1 |
On Tue, Oct 27, 2009 at 11:32:30AM -0700, Zac Medico wrote: |
2 |
> Brian Harring wrote: |
3 |
> > The proposal is pretty simple; if code modifies the vdb in any |
4 |
> > fashion, it needs to update the mtime on a file named |
5 |
> > '.modification_time' in the root of the vdb. |
6 |
> |
7 |
> I'd to prefer using the mtime of the /var/db/pkg directory itself, |
8 |
> since existence of a '.modification_time' file isn't going to prove |
9 |
> that an programs that don't recognize that file haven't made any |
10 |
> modifications. |
11 |
|
12 |
Grumble. Works for me. |
13 |
|
14 |
|
15 |
> We can also use the mtimes of category subdirectories, in order to |
16 |
> indicate whether a modification has occurred in any given category. |
17 |
|
18 |
Pkgcore already relies on that for old style virtuals cache. The |
19 |
pisser there is that modifications w/in a node don't result in a |
20 |
category level mtime- it certainly would be nice to have it formalized |
21 |
in some fashion so that cache regeneration could just work on the |
22 |
areas it needs to. |
23 |
|
24 |
~brian |