1 |
On Thursday 26 January 2006 19:53, Mike Frysinger wrote: |
2 |
> On Thursday 26 January 2006 11:06, Paul de Vrieze wrote: |
3 |
> > On Thursday 26 January 2006 14:51, Mike Frysinger wrote: |
4 |
> > > On Thursday 26 January 2006 05:43, Paul de Vrieze wrote: |
5 |
> > > > Another candidate would be the strip binary which might be called |
6 |
> > > > by certain makefiles instead of being portage controlled. |
7 |
> > > |
8 |
> > > packages should never strip, only portage should |
9 |
> > |
10 |
> > ebuilds don't, some makefiles do. |
11 |
> |
12 |
> exactly, thus i said "packages" and not "ebuilds" |
13 |
|
14 |
Btw. I do agree with you that they shouldn't. |
15 |
|
16 |
> |
17 |
> > Sometimes when calling the strip option |
18 |
> > of install. A strip wrapper prevents this broken behaviour once and for |
19 |
> > all. It could even be written to show a big fat warning. |
20 |
> |
21 |
> i know ... it isnt uncommon to see like `install -s` or `$(STRIP)` in |
22 |
> packages and those need to be removed |
23 |
> |
24 |
> while this is a neat idea (catching those people who do `install -s`), i'm |
25 |
> not sure it'd work as there isnt a clean way to detect whether it's the |
26 |
> package calling `strip` or the ebuild/portage ... you could try passing |
27 |
> info via an env var, but that's no fun :) |
28 |
|
29 |
Well, portage uses prepstrip to do stripping. As such this prepstrip script |
30 |
could take care not to use the wrong strip binary. Shouldn't be hard to do |
31 |
even without hardcoding the path to the strip binary. |
32 |
|
33 |
For ebuilds calling strip, I see no reason why they would. If at some point it |
34 |
is found necessary, it would be easy to have an estrip command to do this. |
35 |
|
36 |
Paul |
37 |
|
38 |
-- |
39 |
Paul de Vrieze |
40 |
Gentoo Developer |
41 |
Mail: pauldv@g.o |
42 |
Homepage: http://www.devrieze.net |