[gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

Hi,

In case you aren't familiar with FEATURES=userpriv, here's the
description from the make.conf(5) man page:

  Allow portage to drop root privileges and compile packages as
  portage:portage without a sandbox (unless usersandbox is also used).

The rationale for having the separate "usersandbox" setting, to enable
use of sys-apps/sandbox, is that people who enable userpriv sometimes
prefer to have sandbox disabled in order to slightly improve
performance. However, I would recommend to enable usersandbox by
default, for the purpose of logging sandbox violations.

Note that ebuilds can set RESTRICT="userpriv" if they require superuser
privileges during any of the src_* phases that userpriv affects.

I've been using FEATURES="userpriv usersandbox" for years, and I don't
remember experiencing any problems because of it, so I think that it
would be reasonable to have it enabled by default. Objections?
-- 
Thanks,
Zac

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Andreas K. Huettel]

[-- Attachment #1: Type: Text/Plain, Size: 403 bytes --]

Am Montag 28 Mai 2012, 23:34:22 schrieb Zac Medico:
> I've been using FEATURES="userpriv usersandbox" for years, and I don't
> remember experiencing any problems because of it, so I think that it
> would be reasonable to have it enabled by default. Objections?

No objections. Excellent idea.

-- 

Andreas K. Huettel
Gentoo Linux developer 
dilfridge@gentoo.org
http://www.akhuettel.de/


[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Davide Pesavento]

On Mon, May 28, 2012 at 11:34 PM, Zac Medico <zmedico@gentoo.org> wrote:
> Hi,
>
> In case you aren't familiar with FEATURES=userpriv, here's the
> description from the make.conf(5) man page:
>
>  Allow portage to drop root privileges and compile packages as
>  portage:portage without a sandbox (unless usersandbox is also used).
>
> The rationale for having the separate "usersandbox" setting, to enable
> use of sys-apps/sandbox, is that people who enable userpriv sometimes
> prefer to have sandbox disabled in order to slightly improve
> performance. However, I would recommend to enable usersandbox by
> default, for the purpose of logging sandbox violations.
>
> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
> privileges during any of the src_* phases that userpriv affects.
>
> I've been using FEATURES="userpriv usersandbox" for years, and I don't
> remember experiencing any problems because of it, so I think that it
> would be reasonable to have it enabled by default. Objections?
> --
> Thanks,
> Zac
>

I've been using both FEATURES for a few years too, seemingly without
adverse effects, so +1 from me.

Thanks,
Pesa

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Michael Weber]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 05/28/2012 11:34 PM, Zac Medico wrote:
> I've been using FEATURES="userpriv usersandbox" for years, and I
> don't remember experiencing any problems because of it, so I think
> that it would be reasonable to have it enabled by default.
> Objections?

It was/is default on default/linux/amd64/10.0/developer (the one we
all should use ?)

+1

- -- 
- --
Gentoo Dev
http://xmw.de/
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iF4EAREIAAYFAk/EB5EACgkQknrdDGLu8JBVyAD/a/Szj+swzSIkAgZv2bGzezIQ
M/2+tZUUk+ZE6HlkDrsA/RufmJGlAEa9MJtImaTo/h9svEG/BhioQNvo49nT2ssi
=IRjv
-----END PGP SIGNATURE-----

[gentoo-dev] Re: RFC: Enable FEATURES="userpriv usersandbox" by default?

[Duncan]

Zac Medico posted on Mon, 28 May 2012 14:34:22 -0700 as excerpted:

> In case you aren't familiar with FEATURES=userpriv, here's the
> description from the make.conf(5) man page:
> 
>   Allow portage to drop root privileges and compile packages as
>   portage:portage without a sandbox (unless usersandbox is also used).
> 
> The rationale for having the separate "usersandbox" setting, to enable
> use of sys-apps/sandbox, is that people who enable userpriv sometimes
> prefer to have sandbox disabled in order to slightly improve
> performance. However, I would recommend to enable usersandbox by
> default, for the purpose of logging sandbox violations.
> 
> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
> privileges during any of the src_* phases that userpriv affects.
> 
> I've been using FEATURES="userpriv usersandbox" for years, and I don't
> remember experiencing any problems because of it, so I think that it
> would be reasonable to have it enabled by default. Objections?

I saw the thread on portage-dev so was waiting for the thread here that 
you mentioned you'd start...

Some years ago I had some problem or other with the usersandbox and 
userpriv combination (AFAIK it would work with just one of the two, but 
not both), but that was several years ago now, and was almost certainly 
~arch (and possibly pre-unmask), so yes, I'd say have them both on by 
default.  I've had no problem with it recently.

As is traditional for this sort of defaults-change, I'd suggest creating 
a news item for it, with the usual paragraph explanation and referral to 
the manpage and/or handbook for more information.

If I don't miss my guess, there's likely a number of folks that had 
either userpriv or userstandbox disabled for some package or other, years 
ago, who simply forgot about it and never reenabled.  I'm usually pretty 
good about that, and only probably 6-8 months ago realized I had one of 
the two disabled, and couldn't remember why (probably 2-3 years ago I 
started putting dated comments in the config when I did stuff like that, 
so whatever it was, was awhile back...), so it had obviously been 
disabled for awhile.  (I've done at least one and I think two full emerge 
--emptytree @worlds since then, however, so as I said above, everything 
that's installed now is fine.)  A news item will help remind folks with 
older installs to check their status as well, which can only be a good 
thing. =:^)

So from this user, +1 (+1000? =:^), news item requested. =:^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Maxim Kammerer]

On Tue, May 29, 2012 at 12:34 AM, Zac Medico <zmedico@gentoo.org> wrote:
> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
> privileges during any of the src_* phases that userpriv affects.

Current list of packages in portage using userpriv restriction:

app-laptop/tp_smapi
dev-db/firebird
games-board/gnuchess-book
games-fps/quakeforge
games-rpg/wastesedge
gnome-extra/gnome-lirc-properties
mail-filter/qmail-scanner (vpopmail)
media-gfx/gtkimageview
media-gfx/imagemagick (when USE=perl)
net-dialup/ltmodem
net-libs/courier-authlib (vpopmail)
net-mail/courier-imap (vpopmail)
net-mail/qmailadmin (vpopmail)
net-mail/vpopmail (old stable)
net-misc/icaclient
sys-fs/udev (when USE=test for udev-9999 only)

It could also be that anything vpopmail-related doesn't need
RESTRICT=userpriv anymore.

> I've been using FEATURES="userpriv usersandbox" for years, and I don't
> remember experiencing any problems because of it, so I think that it
> would be reasonable to have it enabled by default.

