1 |
On Sunday 20 March 2005 03:38, Ciaran McCreesh wrote: |
2 |
> On Sun, 20 Mar 2005 02:54:46 +0900 Jason Stubbs <jstubbs@g.o> |
3 |
> |
4 |
> wrote: |
5 |
> | With package.mask'ed packages, it would have been better to just drop |
6 |
> | them altogether. For unmasked packages, you would just replace them |
7 |
> | with skeletons that dep on the replacement package. See the still |
8 |
> | existing sys-apps/fileutils-4.1.11-r2 ebuild. |
9 |
> |
10 |
> Bleeeeuhhhhh. If I file a bug asking for MERGE functionality in updates, |
11 |
> will it get anywhere? Or is this another one of those "if portage finds |
12 |
> something it doesn't understand in a file it will traceback" situations? |
13 |
|
14 |
It'd be one of those ones that sit for months/years. It should be a rare |
15 |
situation and work required to have it done automatically far outweighs the |
16 |
benefit at this time. |
17 |
|
18 |
The first issue that comes to mind is a different record of md5/mtime - which |
19 |
to use and what if neither match? Another issue is what if there are three |
20 |
packages being merged into one but the user only has two of them installed - |
21 |
what to do? |
22 |
|
23 |
There are probably more issues, but it should be plain that it's not so |
24 |
simple. And, for the record, portage will just emit an error if a malformed |
25 |
line is found in an updates file. |
26 |
|
27 |
Regards, |
28 |
Jason Stubbs |
29 |
-- |
30 |
gentoo-dev@g.o mailing list |