1 |
solar <solar@g.o> wrote on 2011/12/26 21:38:52: |
2 |
> |
3 |
> On Mon, 2011-12-26 at 19:03 +0100, Joakim Tjernlund wrote: |
4 |
> > solar <solar@g.o> wrote on 2011/12/25 18:52:52: |
5 |
> > > |
6 |
> > > On Sun, 2011-12-25 at 11:34 +0200, Kfir Lavi wrote: |
7 |
> > > > |
8 |
> > > > |
9 |
> > > > On Fri, Dec 23, 2011 at 3:23 PM, Joakim Tjernlund |
10 |
> > > > <joakim.tjernlund@×××××××××.se> wrote: |
11 |
> > > > |
12 |
> > > > I got the impression from docs that PKG_INSTALL_MASK would |
13 |
> > > > actually |
14 |
> > > > mask files out so they never get into the binary package, this |
15 |
> > > > doesn't |
16 |
> > > > seem to happen. |
17 |
> > > > Did I misunderstand? If I did, I think an MASK to do the above |
18 |
> > > > would be a worthy addition to quickpkg, very useful for |
19 |
> > > > embedded targets to keep the pkg size down. |
20 |
> > > > Also PKG_INSTALL_KEEP which would list files keep(rest is |
21 |
> > > > dropped) would be nice. |
22 |
> > > > |
23 |
> > > > Oh, something else I wonder about. How does pre/post install |
24 |
> > > > work with |
25 |
> > > > quickpkgs? Is it possible to have such scripts and then have |
26 |
> > > > qmerge execute |
27 |
> > > > them? |
28 |
> > > > |
29 |
> > > > Jocke |
30 |
> > > > |
31 |
> > > > |
32 |
> > > > Yes, PKG_INSTALL_MASK on time of installation will mask the files |
33 |
> > > > defined in it. |
34 |
> > > > Binary package do contain all files, even the masked files. |
35 |
> > > > Kfir |
36 |
> > > |
37 |
> > > |
38 |
> > > $PKG_INSTALL_MASK is supposed to omit the files in it's list from making |
39 |
> > > it into the binary pkgs in the first place. The idea there was to make |
40 |
> > > smaller binary pkgs for embedded devices and such (for use only with |
41 |
> > > private repos). |
42 |
> > > |
43 |
> > > $INSTALL_MASK is supposed to omit the files listed in it from being |
44 |
> > > installed on the file system. |
45 |
> > |
46 |
> > Hi Solar, long time no see :) |
47 |
> > |
48 |
> > So the current behaviour is a bug, good to know. |
49 |
> |
50 |
> I would not really call it a bug. *INSTALL_MASK is a portage feature |
51 |
> itself. quickpkg does not support all the features of portage itself. |
52 |
> You could/should file a feature request bug for quickpkg to add such |
53 |
> support. Maybe file a bug for portage-utils@ as well because I just |
54 |
> checked and looks like we never added support for PKG_INSTALL_MASK in |
55 |
> qpkg.c (c version of quickpkg) |
56 |
|
57 |
I don't really get the difference between the two as impl. today. |
58 |
What can PKG_INSTALL_MASK do that INSTALL_MASK can't(or vice versa)? |
59 |
|
60 |
> |
61 |
> |
62 |
> > I am contemplating a greater problem too. Our system require we can install |
63 |
> > multiple versions of our SW and switch between them. This is easy to do when it |
64 |
> > comes to our own app but not when one want to upgrade core parts of the system, like |
65 |
> > libc etc. |
66 |
> > |
67 |
> > So I am thinking one could use --bind mounts and switch_root to solve that. Basically |
68 |
> > one has a skeleton root FS with /bin, /lib, /usr /opt etc. |
69 |
> > Each upgrade goes into dirs like: |
70 |
> > bin_1.x.y/ usr_1.x.y/ lib_1.x.y/ opt_1.x.y/ sbin_1.x.y/ |
71 |
> > Then, from an initramfs, one selects which of xxx_1.x.y dir one wants to use |
72 |
> > and --bind mounts them under the corresponding skeleton dir. |
73 |
> > |
74 |
> > To do that one needs an way to repackage a root fs created by a bunch of different |
75 |
> > ebuilds, some of them spanning several of the above dirs, into a package per xx_1.x.y dir. |
76 |
> > Not sure if this can be done with the current portage/portage-utils and how, any |
77 |
> > ideas welcome :) |
78 |
> > |
79 |
> > Perhaps there is a better way then the above? |
80 |
> > |
81 |
> > Jocke |
82 |
> |
83 |
> Not sure about all this. |
84 |
|
85 |
Yeah, it is a bit much :) I either case it would be nice if one |
86 |
could group several bianry pkgs into one that qmerge can install. Is that possible? |
87 |
|
88 |
Jocke |