Ditto, ~2 years with regular full @world rebuild.

-- 
Maxim Kammerer
Liberté Linux: http://dee.su/liberte

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Rich Freeman]

On Mon, May 28, 2012 at 9:09 PM, Maxim Kammerer <mk@dee.su> wrote:
> Ditto, ~2 years with regular full @world rebuild.
>

Yup, been years since the last time I even saw a bug for this.

Probably wouldn't hurt to announce in news if it will impact existing
users.  I doubt anybody would actually remove the portage user, but
never hurts just to make people aware...

Rich

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Agostino Sarubbo]

[-- Attachment #1: Type: text/plain, Size: 1157 bytes --]

On Monday 28 May 2012 14:34:22 Zac Medico wrote:
> Hi,
> 
> In case you aren't familiar with FEATURES=userpriv, here's the
> description from the make.conf(5) man page:
> 
>   Allow portage to drop root privileges and compile packages as
>   portage:portage without a sandbox (unless usersandbox is also used).
> 
> The rationale for having the separate "usersandbox" setting, to enable
> use of sys-apps/sandbox, is that people who enable userpriv sometimes
> prefer to have sandbox disabled in order to slightly improve
> performance. However, I would recommend to enable usersandbox by
> default, for the purpose of logging sandbox violations.
> 
> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
> privileges during any of the src_* phases that userpriv affects.
> 
> I've been using FEATURES="userpriv usersandbox" for years, and I don't
> remember experiencing any problems because of it, so I think that it
> would be reasonable to have it enabled by default. Objections?

I'm using usersync since a long time, how about add it too?
-- 
Agostino Sarubbo		ago -at- gentoo.org
Gentoo/AMD64 Arch Security Liaison
GPG: 0x7CD2DC5D

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Richard Yao]

[-- Attachment #1: Type: text/plain, Size: 151 bytes --]

On 05/29/12 04:43, Agostino Sarubbo wrote:
> I'm using usersync since a long time, how about add it too?

This is also a good idea. I second it.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 900 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

On 05/29/2012 01:43 AM, Agostino Sarubbo wrote:
> On Monday 28 May 2012 14:34:22 Zac Medico wrote:
>> Hi,
>>
>> In case you aren't familiar with FEATURES=userpriv, here's the
>> description from the make.conf(5) man page:
>>
>>   Allow portage to drop root privileges and compile packages as
>>   portage:portage without a sandbox (unless usersandbox is also used).
>>
>> The rationale for having the separate "usersandbox" setting, to enable
>> use of sys-apps/sandbox, is that people who enable userpriv sometimes
>> prefer to have sandbox disabled in order to slightly improve
>> performance. However, I would recommend to enable usersandbox by
>> default, for the purpose of logging sandbox violations.
>>
>> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
>> privileges during any of the src_* phases that userpriv affects.
>>
>> I've been using FEATURES="userpriv usersandbox" for years, and I don't
>> remember experiencing any problems because of it, so I think that it
>> would be reasonable to have it enabled by default. Objections?
> 
> I'm using usersync since a long time, how about add it too?

Yeah, I think that would be a good default too. I guess the portage
ebuild can do a recursive adjustment of $PORTDIR permissions in
pkg_postinst, in order to solve bug #277970 [1].

For userpriv, it will have to do a similar recursive adjustment of
permissions for directories inside $DISTDIR (such as git-src and
svn-src), since userpriv causes src_unpack to run with lower privileges.

[1] https://bugs.gentoo.org/show_bug.cgi?id=277970
-- 
Thanks,
Zac

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1664 bytes --]

On Tue, 29 May 2012 02:05:08 -0700
Zac Medico <zmedico@gentoo.org> wrote:

> On 05/29/2012 01:43 AM, Agostino Sarubbo wrote:
> > On Monday 28 May 2012 14:34:22 Zac Medico wrote:
> >> Hi,
> >>
> >> In case you aren't familiar with FEATURES=userpriv, here's the
> >> description from the make.conf(5) man page:
> >>
> >>   Allow portage to drop root privileges and compile packages as
> >>   portage:portage without a sandbox (unless usersandbox is also
> >> used).
> >>
> >> The rationale for having the separate "usersandbox" setting, to
> >> enable use of sys-apps/sandbox, is that people who enable userpriv
> >> sometimes prefer to have sandbox disabled in order to slightly
> >> improve performance. However, I would recommend to enable
> >> usersandbox by default, for the purpose of logging sandbox
> >> violations.
> >>
> >> Note that ebuilds can set RESTRICT="userpriv" if they require
> >> superuser privileges during any of the src_* phases that userpriv
> >> affects.
> >>
> >> I've been using FEATURES="userpriv usersandbox" for years, and I
> >> don't remember experiencing any problems because of it, so I think
> >> that it would be reasonable to have it enabled by default.
> >> Objections?
> > 
> > I'm using usersync since a long time, how about add it too?
> 
> Yeah, I think that would be a good default too. I guess the portage
> ebuild can do a recursive adjustment of $PORTDIR permissions in
> pkg_postinst, in order to solve bug #277970 [1].

Wouldn't that break users who sync using a regular user? And then break
again, and again every time portage is merged?


-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Rich Freeman]

On Tue, May 29, 2012 at 10:11 AM, Michał Górny <mgorny@gentoo.org> wrote:
>
> Wouldn't that break users who sync using a regular user? And then break
> again, and again every time portage is merged?

Yup, unless that regular user is the same one used for userpriv (if
I'm correctly understanding the problem that you're pointing at).  I
don't see this as a show-stopper - just a reason to have a news item.
Those not using userpriv can always disable it and run as root as they
are already doing.  Those who are using a regular user to sync could
ensure that their make.conf uses the same user for userpriv.

