1 |
Volker Armin Hemmann wrote: |
2 |
> On Montag, 5. Mai 2008, Wolf Canis wrote: |
3 |
> |
4 |
>> Volker Armin Hemmann wrote: |
5 |
>> |
6 |
>>> On Montag, 5. Mai 2008, Wolf Canis wrote: |
7 |
>>> |
8 |
>>>> Volker Armin Hemmann wrote: |
9 |
>>>> |
10 |
>>>>> extremly long. So long that you have to start ooo several times a day |
11 |
>>>>> for a year so that the saved startup time equalizes the time spent |
12 |
>>>>> compiling it. |
13 |
>>>>> |
14 |
>>>> "ccache" in make.conf is enabled and MAKEOPTS has a reasonable value, I |
15 |
>>>> have set it |
16 |
>>>> to "-j2". I follow the rule MAKEOPTS=<number CPUS>. But in the case of |
17 |
>>>> openoffice, the |
18 |
>>>> ebuild overwrite this value with "-j1". For the version 2.3.x I had set |
19 |
>>>> the variable |
20 |
>>>> WANT_MP but with version 2.4 it breaks the build. But how you can see |
21 |
>>>> in the following, |
22 |
>>>> that's only a minor problem. |
23 |
>>>> |
24 |
>>> or not. So everything bigger than -j1 breaks the built. Which makes dual |
25 |
>>> core cpus useless to speed up compilation. |
26 |
>>> |
27 |
>> Not really, because if you have set -pipe in CFLAGS than you can |
28 |
>> easily, with top, check how the cpus are used. But that's it, of course. |
29 |
>> |
30 |
>> How I mentioned earlier with version 2.3.x I had set WANT_MP=true |
31 |
>> and MAKEOPTS=-j2 (and with my first builds -j4 and -j5 but that was pretty |
32 |
>> much useless, because the processes are hinder them self but they don't |
33 |
>> break |
34 |
>> the build) and that works for me. The only problem which occurred was this |
35 |
>> |
36 |
>> https://bugs.gentoo.org/show_bug.cgi?id=210065 |
37 |
>> |
38 |
>> |
39 |
>>>> wolf-di6400 0(0) 03:04 PM ~ # qlop -gH openoffice |
40 |
>>>> openoffice: Fri May 2 16:22:23 2008: 1 hour, 20 minutes, 38 seconds |
41 |
>>>> openoffice: Sat May 3 04:06:11 2008: 1 hour, 19 minutes, 12 seconds |
42 |
>>>> openoffice: 2 times |
43 |
>>>> |
44 |
>>> emerge -p openoffice-bin|genlop -p |
45 |
>>> These are the pretended packages: (this may take a while; wait...) |
46 |
>>> |
47 |
>>> [ebuild R ] app-office/openoffice-bin-2.4.0 |
48 |
>>> |
49 |
>>> |
50 |
>>> Estimated update time: 2 minutes. |
51 |
>>> |
52 |
>> Yeh, of course is that faster but why we use Gentoo? Because |
53 |
>> of the fast binary install? ;-) |
54 |
>> |
55 |
> |
56 |
> with packages that are only needed once in a while (ooo, frickelfox) binaries |
57 |
> might be the right thing to do. |
58 |
> |
59 |
How I said, everyone's own decision. |
60 |
> I have compiled ooo in the past - on much, much slower machines. Ever compiled |
61 |
> it on a 900mhz thunderbird? I did (and later faster cpus, of course). |
62 |
> |
63 |
I don't know a machine with the name thunderbird :-[ . But I started |
64 |
with Gentoo |
65 |
on a Toshiba Tecra 8100, that's a PIII Copermine 800MHz and 512 MB RAM. |
66 |
In this |
67 |
respect, I can say: Yes, I did. :-) An emerge -e world lasted 11 |
68 |
hours, without OOO, |
69 |
OOO alone needs 16 hours to build, _but_ that, for me, was the |
70 |
fascinating thing - |
71 |
The build runs faultless, not even this strange segfaults of |
72 |
typesconfig. :-D |
73 |
> Inclusive seeing it fail after 8h because the wrong java version was |
74 |
> installed. It took less time to emerge ALL of kde than ooo. And one day I |
75 |
> compared the differences. ooo started maybe 3 seconds faster than ooo-bin. As |
76 |
> soon as started, no difference at all. |
77 |
> |
78 |
That are bad experiences, but those things don't happened to me. |
79 |
Perhaps God has an eye on me. :-D |
80 |
For me isn't the start time of a program that important, but that all |
81 |
fits perfect together. |
82 |
> That was not worth the trouble. |
83 |
> |
84 |
In your case, maybe. |
85 |
|
86 |
> |
87 |
>> Although I conduct all emerges at the console _not_ in X. Perhaps |
88 |
>> that's it. However, every user should do how he/she likes. |
89 |
>> |
90 |
> |
91 |
> it does not matter where - ooo is huge - bloated. And whereever you emerge it, |
92 |
> it is the package needing the most time. |
93 |
|
94 |
That's absolutely right. |
95 |
|
96 |
W. Canis |