1 |
On 5 October 2010 02:55, Diego Elio Pettenò <flameeyes@×××××.com> wrote: |
2 |
> USE flags add complexity, and in real use cases there are near to no |
3 |
> good reasons at all to keep .la files around. |
4 |
|
5 |
That's why I initially suggested a variable rather than a USE flag, as |
6 |
if it was implemented in a centralised function it wouldn't add |
7 |
complexity to individual packages (no need to declare it in IUSE, for |
8 |
example). Although the USE flag approach has advantages too, as has |
9 |
been mentioned. |
10 |
|
11 |
> Also, I would like to ask again that if you're going to argue "you never |
12 |
> know who might use them", you're going to have to actually understand |
13 |
> _what_ the files are used for, _which_ software uses them, and come up |
14 |
> with a use case for them, not a vague "oh there might be a project that |
15 |
> use them". |
16 |
|
17 |
The fact is that you /don't/ know what might use them. Gentoo is a |
18 |
general-purpose system, you can't predict what people will want to do |
19 |
with it, including use software that relies on .la files from |
20 |
libraries in the tree. Obviously you can't support everything that |
21 |
anyone might conceivably dream up, but keeping packages compatible |
22 |
with their upstream state is a good place to start. |
23 |
|
24 |
> Things like "taking in account what isn't in the tree" (without actually |
25 |
> having a clue on what .la files are for) |
26 |
|
27 |
So taking into account what isn't in the tree while knowing what .la |
28 |
files are for is OK? |
29 |
|
30 |
> looking for "alternative approaches" (to do what, exactly?) |
31 |
|
32 |
Alternative approaches to fix the excessive dependency propagation |
33 |
without breaking the semantics of libtool under static linking etc. |
34 |
|
35 |
> Should we remove /usr/lib/libxml2.la? Yes. Why? Because upstream does |
36 |
> not explicitly provide it (it's a byproduct of libtool) |
37 |
|
38 |
What's the difference? |
39 |
|
40 |
> and they don'trequire/ask to rely on it (otherwise it wouldn't |
41 |
> provide pkg-config or .pc files). |
42 |
|
43 |
pkg-config is useful because it's a standardised way to check for the |
44 |
presence of other packages, including specific versions if necessary, |
45 |
and to get the necessary include and library paths to use the library |
46 |
- it's not the case that pkg-config is only supported for the sake of |
47 |
avoiding libtool. pkg-config /also/ overlaps with some functionality |
48 |
of libtool, but that's not to say that no-one must ever use the |
49 |
libtool functionality. |