Rich

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[hasufell]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 05/29/2012 04:50 PM, Rich Freeman wrote:
> On Tue, May 29, 2012 at 10:11 AM, Michał Górny <mgorny@gentoo.org>
> wrote:
>> 
>> Wouldn't that break users who sync using a regular user? And then
>> break again, and again every time portage is merged?
> 
> Yup, unless that regular user is the same one used for userpriv
> (if I'm correctly understanding the problem that you're pointing
> at).  I don't see this as a show-stopper - just a reason to have a
> news item. Those not using userpriv can always disable it and run
> as root as they are already doing.  Those who are using a regular
> user to sync could ensure that their make.conf uses the same user
> for userpriv.
> 
> Rich
> 

- -1

I am against too many defaults. It's documented and people can
activate it.
I'm already annoyed by pre-set stuff like "cups" in
releases/make.defaults.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJPxOPzAAoJEFpvPKfnPDWz9d8H/A0AQr57nDv/0n0+jN8bxdxc
nPQyBN9faSuh8IYztQmFe1Xpn/JFx9LoqRGQrvncMmzjmPkM1iaoXUpuo/qw5Fys
ar9pN84yZoAJuzgMdLzLs0U/6lqkvLzO+x1Y5DkNU2F+h3Bx9sAk+4vCUjEYg/pC
UdXkeRONaB62p/D2T2ucP6IuG6qBI/raW7vvDvkiDGzVbNnDBe4hGESh3Fb4Gd/Y
x/P/QJ+cZvFF3SvqhORMeXlgccbqU2kBy2Bwcq2GwKKmYIdKwnA2J0KKwqLkHraD
8pkTzUsvqxnQVqFGfCvFyJe3uwiJKQoTIAGugf3n9irvczuZTQ9MDWoZkGKiaNI=
=eo74
-----END PGP SIGNATURE-----

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Rich Freeman]

On Tue, May 29, 2012 at 10:57 AM, hasufell <hasufell@gentoo.org> wrote:
> I am against too many defaults. It's documented and people can
> activate it.
> I'm already annoyed by pre-set stuff like "cups" in
> releases/make.defaults.

While universal agreement is a bit much to hope for, I just wanted to
point out that fewer defaults is really just an illusion.

There is ALWAYS a default, anytime you have an option.  The default
might be one thing, or it might be another, but there is ALWAYS a
default.  My thinking is that our defaults should generally reflect
the most mainstream or least-surprising behavior, especially where
there are upstream projects.  in the case of portage, we are the
upstream, so we should do whatever is most useful and least obnoxious
to our users.

If you're running something other than a generic desktop/server, there
will always be a need to tweak things.

Rich

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[hasufell]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 05/29/2012 05:23 PM, Rich Freeman wrote:
> On Tue, May 29, 2012 at 10:57 AM, hasufell <hasufell@gentoo.org>
> wrote:
>> I am against too many defaults. It's documented and people can 
>> activate it. I'm already annoyed by pre-set stuff like "cups" in 
>> releases/make.defaults.
> 
> While universal agreement is a bit much to hope for, I just wanted
> to point out that fewer defaults is really just an illusion.
> 
> There is ALWAYS a default, anytime you have an option.  The
> default might be one thing, or it might be another, but there is
> ALWAYS a default.  My thinking is that our defaults should
> generally reflect the most mainstream or least-surprising behavior,
> especially where there are upstream projects.  in the case of
> portage, we are the upstream, so we should do whatever is most
> useful and least obnoxious to our users.
> 
> If you're running something other than a generic desktop/server,
> there will always be a need to tweak things.
> 
> Rich
> 

Well then let my clarify: I'm against too many pre-set (meaning
"activated") features/useflags.

That's probably a seperate discussion, but I myself would expect the
_default_ profile/config to have almost nothing activated. No
useflags, no features etc.

That may imply that this default is "broken", but it takes more time
to do reverse-configuration while looking for things that someone
considered "sane" and has set for your "convenience".

I discovered this the first time I set up a blank chroot and got a
load of stuff pulled in by some trivial emerges. Some set by already
mentioned releases/make.defaults and similar, some set by ebuilds etc.

What you do with other profiles is a completely different topic,
because I'm not forced to use them.

means: I don't like the fact that I have to set
FEATURES="-foobar"
or
USE="-foobar"

That should almost never be the case (unless I set some globally and
unset some locally or use desktop-profiles etc).

am I offtopic already? Hope you got the point though.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJPxPkHAAoJEFpvPKfnPDWzejcH/3g1VGmSRHufoQMHUpi6X1x3
31pNy2Q+SKxo4voy5Y1/mt+0lKGrhyDq6npmBY+7n5RlhdKrn8J3VyQ7HQ1jBGiS
nEdSVb6BCHtFeWWWYRo6efooQFsGT+6NOFQgX/xXXgk9Ndzk8LtURGp8oP0oucNt
YWfhDruoUzJXRyIMP9u6SbbDVXOnYVP+WUniNJ855l2Q1jg5lrwE6f6dD7wsbtyp
3PGBEtMqX9nAtzFZ8blUHngyrMP9J/GcJ3OVQkLXla7WBCWLqKlN0pIIiVqe2L5V
45MPQ/Muhyy0JUKLmLJLvx/2c+1I4mCt1lrfZNNN3zhepnjZSLn/uiGZk3JVEQs=
=KNF8
-----END PGP SIGNATURE-----

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Jeff Horelick]

On 29 May 2012 12:27, hasufell <hasufell@gentoo.org> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 05/29/2012 05:23 PM, Rich Freeman wrote:
>> On Tue, May 29, 2012 at 10:57 AM, hasufell <hasufell@gentoo.org>
>> wrote:
>>> I am against too many defaults. It's documented and people can
>>> activate it. I'm already annoyed by pre-set stuff like "cups" in
>>> releases/make.defaults.
>>
>> While universal agreement is a bit much to hope for, I just wanted
>> to point out that fewer defaults is really just an illusion.
>>
>> There is ALWAYS a default, anytime you have an option.  The
>> default might be one thing, or it might be another, but there is
>> ALWAYS a default.  My thinking is that our defaults should
>> generally reflect the most mainstream or least-surprising behavior,
>> especially where there are upstream projects.  in the case of
>> portage, we are the upstream, so we should do whatever is most
>> useful and least obnoxious to our users.
>>
>> If you're running something other than a generic desktop/server,
>> there will always be a need to tweak things.
>>
>> Rich
>>
>
> Well then let my clarify: I'm against too many pre-set (meaning
> "activated") features/useflags.
>
> That's probably a seperate discussion, but I myself would expect the
> _default_ profile/config to have almost nothing activated. No
> useflags, no features etc.
>
> That may imply that this default is "broken", but it takes more time
> to do reverse-configuration while looking for things that someone
> considered "sane" and has set for your "convenience".
>
> I discovered this the first time I set up a blank chroot and got a
> load of stuff pulled in by some trivial emerges. Some set by already
> mentioned releases/make.defaults and similar, some set by ebuilds etc.
>
> What you do with other profiles is a completely different topic,
> because I'm not forced to use them.
>
> means: I don't like the fact that I have to set
> FEATURES="-foobar"
> or
> USE="-foobar"
>
> That should almost never be the case (unless I set some globally and
> unset some locally or use desktop-profiles etc).
>
> am I offtopic already? Hope you got the point though.
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v2.0.19 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
>
> iQEcBAEBAgAGBQJPxPkHAAoJEFpvPKfnPDWzejcH/3g1VGmSRHufoQMHUpi6X1x3
> 31pNy2Q+SKxo4voy5Y1/mt+0lKGrhyDq6npmBY+7n5RlhdKrn8J3VyQ7HQ1jBGiS
> nEdSVb6BCHtFeWWWYRo6efooQFsGT+6NOFQgX/xXXgk9Ndzk8LtURGp8oP0oucNt
> YWfhDruoUzJXRyIMP9u6SbbDVXOnYVP+WUniNJ855l2Q1jg5lrwE6f6dD7wsbtyp
> 3PGBEtMqX9nAtzFZ8blUHngyrMP9J/GcJ3OVQkLXla7WBCWLqKlN0pIIiVqe2L5V
> 45MPQ/Muhyy0JUKLmLJLvx/2c+1I4mCt1lrfZNNN3zhepnjZSLn/uiGZk3JVEQs=
> =KNF8
> -----END PGP SIGNATURE-----
>

