1 |
On Fri, 10 Jan 2014 02:19:03 +0100 |
2 |
Tom Wijsman <TomWij@g.o> wrote: |
3 |
|
4 |
> On Fri, 10 Jan 2014 08:31:21 +0800 |
5 |
> Patrick Lauer <patrick@g.o> wrote: |
6 |
> |
7 |
> > On 01/10/2014 08:16 AM, heroxbd@g.o wrote: |
8 |
> > > Igor <lanthruster@×××××.com> writes: |
9 |
> > > |
10 |
> > >> The ebuilds have approximately the same time to install, the |
11 |
> > >> failure rate is about the same, emerge is getting slower. |
12 |
> > > |
13 |
> > > I am curious about the slowness of emerge. |
14 |
> > > |
15 |
> > > How about profile the portage and rewrite the time-crucial part in |
16 |
> > > C/C++, or ideally, borrowing the counterpart from paludis? How |
17 |
> > > feasible is that? |
18 |
> > |
19 |
> > Last I checked paludis wasn't faster - on average portage was a few |
20 |
> > percents faster. |
21 |
> |
22 |
|
23 |
> > For python things you really want python or C instead of C++... |
24 |
> |
25 |
> Why is this a Python thing? What's the reason to exclude a language? |
26 |
> |
27 |
> > So, what you wanted to ask was: |
28 |
> > "Which parts of pkgcore can be migrated into portage?" |
29 |
> |
30 |
> Or rather: "What does it take to migrate parts of pkgcore into |
31 |
> portage?" |
32 |
|
33 |
Why not just switch to using pkgcore as the default package manager. |
34 |
radhermit has been doing a lot of work lately getting EAPI 5 support |
35 |
added, and generally fixing bugs etc. |
36 |
|
37 |
Since we no longer have anyone intimately familiar with the |
38 |
portage code, we should just switch to a more modern and readable |
39 |
system. Rather than having random people trying to learn the convoluted |
40 |
portage code, let's concentrate on getting pkgcore to a point where |
41 |
we can replace portage with it. |
42 |
|
43 |
> > > I guess the dep-tree calculation is the slowest part. |
44 |
> > Yes, it's doing lots of silly dynamic things (backtracking), |
45 |
> |
46 |
> Exactly, without backtracking (which I can live without) it takes |
47 |
> around half a minute as compared to a lot of minutes; when it started |
48 |
> to take almost half an hour it was time to completely nuke |
49 |
> backtracking. |
50 |
> |
51 |
> Now I happily live without ever needing to enable backtracking |
52 |
> again. :) |
53 |
> |
54 |
> > and portage codebase on average is not designed for speed. |
55 |
> |
56 |
> Yes, inspecting the profiler results has me worried; though I find |
57 |
> half a minute acceptable for the amount of packages that are involved |
58 |
> on my system, so, I'm less concerned but it is still interesting that |
59 |
> there is the possibility to do things faster. It's just some work to |
60 |
> actually get it to do that, which requires much more dedication, time |
61 |
> and knowledge than doing an occasional commit. |
62 |
> |
63 |
> > As a first step I would recommend profiling it and removing unneeded |
64 |
> > stuff (do less work!), rewriting parts in C is a lot of work and not |
65 |
> > needed for the first round of speedups. |
66 |
> |
67 |
> See my other mail <20140110020218.0f6244d5@TOMWIJ-GENTOO> for recent |
68 |
> profiling images; we should indeed start with dealing with the low |
69 |
> hanging fruit (there are quite some 0-2% speed improvements possible), |
70 |
> as that can get us to speed it up faster than trying to deal with some |
71 |
> algorithmic complexity that needs far more insight to redesign, let |
72 |
> alone to properly re-implement it. |
73 |
> |