1 |
On Fri, Aug 12, 2005 at 10:14:35AM +0900, Jason Stubbs wrote: |
2 |
> On Friday 12 August 2005 02:26, Marius Mauch wrote: |
3 |
> > Hi, |
4 |
> > |
5 |
> > For all those drooling over their keyboards after reading this topic, |
6 |
> > please also read the rest of this mail. |
7 |
> > So yes, finally a portage-2.1 pre-pre-pre-alpha version is out and in |
8 |
> > the tree (p.masked). However, it's not the 2.1 that some of you might |
9 |
> > expect as it doesn't have a new dep resolver, so still no use deps. |
10 |
> > |
11 |
|
12 |
Couple of other things... |
13 |
1) use sandbox 1.2.12. if you use 1.2.11 (stable), ebuild-daemon will |
14 |
be a bit borked whenever sandbox is enabled. |
15 |
|
16 |
2) Cache rewrite, eclass_cache and actual tree cache are now one, and |
17 |
cache classes are repository wide. |
18 |
English: remote cache is sane now, but the rewrite in 2.1 *will* |
19 |
lack a faster searchDesc; the short version is that 2.1 lacks the |
20 |
restriction subsystem, as such it's still the old crappy approach |
21 |
|
22 |
3) fakeroot integration. Allows you to run the install phase as |
23 |
non-root, enabled via FEATURES="userpriv_fakeroot". Remaining |
24 |
ebuild phases that are ran as root are prerm, postrm, preinst, |
25 |
postinst, and setup (why again must setup be root? :) |
26 |
|
27 |
4) confcache. Stuart came up with a nifty hack that relies |
28 |
on sandbox debug logs to track what autoconf checks for configure |
29 |
calls; uses this information to maintain an autoconf cache, |
30 |
speeding up configures pretty massively. Doesn't play nice on |
31 |
multifreq procs (sorry cpu-freq users) due to changing md5 of |
32 |
/proc/cpuinfo, and still has a few bugs. Useful, but a bit raw. |
33 |
|
34 |
5) parallelized fetching when FEATURES="parallel-fetch" is on. Kicks |
35 |
in whenever pkg list is > 1, requires FEATURES="distlocks" to be |
36 |
enabled. |
37 |
|
38 |
6) save/restoration of ebuild envs; bug 56408. So... a few ebuilds |
39 |
may not play nice. If you find one, note it's a bug *strictly* |
40 |
for an experimental/alpha version of portage, so do the legwork, |
41 |
and notify the maintainer of what needs to be changed so that the |
42 |
ebuild works with stable *and* with upcoming changes. |
43 |
Haven't hit many of them, but their are still a few lingering I'm |
44 |
sure. |
45 |
Bonus of it, is that you can yank eclasses from the tree once this |
46 |
is deployed (do it now, and the long standing threat of breaking |
47 |
the devs knees will be executed). |
48 |
|
49 |
7) metadata.xml *basic* integration. emerge -sv uses it. |
50 |
This is *strictly* just a UI parsing of metadata.xml's, dumping |
51 |
maintainer, herd, etc. |
52 |
|
53 |
8) code clean up of non-core portage lib (ongoing), and start of |
54 |
modularization, and start of class docs for new stuff. |
55 |
|
56 |
~harring |