I disagree with this. I think Gentoo should be about SANE defaults. If
you want a minimal system, you can turn off all the USE flags and/or
FEATURES and/or use the standard (not desktop/) profile. SANE defaults
like FEATURES="userpriv usersandbox" are optimal for probably 90% of
users and if you're not one of those 90%, there'll be a news item,
just turn them off...

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Ralph Sennhauser]

[-- Attachment #1: Type: text/plain, Size: 444 bytes --]

On Tue, 29 May 2012 18:27:51 +0200
hasufell <hasufell@gentoo.org> wrote:

> Well then let my clarify: I'm against too many pre-set (meaning
> "activated") features/useflags.

Think of it as nouserpriv feature. ;) Either way, to disable userpriv
is kind of working against QA as a package really should be build-able
as non root user but then.

Have userpriv and usersandbox enabled since it's became available, no
issues to report.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Michael Orlitzky]

How about introducing e.g. FEATURES="nouserpriv", and make the current
userpriv behavior the default?

The migration might be a bit more confusing, but it allows portage to
gradually adopt better stuff without having FEATURES="everything under
the sun".

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Mike Gilbert]

On Tue, May 29, 2012 at 3:46 PM, Michael Orlitzky <michael@orlitzky.com> wrote:
> How about introducing e.g. FEATURES="nouserpriv", and make the current
> userpriv behavior the default?
>

Portage currently defaults to running the build process as root. The
entire point of this thread is that Zac wants to change the default to
build as the portage user (FEATURES="userpriv" in make.globals).

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[William Hubbs]

[-- Attachment #1: Type: text/plain, Size: 348 bytes --]

On Tue, May 29, 2012 at 03:46:39PM -0400, Michael Orlitzky wrote:
> How about introducing e.g. FEATURES="nouserpriv", and make the current
> userpriv behavior the default?

No. Please stay away from things like this.
It is reverse logic and can be very confusing. Just adding "-userpriv"
to your features would do exactly the same thing.

William


[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Michael Orlitzky]

On 05/29/12 15:58, Mike Gilbert wrote:
> On Tue, May 29, 2012 at 3:46 PM, Michael Orlitzky <michael@orlitzky.com> wrote:
>> How about introducing e.g. FEATURES="nouserpriv", and make the current
>> userpriv behavior the default?
>>
> 
> Portage currently defaults to running the build process as root. The
> entire point of this thread is that Zac wants to change the default to
> build as the portage user (FEATURES="userpriv" in make.globals).
> 

Right, I was just offering a way to change the default behavior without
adding another value to the FEATURES variable, which seems to be
hasufell's objection.

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

On 05/29/2012 07:11 AM, Michał Górny wrote:
> On Tue, 29 May 2012 02:05:08 -0700
> Zac Medico <zmedico@gentoo.org> wrote:
> 
>> On 05/29/2012 01:43 AM, Agostino Sarubbo wrote:
>>> I'm using usersync since a long time, how about add it too?
>>
>> Yeah, I think that would be a good default too. I guess the portage
>> ebuild can do a recursive adjustment of $PORTDIR permissions in
>> pkg_postinst, in order to solve bug #277970 [1].
> 
> Wouldn't that break users who sync using a regular user?

No, because the "usersync" feature causes the rsync process to inherit
the UID and GID of the PORTDIR directory (obtained using the stat function).

> And then break
> again, and again every time portage is merged?

No, I would not want to trigger an relatively expensive operation like
this more that once. So, it would only be triggered in pkg_postinst if
the replaced version of portage did not have usersync enabled by default.
-- 
Thanks,
Zac

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Hilco Wijbenga]

On 29 May 2012 12:46, Michael Orlitzky <michael@orlitzky.com> wrote:
> How about introducing e.g. FEATURES="nouserpriv", and make the current
> userpriv behavior the default?

rootpriv instead of nouserpriv?

> The migration might be a bit more confusing, but it allows portage to
> gradually adopt better stuff without having FEATURES="everything under
> the sun".

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 05/29/2012 07:57 AM, hasufell wrote:
> I am against too many defaults. It's documented and people can 
> activate it. I'm already annoyed by pre-set stuff like "cups" in 
> releases/make.defaults.

In the case of userpriv and usersync, I expect that we can eventually
make them unconditional, so that they'll no longer need to be listed
in FEATURES.
- -- 
Thanks,
Zac
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk/FSSIACgkQ/ejvha5XGaPTnwCg0QAe1WtZv/wMlMvb5WrxbTk+
jq4AnjTTo77BXYr0d+4F/6P3/447Jk7t
=CuDh
-----END PGP SIGNATURE-----

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

On 05/29/2012 02:47 PM, Hilco Wijbenga wrote:
> On 29 May 2012 12:46, Michael Orlitzky <michael@orlitzky.com> wrote:
>> How about introducing e.g. FEATURES="nouserpriv", and make the current
>> userpriv behavior the default?
> 
> rootpriv instead of nouserpriv?

What's the use case for this? Can't we just enable userpriv
unconditionally, so that it doesn't have to be listed in FEATURES? Note
that ebuilds will still be able to use RESTRICT=userpriv if necessary.
-- 
Thanks,
Zac

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Richard Yao]

[-- Attachment #1: Type: text/plain, Size: 646 bytes --]

On 05/29/12 18:11, Zac Medico wrote:
> On 05/29/2012 02:47 PM, Hilco Wijbenga wrote:
>> On 29 May 2012 12:46, Michael Orlitzky <michael@orlitzky.com> wrote:
>>> How about introducing e.g. FEATURES="nouserpriv", and make the current
>>> userpriv behavior the default?
>>
>> rootpriv instead of nouserpriv?
> 
> What's the use case for this? Can't we just enable userpriv
> unconditionally, so that it doesn't have to be listed in FEATURES? Note
> that ebuilds will still be able to use RESTRICT=userpriv if necessary.

