1 |
On 09/10/14 13:16, Rick "Zero_Chaos" Farina wrote: |
2 |
> On 09/09/2014 02:38 PM, Brian Dolbec wrote: |
3 |
>> On Mon, 1 Sep 2014 22:05:45 -0700 |
4 |
>> Brian Dolbec <dolsen@g.o> wrote: |
5 |
>> |
6 |
>>> This code had portage bin path hard coded. That path needed to be |
7 |
>>> changed for a new portage ebuild and install system. |
8 |
>>> After testing the origianl code and comparing it with some updated |
9 |
>>> code supplied by Douglas Freed. It turned out both code chunks |
10 |
>>> resulted in nothing being cleaned. |
11 |
>>> |
12 |
>>> Tested and confirmed by zero_chaos. |
13 |
>> |
14 |
>> I have gone over things more and tested the new find command. It does |
15 |
>> work on my host system. However, the question remains... DOES this |
16 |
>> particular cleaning operation NEED to be performed? |
17 |
>> |
18 |
>> With current the tree snapshot for my testing 20140829. It does not |
19 |
>> find anything to clean. BUT, will that remain the same in the future |
20 |
>> as pkgs are bumped? |
21 |
>> |
22 |
>> For safety, I'd be inclined to keep the find command (v1 of the |
23 |
>> patch) and clean any it does find just in case. |
24 |
>> |
25 |
> |
26 |
> If we truly need to remove these files, I don't think catalyst was ever |
27 |
> the place to do this. We have USE=static and USE=static-libs for a |
28 |
> reason. Randomly removing files from the filesystem was a hack then, |
29 |
> and if we are cleaning this up let's just remove it. If I want to build |
30 |
> things with USE=static or USE=static-libs then catalyst shouldn't be |
31 |
> pointlessly crippling my builds. |
32 |
> |
33 |
> v2 means one less horrible hack in catalyst, and one at a time is the |
34 |
> only way will remove all the horror. |
35 |
> |
36 |
> -Zero |
37 |
> |
38 |
|
39 |
I just did some testing here. You have to be very careful not to remove |
40 |
the wrong .a's else you break the toolchain. So I agree, remove the |
41 |
.a's in the ebuilds switched on USE=statlic-libs, and not in catalyst. |
42 |
The biggest .a's in stage1 are libpythonXX.a which are 3 to 4 MB, but |
43 |
the others are like a few hundred K's at most. The gain is a smaller |
44 |
stage, but the downside is headaches making sure we don't break something. |
45 |
|
46 |
So, to slim down stage1, see if you can get USE=static-libs in python's |
47 |
ebuilds and build USE=-static-libs. That itself will reduce that stage |
48 |
by about 8 MBs which is more than all the "expendable" .a's put together. |
49 |
|
50 |
-- |
51 |
Anthony G. Basile, Ph. D. |
52 |
Chair of Information Technology |
53 |
D'Youville College |
54 |
Buffalo, NY 14201 |
55 |
(716) 829-8197 |