[gentoo-dev] some questions about using subslots and EAPI5

[gentoo-dev] some questions about using subslots and EAPI5

[Alexandre Rostovtsev]

Suppose I am upgrading an existing ebuild for dev-libs/libfoo-2.5 to
EAPI5 and adding a subslot, e.g. SLOT="2/5". Should I then add
"slotmove =dev-libs/libfoo-2.5 2 2/5" in profiles/updates? Or should it
be "slotmove =dev-libs/libfoo-2.5 2/2 2/5" (since the PMS states that
"When the sub-slot part is omitted from the SLOT definition, the package
is considered to have an implicit sub-slot which is equal to the regular
slot")? Will this update even do what one would intuitively expect it to
do? Might it break older versions of portage on emerge --sync?

Is there any difference between RDEPEND="dev-libs/libfoo:2" and
RDEPEND="dev-libs/libfoo:2=" in an EAPI5 ebuild? Will both variants
force a rebuild when the subslot of libfoo:2 changes?

Can I use EAPI5 in a stable ebuild, as long as it's only marked stable
on arches that have >=portage-2.1.11.31 stabilized? Or is EAPI5 still
~arch-only until the council declares otherwise?

Re: [gentoo-dev] some questions about using subslots and EAPI5

[Ian Stakenvicius]

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

On 05/12/12 04:42 AM, Alexandre Rostovtsev wrote:
> Suppose I am upgrading an existing ebuild for dev-libs/libfoo-2.5
> to EAPI5 and adding a subslot, e.g. SLOT="2/5". Should I then add 
> "slotmove =dev-libs/libfoo-2.5 2 2/5" in profiles/updates? Or
> should it be "slotmove =dev-libs/libfoo-2.5 2/2 2/5" (since the PMS
> states that "When the sub-slot part is omitted from the SLOT
> definition, the package is considered to have an implicit sub-slot
> which is equal to the regular slot")? Will this update even do what
> one would intuitively expect it to do? Might it break older
> versions of portage on emerge --sync?
> 
> Is there any difference between RDEPEND="dev-libs/libfoo:2" and 
> RDEPEND="dev-libs/libfoo:2=" in an EAPI5 ebuild? Will both
> variants force a rebuild when the subslot of libfoo:2 changes?
> 
> Can I use EAPI5 in a stable ebuild, as long as it's only marked
> stable on arches that have >=portage-2.1.11.31 stabilized? Or is
> EAPI5 still ~arch-only until the council declares otherwise?
> 
> 

1.  slotmove doesn't work for sub-slots, so no.

1a.  the implicit subslot text essentially means that if you only set
"SLOT=2", then it rolls itself out to SLOT=2/2 in the background
(vdb).  This doesn't affect how you manage the package yourself,
afaik.  Note also that I haven't actually examined this so i can't
confirm that SLOT=2/2 would be the actual value written in the vdb.

2.  dev-libs/libfoo:2 is the same as dev-libs/libfoo:2* , which
differs from dev-libs/libfoo:2= .  Essentially you would use :2= when
you want this package to remain solely dependent on slot 2, but
rebuild when sub-slot changes occur in libfoo within slot 2.  Without
the = slot-operator, rebuilds will not be triggered.

3.  EAPI5 can be used in stable ebuilds as long as >=portage-2.1.11.20
is stable on all of the stable keywords in that ebuild.  Council was
clear that as soon as an EAPI5-capable portage was stable it could be
used in stable ebuilds.

That said, it is recommended that you revbump for EAPI5; it is not
strictly necessary, but without the revbump users with that version
already installed will not receive the benefits of the
sub-slot/slot-operator, since the sub-slot has to be written to the
vdb for both the package and all rdeps with slot-operator atoms.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)

iF4EAREIAAYFAlC/TbgACgkQ2ugaI38ACPC23wD/T5fEzWbJJs/VptIHE37l6F1M
6ntTByLRkHcVxiy3CJcBALvVvaC7Y0Y5XTzCmdKXvGiuqRKhKMu/+8Iq8SuKZ82T
=a+HV
-----END PGP SIGNATURE-----

Re: [gentoo-dev] some questions about using subslots and EAPI5

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 765 bytes --]

On Wed, 05 Dec 2012 04:42:23 -0500
Alexandre Rostovtsev <tetromino@gentoo.org> wrote:
> Suppose I am upgrading an existing ebuild for dev-libs/libfoo-2.5 to
> EAPI5 and adding a subslot, e.g. SLOT="2/5". Should I then add
> "slotmove =dev-libs/libfoo-2.5 2 2/5" in profiles/updates? Or should
> it be "slotmove =dev-libs/libfoo-2.5 2/2 2/5" (since the PMS states
> that "When the sub-slot part is omitted from the SLOT definition, the
> package is considered to have an implicit sub-slot which is equal to
> the regular slot")? Will this update even do what one would
> intuitively expect it to do? Might it break older versions of portage
> on emerge --sync?

You can't put subslots in updates, since it's not EAPI controlled.

-- 
Ciaran McCreesh

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

Re: [gentoo-dev] some questions about using subslots and EAPI5

[Ciaran McCreesh]

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

On Wed, 05 Dec 2012 08:35:52 -0500
Ian Stakenvicius <axs@gentoo.org> wrote:
> That said, it is recommended that you revbump for EAPI5; it is not
> strictly necessary, but without the revbump users with that version
> already installed will not receive the benefits of the
> sub-slot/slot-operator, since the sub-slot has to be written to the
> vdb for both the package and all rdeps with slot-operator atoms.

If subslots are involved, a revbump is necessary when switching to
EAPI 5.

- -- 
Ciaran McCreesh
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)

iEYEARECAAYFAlC/YuUACgkQ96zL6DUtXhGRXACg4tDMvPeFXehpLY8sK2Y9QImH
yS0AoJ2sjC7xBBniacnTLCToRnXjSuCz
=DAjV
-----END PGP SIGNATURE-----

Re: [gentoo-dev] some questions about using subslots and EAPI5

[Ian Stakenvicius]

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

On 05/12/12 10:06 AM, Ciaran McCreesh wrote:
> On Wed, 05 Dec 2012 08:35:52 -0500 Ian Stakenvicius
> <axs@gentoo.org> wrote:
>> That said, it is recommended that you revbump for EAPI5; it is
>> not strictly necessary, but without the revbump users with that
>> version already installed will not receive the benefits of the 
>> sub-slot/slot-operator, since the sub-slot has to be written to
>> the vdb for both the package and all rdeps with slot-operator
>> atoms.
> 
> If subslots are involved, a revbump is necessary when switching to 
> EAPI 5.
> 

Only if you want existing users to get the update.  Otherwise the
subslot will just be ignored until that package is rebuilt locally.
If no other changes occur in the ebuild, this is not necessarily a bad
thing, given that slot-operator rebuilds are not -strictly- necessary
(ie the lack of them is not a bug and will not cause a failure -- it
is perfectly valid for an EAPI5 ebuild to have a slot-operator dep on
an EAPI4 dependency, for instance--useless, but still valid.  the case
here is identical to that).

New emerges of the package would still immediately receive the
benefits of the EAPI5 sub-slot.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)

iF4EAREIAAYFAlC/dA8ACgkQ2ugaI38ACPDh4gEAj3S8jHGSS/3qASCpdFgGa7Bl
JzEFXF9QvXKVzD/ZnEkA/0lvj1Ao4/MgcF7kYv22+1GeTCdvtpoWerV6WLdooyNW
=LaQQ
-----END PGP SIGNATURE-----