Would FEATURES=-userpriv still work at the command line? It could be
useful for debugging to keep that working.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 900 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

On 05/29/2012 04:22 PM, Richard Yao wrote:
> On 05/29/12 18:11, Zac Medico wrote:
>> On 05/29/2012 02:47 PM, Hilco Wijbenga wrote:
>>> On 29 May 2012 12:46, Michael Orlitzky <michael@orlitzky.com> wrote:
>>>> How about introducing e.g. FEATURES="nouserpriv", and make the current
>>>> userpriv behavior the default?
>>>
>>> rootpriv instead of nouserpriv?
>>
>> What's the use case for this? Can't we just enable userpriv
>> unconditionally, so that it doesn't have to be listed in FEATURES? Note
>> that ebuilds will still be able to use RESTRICT=userpriv if necessary.
> 
> Would FEATURES=-userpriv still work at the command line? It could be
> useful for debugging to keep that working.

Yeah, I guess it would be bad for it to be unconditional, because
permission issues seem to be a really common source of trouble for
people. Even something as seemingly simple as userfetch probably
shouldn't be unconditional, due to issues like the ACLs discussed in bug
#416705 [1].

[1] https://bugs.gentoo.org/show_bug.cgi?id=416705
-- 
Thanks,
Zac

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Hilco Wijbenga]

On 29 May 2012 15:11, Zac Medico <zmedico@gentoo.org> wrote:
> On 05/29/2012 02:47 PM, Hilco Wijbenga wrote:
>> On 29 May 2012 12:46, Michael Orlitzky <michael@orlitzky.com> wrote:
>>> How about introducing e.g. FEATURES="nouserpriv", and make the current
>>> userpriv behavior the default?
>>
>> rootpriv instead of nouserpriv?
>
> What's the use case for this? Can't we just enable userpriv
> unconditionally, so that it doesn't have to be listed in FEATURES? Note
> that ebuilds will still be able to use RESTRICT=userpriv if necessary.

Absolutely, this was more in response to the "please no reverse logic"
(which I agree with).

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Pacho Ramos]

[-- Attachment #1: Type: text/plain, Size: 1142 bytes --]

El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
> Hi,
> 
> In case you aren't familiar with FEATURES=userpriv, here's the
> description from the make.conf(5) man page:
> 
>   Allow portage to drop root privileges and compile packages as
>   portage:portage without a sandbox (unless usersandbox is also used).
> 
> The rationale for having the separate "usersandbox" setting, to enable
> use of sys-apps/sandbox, is that people who enable userpriv sometimes
> prefer to have sandbox disabled in order to slightly improve
> performance. However, I would recommend to enable usersandbox by
> default, for the purpose of logging sandbox violations.
> 
> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
> privileges during any of the src_* phases that userpriv affects.
> 
> I've been using FEATURES="userpriv usersandbox" for years, and I don't
> remember experiencing any problems because of it, so I think that it
> would be reasonable to have it enabled by default. Objections?

Looks like non important problems arised and, then, these could probably
be enabled by default, no? :)

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

On 07/02/2012 12:48 PM, Pacho Ramos wrote:
> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
>> Hi,
>>
>> In case you aren't familiar with FEATURES=userpriv, here's the
>> description from the make.conf(5) man page:
>>
>>   Allow portage to drop root privileges and compile packages as
>>   portage:portage without a sandbox (unless usersandbox is also used).
>>
>> The rationale for having the separate "usersandbox" setting, to enable
>> use of sys-apps/sandbox, is that people who enable userpriv sometimes
>> prefer to have sandbox disabled in order to slightly improve
>> performance. However, I would recommend to enable usersandbox by
>> default, for the purpose of logging sandbox violations.
>>
>> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
>> privileges during any of the src_* phases that userpriv affects.
>>
>> I've been using FEATURES="userpriv usersandbox" for years, and I don't
>> remember experiencing any problems because of it, so I think that it
>> would be reasonable to have it enabled by default. Objections?
> 
> Looks like non important problems arised and, then, these could probably
> be enabled by default, no? :)

I'm not sure about the best way to handle migration for directories
inside $DISTDIR that are used by live ebuilds, since src_unpack will run
with different privileges when userpriv is enabled.
-- 
Thanks,
Zac

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[vivo75]

Il 02/07/2012 22:01, Zac Medico ha scritto:
> On 07/02/2012 12:48 PM, Pacho Ramos wrote:
>> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
>>> Hi,
>>>
>>> In case you aren't familiar with FEATURES=userpriv, here's the
>>> description from the make.conf(5) man page:
>>>
>>>    Allow portage to drop root privileges and compile packages as
>>>    portage:portage without a sandbox (unless usersandbox is also used).
>>>
>>> The rationale for having the separate "usersandbox" setting, to enable
>>> use of sys-apps/sandbox, is that people who enable userpriv sometimes
>>> prefer to have sandbox disabled in order to slightly improve
>>> performance. However, I would recommend to enable usersandbox by
>>> default, for the purpose of logging sandbox violations.
>>>
>>> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
>>> privileges during any of the src_* phases that userpriv affects.
>>>
>>> I've been using FEATURES="userpriv usersandbox" for years, and I don't
>>> remember experiencing any problems because of it, so I think that it
>>> would be reasonable to have it enabled by default. Objections?
>> Looks like non important problems arised and, then, these could probably
>> be enabled by default, no? :)
> I'm not sure about the best way to handle migration for directories
> inside $DISTDIR that are used by live ebuilds, since src_unpack will run
> with different privileges when userpriv is enabled.
tell the user to chown/remove the files/directories if and when needed, 
unless there is a very good reason (try) to automate it.

