implementing something simpler. Thus, |
| 11 |
> I propose introducing the following new command to profiles/updates: |
| 12 |
> |
| 13 |
> fsmove <old-location> <new-location> |
| 14 |
> |
| 15 |
> which -- at the moment of update -- will cause all PM-owned files |
| 16 |
> in the old-location to be moved to the new one (recursively), updating |
| 17 |
> the vdb as necessary. |
| 18 |
> |
| 19 |
> What remains to be solved/decided: |
| 20 |
> |
| 21 |
> 1. How to treat non-owned files? (leave them there, refuse to proceed |
| 22 |
> with updates?) |
| 23 |
> |
| 24 |
> 2. How to handle relevant required updates? (packages which |
| 25 |
> actually *have* to be updated before moving files) |
| 26 |
> |
| 27 |
|
| 28 |
I suggest, that due to the volatility of such actions, a user should |
| 29 |
have to approve each bulk move before it is done, to avoid breaking |
| 30 |
things. |
| 31 |
|
| 32 |
Sort of like etc-update: |
| 33 |
|
| 34 |
An update file is added to the repository |
| 35 |
PMS's detect the new update, and detect the update has not been |
| 36 |
performed, and starts notifying the user that pending updates are |
| 37 |
needed. |
| 38 |
User performs action(s) when ready via some client ( eselect ? PMS |
| 39 |
specific? ~~ ) |
| 40 |
|
| 41 |
Additionally, move batches could have annotations preceding them |
| 42 |
indicating either instructional ( einfo ) or automated ( like emerge |
| 43 |
--config ) to handle things like "you'll want to close postgresql |
| 44 |
before you do this or you'll get database corruption" . |
| 45 |
|
| 46 |
I guess what I'm saying basically, is a hybrid concept, sort-of like |
| 47 |
eselect news , except with executable behaviour attached. |
| 48 |
|
| 49 |
|
| 50 |
At least, thats my 2c. |
| 51 |
|
| 52 |
-- |
| 53 |
Kent |
| 54 |
|
| 55 |
perl -e "print substr( \"edrgmaM SPA NOcomil.ic\\@tfrken\", \$_ * 3, |
| 56 |
3 ) for ( 9,8,0,7,1,6,5,4,3,2 );" |
| 57 |
|
| 58 |
http://kent-fredric.fox.geek.nz |
Archives