1 |
On Saturday 01 May 2004 19:46, Jerry McBride wrote: |
2 |
> Move all documentation to html and place it into a directory that apache |
3 |
> can disseminate from. Having a central, online, browseable doc tree would |
4 |
> be beautiful. Priceless... |
5 |
|
6 |
All package documentation is accessible from http://localhost/doc/ in the |
7 |
default apache configuration files. |
8 |
|
9 |
> Move away from the filesystem database that portage currently uses to |
10 |
> something with a bit more perfomance... mysql, sqlite even postgres... An |
11 |
> SQL backend interface would be WONDERFUL. |
12 |
|
13 |
The biggest problem is that the ebuilds can only be parsed using bash. If that |
14 |
restriction can be lifted (by restricting / changing the ebuild format) |
15 |
parsing should go a lot faster. Really the installed package database is not |
16 |
the problem. |
17 |
|
18 |
> Port portage to c or c++. Python sucks in the performance department... If |
19 |
> not port to c then atleast start using psyco for a small perf boost... |
20 |
> Rebuilding the cache (reading files) after a sync is... terrible... What a |
21 |
> kludge... |
22 |
|
23 |
Ever heard of the 10/90 rule. Really python is not the problem, the cache |
24 |
rebuilding is again ebuild parsing. |
25 |
|
26 |
> |
27 |
> Work on eliminating the constant caching of dependencies and other caching |
28 |
> processes during boot up... On some of my boxes, these caching periods take |
29 |
> longer than it does for KDE to boot to a usable desktop... What a drag on a |
30 |
> laptop that I use to display linux with.... It's the first thing that |
31 |
> windows advocates point and laugh at... Give me a break and cache the damn |
32 |
> things when changes are MADE, not everytime you boot the box... |
33 |
|
34 |
Probably those calculations should be rethought and a better algorithm should |
35 |
be devised. Calculating at boot time does make sense though as we have no way |
36 |
to know when the calculations are invalid. |
37 |
|
38 |
> Tone down the amount of noise that emerge prints to the screen when using |
39 |
> it. I don't need or want to see maybe 99.99% of all the stuff it prints |
40 |
> during it operations. All you need is... either "THERE WAS AN ERROR" or |
41 |
> "THERE WERE NO ERRORS". That in and of itself would boost performance a |
42 |
> lot!... Less console updates, more cpu for the task at hand... Less user |
43 |
> confusion... |
44 |
> |
45 |
> PROGRESS BARS...PROGRESS BARS...PROGRESS BARS...PROGRESS BARS...PROGRESS |
46 |
> BARS...PROGRESS BARS...PROGRESS BARS...PROGRESS BARS...PROGRESS |
47 |
> BARS...PROGRESS BARS... |
48 |
> |
49 |
> Got the point??? :') |
50 |
|
51 |
I see what you mean. There are two problems. It is very hard to make a |
52 |
progress bar as using useflags means that we cannot give a good idea of the |
53 |
amount of code to be compiled in total. Even measuring what has allready been |
54 |
done is hard. Also removing the build output creates much problems for |
55 |
bug-hunting. |
56 |
|
57 |
If you don't want to see the compilation happen. Use "emerge foo &>/dev/null&" |
58 |
and you will see nothing. When bash has finished it will tell you whether it |
59 |
was successfull. (Instead you could direct to a file that you can check upon |
60 |
once in a while) |
61 |
|
62 |
Paul |
63 |
|
64 |
-- |
65 |
Paul de Vrieze |
66 |
Gentoo Developer |
67 |
Mail: pauldv@g.o |
68 |
Homepage: http://www.devrieze.net |