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

On 07/02/2012 01:36 PM, vivo75@gmail.com wrote:
> Il 02/07/2012 22:01, Zac Medico ha scritto:
>> On 07/02/2012 12:48 PM, Pacho Ramos wrote:
>>> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
>>>> Hi,
>>>>
>>>> In case you aren't familiar with FEATURES=userpriv, here's the
>>>> description from the make.conf(5) man page:
>>>>
>>>>    Allow portage to drop root privileges and compile packages as
>>>>    portage:portage without a sandbox (unless usersandbox is also used).
>>>>
>>>> The rationale for having the separate "usersandbox" setting, to enable
>>>> use of sys-apps/sandbox, is that people who enable userpriv sometimes
>>>> prefer to have sandbox disabled in order to slightly improve
>>>> performance. However, I would recommend to enable usersandbox by
>>>> default, for the purpose of logging sandbox violations.
>>>>
>>>> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
>>>> privileges during any of the src_* phases that userpriv affects.
>>>>
>>>> I've been using FEATURES="userpriv usersandbox" for years, and I don't
>>>> remember experiencing any problems because of it, so I think that it
>>>> would be reasonable to have it enabled by default. Objections?
>>> Looks like non important problems arised and, then, these could probably
>>> be enabled by default, no? :)
>> I'm not sure about the best way to handle migration for directories
>> inside $DISTDIR that are used by live ebuilds, since src_unpack will run
>> with different privileges when userpriv is enabled.
> tell the user to chown/remove the files/directories if and when needed,

How should we tell them? Elog message, news item, or both?

> unless there is a very good reason (try) to automate it.

I guess something like this might work in pkg_postinst of the portage
ebuild:

  find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
portage:portage

I would only trigger something like this once, when upgrading from a
version that doesn't have userpriv enabled by default.
-- 
Thanks,
Zac

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Pacho Ramos]

[-- Attachment #1: Type: text/plain, Size: 2229 bytes --]

El lun, 02-07-2012 a las 13:45 -0700, Zac Medico escribió:
> On 07/02/2012 01:36 PM, vivo75@gmail.com wrote:
> > Il 02/07/2012 22:01, Zac Medico ha scritto:
> >> On 07/02/2012 12:48 PM, Pacho Ramos wrote:
> >>> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
> >>>> Hi,
> >>>>
> >>>> In case you aren't familiar with FEATURES=userpriv, here's the
> >>>> description from the make.conf(5) man page:
> >>>>
> >>>>    Allow portage to drop root privileges and compile packages as
> >>>>    portage:portage without a sandbox (unless usersandbox is also used).
> >>>>
> >>>> The rationale for having the separate "usersandbox" setting, to enable
> >>>> use of sys-apps/sandbox, is that people who enable userpriv sometimes
> >>>> prefer to have sandbox disabled in order to slightly improve
> >>>> performance. However, I would recommend to enable usersandbox by
> >>>> default, for the purpose of logging sandbox violations.
> >>>>
> >>>> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
> >>>> privileges during any of the src_* phases that userpriv affects.
> >>>>
> >>>> I've been using FEATURES="userpriv usersandbox" for years, and I don't
> >>>> remember experiencing any problems because of it, so I think that it
> >>>> would be reasonable to have it enabled by default. Objections?
> >>> Looks like non important problems arised and, then, these could probably
> >>> be enabled by default, no? :)
> >> I'm not sure about the best way to handle migration for directories
> >> inside $DISTDIR that are used by live ebuilds, since src_unpack will run
> >> with different privileges when userpriv is enabled.
> > tell the user to chown/remove the files/directories if and when needed,
> 
> How should we tell them? Elog message, news item, or both?
> 
> > unless there is a very good reason (try) to automate it.
> 
> I guess something like this might work in pkg_postinst of the portage
> ebuild:
> 
>   find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
> portage:portage
> 
> I would only trigger something like this once, when upgrading from a
> version that doesn't have userpriv enabled by default.

This looks reasonable, I think

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 2586 bytes --]

On Mon, 02 Jul 2012 13:45:26 -0700
Zac Medico <zmedico@gentoo.org> wrote:

> On 07/02/2012 01:36 PM, vivo75@gmail.com wrote:
> > Il 02/07/2012 22:01, Zac Medico ha scritto:
> >> On 07/02/2012 12:48 PM, Pacho Ramos wrote:
> >>> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
> >>>> Hi,
> >>>>
> >>>> In case you aren't familiar with FEATURES=userpriv, here's the
> >>>> description from the make.conf(5) man page:
> >>>>
> >>>>    Allow portage to drop root privileges and compile packages as
> >>>>    portage:portage without a sandbox (unless usersandbox is also
> >>>> used).
> >>>>
> >>>> The rationale for having the separate "usersandbox" setting, to
> >>>> enable use of sys-apps/sandbox, is that people who enable
> >>>> userpriv sometimes prefer to have sandbox disabled in order to
> >>>> slightly improve performance. However, I would recommend to
> >>>> enable usersandbox by default, for the purpose of logging
> >>>> sandbox violations.
> >>>>
> >>>> Note that ebuilds can set RESTRICT="userpriv" if they require
> >>>> superuser privileges during any of the src_* phases that
> >>>> userpriv affects.
> >>>>
> >>>> I've been using FEATURES="userpriv usersandbox" for years, and I
> >>>> don't remember experiencing any problems because of it, so I
> >>>> think that it would be reasonable to have it enabled by default.
> >>>> Objections?
> >>> Looks like non important problems arised and, then, these could
> >>> probably be enabled by default, no? :)
> >> I'm not sure about the best way to handle migration for directories
> >> inside $DISTDIR that are used by live ebuilds, since src_unpack
> >> will run with different privileges when userpriv is enabled.
> > tell the user to chown/remove the files/directories if and when
> > needed,
> 
> How should we tell them? Elog message, news item, or both?

I think this deserves a news item anyway.

> > unless there is a very good reason (try) to automate it.
> 
> I guess something like this might work in pkg_postinst of the portage
> ebuild:
> 
>   find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
> portage:portage

find "$DISTDIR" -maxdepth 1 -type d -uid 0 -exec \
	chown -R portage:portage {} +

> I would only trigger something like this once, when upgrading from a
> version that doesn't have userpriv enabled by default.

This will work only for users who actually keep those in DISTDIR. Some
of them actually redefine E*_STORE_DIR to a more sane location. But
that's probably irrelevant.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[vivo75]

