1 |
On Thu, 1 Jun 2017 09:38:01 -0400 |
2 |
"Walter Dnes" <waltdnes@××××××××.org> wrote: |
3 |
|
4 |
> As mentioned elsewhere, what happens when two or three other people |
5 |
> do their own forks? Plugin 1 works with vim A and B but not C or D. |
6 |
> Plugin 2 works with vim A and C and D but not B. The number of |
7 |
> directories could potentially be 2^N where N is the number of forks. |
8 |
> And who's going to do the necessary testing across multiple versions? |
9 |
> And remember that each minor version bump of any of the forks could |
10 |
> render another fork's plugin incompatable. This is a classic "moving |
11 |
> target". The only way that works is to have each fork look after their |
12 |
> own copies of plugins. |
13 |
|
14 |
If and when that happens: |
15 |
|
16 |
1. Packages that are available on only one vim still install to the one dir |
17 |
2. Packages that are available on >1 install to a common dir |
18 |
3. Vim runtimes *don't* all use the common dir, but only their own |
19 |
4. Packages that support multiple vims get symlinked into their respective |
20 |
vim paths at install time. |
21 |
|
22 |
"4" could be done in a PYTHON_TARGETS sort of way, but its obvious to see why |
23 |
I'd say "lets not do that unless we absolutely have to". |
24 |
|
25 |
"4" could alternatively be implemented by creating a meta-file that enumerates |
26 |
a list of source files to symlink, and the list of vim implementations that are |
27 |
known to be supported by it, and then a tool can fix up the difference in pkg_postinst |
28 |
or on-demand. |