1 |
On Mon, 31 Dec 2012 16:53:47 +0800 |
2 |
kwkhui@××××.net wrote: |
3 |
|
4 |
> On Mon, 31 Dec 2012 10:03:40 +0200 |
5 |
> Alan McKinnon <alan.mckinnon@×××××.com> wrote: |
6 |
> |
7 |
> > It's not in the profile, the xorg-server ebuild sets USE="suid" on |
8 |
> > by default. |
9 |
> > |
10 |
> > Most likely is that Walter has USE="-suid" in his make.conf and sets |
11 |
> > it back on for things he's checked out personally. Meaning that in |
12 |
> > this case one slipped through. |
13 |
> |
14 |
> I suspect it is a USE="-* (blah)" rather than an explicit USE="-suid" |
15 |
> in the make.conf file. |
16 |
> |
17 |
> One question though --- should the xorg-server ebuild be such that |
18 |
> IUSE="(blah) +suid" when using a hardened-profile? |
19 |
|
20 |
That already has a de-facto answer; USE="suid" must be on by default |
21 |
as without it users cannot run a desktop (xorg-server does not yet run |
22 |
without root permissions) |
23 |
|
24 |
> Also, checking |
25 |
> my PORTDIR, given the global description in use.desc (suid - Enable |
26 |
> setuid root program, with potential security risks), shouldn't the |
27 |
> suid use flag entries (net-analyzer/nagios-plugins:suid and |
28 |
> net-wireless/kismet:suid) be deleted from use.local.desc? |
29 |
|
30 |
I see this is being discussed on -dev ATM. Duncan has this to say: |
31 |
|
32 |
"Promoting a flag to global does mean it gets a global description in |
33 |
use.desc, but per package descriptions (as now maintained in the per- |
34 |
package metadata.xml files, but there's a tree maintenance script that |
35 |
keeps use.local.desc current based on the metadata files, to keep the |
36 |
tools using it working) continue to be encouraged where they are |
37 |
useful, as they can often provide much more detailed per-package |
38 |
descriptions of what the flag actually does in that specific package, |
39 |
than the global description can." |
40 |
|
41 |
The current policy seems to be the sensible one: A global generic |
42 |
description can exist, but more specific package-level descriptions are |
43 |
also supported. I'd agree with that; a policy of "only global |
44 |
descriptions" or "no global description if a local one exists" would be |
45 |
overly restrictive and just cause problems. On the whole, we humans are |
46 |
perfectly OK with the idea of over-loading concepts; this is not |
47 |
something we have problems with. |
48 |
|
49 |
|
50 |
|
51 |
-- |
52 |
Alan McKinnon |
53 |
alan.mckinnon@×××××.com |