Il 02/07/2012 22:45, Zac Medico ha scritto:
> On 07/02/2012 01:36 PM, vivo75@gmail.com wrote:
>> Il 02/07/2012 22:01, Zac Medico ha scritto:
>>> On 07/02/2012 12:48 PM, Pacho Ramos wrote:
>>>> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
>>>>> Hi,
>>>>>
>>>>> In case you aren't familiar with FEATURES=userpriv, here's the
>>>>> description from the make.conf(5) man page:
>>>>>
>>>>>     Allow portage to drop root privileges and compile packages as
>>>>>     portage:portage without a sandbox (unless usersandbox is also used).
>>>>>
>>>>> The rationale for having the separate "usersandbox" setting, to enable
>>>>> use of sys-apps/sandbox, is that people who enable userpriv sometimes
>>>>> prefer to have sandbox disabled in order to slightly improve
>>>>> performance. However, I would recommend to enable usersandbox by
>>>>> default, for the purpose of logging sandbox violations.
>>>>>
>>>>> Note that ebuilds can set RESTRICT="userpriv" if they require superuser
>>>>> privileges during any of the src_* phases that userpriv affects.
>>>>>
>>>>> I've been using FEATURES="userpriv usersandbox" for years, and I don't
>>>>> remember experiencing any problems because of it, so I think that it
>>>>> would be reasonable to have it enabled by default. Objections?
>>>> Looks like non important problems arised and, then, these could probably
>>>> be enabled by default, no? :)
>>> I'm not sure about the best way to handle migration for directories
>>> inside $DISTDIR that are used by live ebuilds, since src_unpack will run
>>> with different privileges when userpriv is enabled.
>> tell the user to chown/remove the files/directories if and when needed,
> How should we tell them? Elog message, news item, or both?
both seem reasonable, additionally emerge will and should fail when it 
meet a incorrect owned directory, the most sensitive place where to 
output the message is exactly there if possible.
"Failed to update $DIR, check permission and/or correctness, as a last 
resort remove it" something like this, written by someone who speak english.
>> unless there is a very good reason (try) to automate it.
> I guess something like this might work in pkg_postinst of the portage
> ebuild:
>
>    find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
> portage:portage
>
> I would only trigger something like this once, when upgrading from a
> version that doesn't have userpriv enabled by default.
baaaaa, I've totally inverted the logic, it was meant "do _not_  
automate it", even if the chown work flawlessy it become additional 
cruft that will be forever with us.

thanks,
Francesco

Re: Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 489 bytes --]

> 
> I guess something like this might work in pkg_postinst of the portage
> ebuild:
> 
>   find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
> portage:portage
> 
> I would only trigger something like this once, when upgrading from a
> version that doesn't have userpriv enabled by default.

If you run ebuild as user (belonging to group portage), that won't help...
better add a "chmod -R g+w" too...


-- 
Andreas K. Huettel
Gentoo Linux developer
kde, sci, arm, tex, printing

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: Re: Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 668 bytes --]

> > I guess something like this might work in pkg_postinst of the portage
> > 
> > ebuild:
> >   find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
> > 
> > portage:portage
> > 
> > I would only trigger something like this once, when upgrading from a
> > version that doesn't have userpriv enabled by default.
> 
> If you run ebuild as user (belonging to group portage), that won't help...
> better add a "chmod -R g+w" too...

Scratch that. It would not have worked before either, so the user has to do 
something him/herself either way. I guess we dont have to care for this case.

-- 
Andreas K. Huettel
Gentoo Linux developer
kde, sci, arm, tex, printing

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Pacho Ramos]

El mar, 03-07-2012 a las 10:02 +0200, Michał Górny escribió:
> On Mon, 02 Jul 2012 13:45:26 -0700
> Zac Medico <zmedico@gentoo.org> wrote:
> 
> > On 07/02/2012 01:36 PM, vivo75@gmail.com wrote:
> > > Il 02/07/2012 22:01, Zac Medico ha scritto:
> > >> On 07/02/2012 12:48 PM, Pacho Ramos wrote:
> > >>> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
> > >>>> Hi,
> > >>>>
> > >>>> In case you aren't familiar with FEATURES=userpriv, here's the
> > >>>> description from the make.conf(5) man page:
> > >>>>
> > >>>>    Allow portage to drop root privileges and compile packages as
> > >>>>    portage:portage without a sandbox (unless usersandbox is also
> > >>>> used).
> > >>>>
> > >>>> The rationale for having the separate "usersandbox" setting, to
> > >>>> enable use of sys-apps/sandbox, is that people who enable
> > >>>> userpriv sometimes prefer to have sandbox disabled in order to
> > >>>> slightly improve performance. However, I would recommend to
> > >>>> enable usersandbox by default, for the purpose of logging
> > >>>> sandbox violations.
> > >>>>
> > >>>> Note that ebuilds can set RESTRICT="userpriv" if they require
> > >>>> superuser privileges during any of the src_* phases that
> > >>>> userpriv affects.
> > >>>>
> > >>>> I've been using FEATURES="userpriv usersandbox" for years, and I
> > >>>> don't remember experiencing any problems because of it, so I
> > >>>> think that it would be reasonable to have it enabled by default.
> > >>>> Objections?
> > >>> Looks like non important problems arised and, then, these could
> > >>> probably be enabled by default, no? :)
> > >> I'm not sure about the best way to handle migration for directories
> > >> inside $DISTDIR that are used by live ebuilds, since src_unpack
> > >> will run with different privileges when userpriv is enabled.
> > > tell the user to chown/remove the files/directories if and when
> > > needed,
> > 
> > How should we tell them? Elog message, news item, or both?
> 
> I think this deserves a news item anyway.
> 
> > > unless there is a very good reason (try) to automate it.
> > 
> > I guess something like this might work in pkg_postinst of the portage
> > ebuild:
> > 
> >   find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
> > portage:portage
> 
> find "$DISTDIR" -maxdepth 1 -type d -uid 0 -exec \
> 	chown -R portage:portage {} +
> 
> > I would only trigger something like this once, when upgrading from a
> > version that doesn't have userpriv enabled by default.
> 
> This will work only for users who actually keep those in DISTDIR. Some
> of them actually redefine E*_STORE_DIR to a more sane location. But
> that's probably irrelevant.
> 

Then, is there any other blocker? (apart of the needing of add a news
item)

Thanks :)

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Zac Medico]

