1 |
On Thu, 4 Nov 2004, Graham Murray wrote: |
2 |
|
3 |
> Ciaran McCreesh <ciaranm@g.o> writes: |
4 |
> |
5 |
>> A user installing any one of these packages would expect to be given |
6 |
>> convenience symlinks which would allow them to access the packages |
7 |
>> via the traditional names. Suddenly we have a problem -- what if a |
8 |
>> user has more than one of these packages installed? |
9 |
> |
10 |
> Would it be possible for the emerge to only install the symlink if it |
11 |
> does not already exist or if it is 'dangling'? Then allow the user to |
12 |
> change the symlink if (s)he wishes to change the program called using |
13 |
> the traditional names. |
14 |
|
15 |
I think that this is the most sensible idea mentioned in this thread. |
16 |
Admittedly, shortly after this message, the Grand System crowd came out |
17 |
in force. |
18 |
|
19 |
Personally, I'd augment the above with, 'Of course, the original package |
20 |
(in this case, vi) would remove any symlinks it needed to remove in |
21 |
order to install its binaries - this is the principle of "the actual |
22 |
product trumps the emulator."', and then feel this is sufficient |
23 |
direction to the coders. |
24 |
|
25 |
Yeah, I know, this is Gentoo, so regardless of how complicated the spec |
26 |
is, they're going to make a Grand System. But it'd be nice if the spec |
27 |
at least were simple. This gives people more opportunity to fix |
28 |
senseless complication later on by distinguishing between what was |
29 |
perceived to be needed, and what was added in simply because it seemed |
30 |
cool. |
31 |
|
32 |
Personally, I'd implement the above with a wrapper around installing |
33 |
links and another around installing files; the former checks to see if |
34 |
os.stat(pathname) exists and is not a symlink, the latter checks to see |
35 |
if os.lstat(pathname) exists and is not a symlink. |
36 |
|
37 |
Ed |
38 |
|
39 |
-- |
40 |
gentoo-dev@g.o mailing list |