On 07/21/2013 03:53 AM, Pacho Ramos wrote:
> El mar, 03-07-2012 a las 10:02 +0200, Michał Górny escribió:
>> On Mon, 02 Jul 2012 13:45:26 -0700
>> Zac Medico <zmedico@gentoo.org> wrote:
>>
>>> On 07/02/2012 01:36 PM, vivo75@gmail.com wrote:
>>>> Il 02/07/2012 22:01, Zac Medico ha scritto:
>>>>> On 07/02/2012 12:48 PM, Pacho Ramos wrote:
>>>>>> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
>>>>>>> Hi,
>>>>>>>
>>>>>>> In case you aren't familiar with FEATURES=userpriv, here's the
>>>>>>> description from the make.conf(5) man page:
>>>>>>>
>>>>>>>    Allow portage to drop root privileges and compile packages as
>>>>>>>    portage:portage without a sandbox (unless usersandbox is also
>>>>>>> used).
>>>>>>>
>>>>>>> The rationale for having the separate "usersandbox" setting, to
>>>>>>> enable use of sys-apps/sandbox, is that people who enable
>>>>>>> userpriv sometimes prefer to have sandbox disabled in order to
>>>>>>> slightly improve performance. However, I would recommend to
>>>>>>> enable usersandbox by default, for the purpose of logging
>>>>>>> sandbox violations.
>>>>>>>
>>>>>>> Note that ebuilds can set RESTRICT="userpriv" if they require
>>>>>>> superuser privileges during any of the src_* phases that
>>>>>>> userpriv affects.
>>>>>>>
>>>>>>> I've been using FEATURES="userpriv usersandbox" for years, and I
>>>>>>> don't remember experiencing any problems because of it, so I
>>>>>>> think that it would be reasonable to have it enabled by default.
>>>>>>> Objections?
>>>>>> Looks like non important problems arised and, then, these could
>>>>>> probably be enabled by default, no? :)
>>>>> I'm not sure about the best way to handle migration for directories
>>>>> inside $DISTDIR that are used by live ebuilds, since src_unpack
>>>>> will run with different privileges when userpriv is enabled.
>>>> tell the user to chown/remove the files/directories if and when
>>>> needed,
>>>
>>> How should we tell them? Elog message, news item, or both?
>>
>> I think this deserves a news item anyway.
>>
>>>> unless there is a very good reason (try) to automate it.
>>>
>>> I guess something like this might work in pkg_postinst of the portage
>>> ebuild:
>>>
>>>   find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
>>> portage:portage
>>
>> find "$DISTDIR" -maxdepth 1 -type d -uid 0 -exec \
>> 	chown -R portage:portage {} +
>>
>>> I would only trigger something like this once, when upgrading from a
>>> version that doesn't have userpriv enabled by default.
>>
>> This will work only for users who actually keep those in DISTDIR. Some
>> of them actually redefine E*_STORE_DIR to a more sane location. But
>> that's probably irrelevant.
>>
> 
> Then, is there any other blocker? (apart of the needing of add a news
> item)
> 
> Thanks :)
> 

I can't think of anything else. I've filed this bug:

  https://bugs.gentoo.org/show_bug.cgi?id=477664

-- 
Thanks,
Zac

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Alex Xu]

[-- Attachment #1: Type: text/plain, Size: 3306 bytes --]

On 21/07/13 02:25 PM, Zac Medico wrote:
> On 07/21/2013 03:53 AM, Pacho Ramos wrote:
>> El mar, 03-07-2012 a las 10:02 +0200, Michał Górny escribió:
>>> On Mon, 02 Jul 2012 13:45:26 -0700
>>> Zac Medico <zmedico@gentoo.org> wrote:
>>>
>>>> On 07/02/2012 01:36 PM, vivo75@gmail.com wrote:
>>>>> Il 02/07/2012 22:01, Zac Medico ha scritto:
>>>>>> On 07/02/2012 12:48 PM, Pacho Ramos wrote:
>>>>>>> El lun, 28-05-2012 a las 14:34 -0700, Zac Medico escribió:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> In case you aren't familiar with FEATURES=userpriv, here's the
>>>>>>>> description from the make.conf(5) man page:
>>>>>>>>
>>>>>>>>    Allow portage to drop root privileges and compile packages as
>>>>>>>>    portage:portage without a sandbox (unless usersandbox is also
>>>>>>>> used).
>>>>>>>>
>>>>>>>> The rationale for having the separate "usersandbox" setting, to
>>>>>>>> enable use of sys-apps/sandbox, is that people who enable
>>>>>>>> userpriv sometimes prefer to have sandbox disabled in order to
>>>>>>>> slightly improve performance. However, I would recommend to
>>>>>>>> enable usersandbox by default, for the purpose of logging
>>>>>>>> sandbox violations.
>>>>>>>>
>>>>>>>> Note that ebuilds can set RESTRICT="userpriv" if they require
>>>>>>>> superuser privileges during any of the src_* phases that
>>>>>>>> userpriv affects.
>>>>>>>>
>>>>>>>> I've been using FEATURES="userpriv usersandbox" for years, and I
>>>>>>>> don't remember experiencing any problems because of it, so I
>>>>>>>> think that it would be reasonable to have it enabled by default.
>>>>>>>> Objections?
>>>>>>> Looks like non important problems arised and, then, these could
>>>>>>> probably be enabled by default, no? :)
>>>>>> I'm not sure about the best way to handle migration for directories
>>>>>> inside $DISTDIR that are used by live ebuilds, since src_unpack
>>>>>> will run with different privileges when userpriv is enabled.
>>>>> tell the user to chown/remove the files/directories if and when
>>>>> needed,
>>>>
>>>> How should we tell them? Elog message, news item, or both?
>>>
>>> I think this deserves a news item anyway.
>>>
>>>>> unless there is a very good reason (try) to automate it.
>>>>
>>>> I guess something like this might work in pkg_postinst of the portage
>>>> ebuild:
>>>>
>>>>   find "$DISTDIR" -maxdepth 1 -type d -uid 0 | xargs chown -R
>>>> portage:portage
>>>
>>> find "$DISTDIR" -maxdepth 1 -type d -uid 0 -exec \
>>> 	chown -R portage:portage {} +
>>>
>>>> I would only trigger something like this once, when upgrading from a
>>>> version that doesn't have userpriv enabled by default.
>>>
>>> This will work only for users who actually keep those in DISTDIR. Some
>>> of them actually redefine E*_STORE_DIR to a more sane location. But
>>> that's probably irrelevant.
>>>
>>
>> Then, is there any other blocker? (apart of the needing of add a news
>> item)
>>
>> Thanks :)
>>
> 
> I can't think of anything else. I've filed this bug:
> 
>   https://bugs.gentoo.org/show_bug.cgi?id=477664
> 

userpriv and usersandbox don't work in pypy because os.setgroups isn't
implemented there.

I had a go at it a while back, but the complete and utter lack of any
documentation whatsoever... kinda threw me off.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [gentoo-dev] RFC: Enable FEATURES="userpriv usersandbox" by default?

[Mike Gilbert]

On Sun, Jul 21, 2013 at 2:30 PM, Alex Xu <alex_y_xu@yahoo.ca> wrote:
> userpriv and usersandbox don't work in pypy because os.setgroups isn't
> implemented there.
>
> I had a go at it a while back, but the complete and utter lack of any
> documentation whatsoever... kinda threw me off.
>

I don't think we need to tailor the default configuration to meet the
limitations imposed by an experimental python interpreter.

If it can be worked around, great, but it should not stop us from
enabling userpriv and usersandbox by default.