[gentoo-dev] EAPI 8 draft for review

2021-05-27 Thread Ulrich Mueller
The first draft of EAPI 8 has been posted to the gentoo-pms mailing
list for review:
https://archives.gentoo.org/gentoo-pms/message/e3a7c931ea369e84d81ee70d2fe9802c

> Here is the series of EAPI 8 patches for review. They include the
> pre-approved items from the 2020-11-08 Council meeting, with two
> modifications:

> - "Empty working directory in pkg_* phase functions" added
> - "Variant of || ( ) with defined runtime behaviour" dropped,
>   because the implementation is not ready

> The complete list of features is:

> - Less strict naming rules for files in updates directory
> - Bash version is 5.0
> - Selective fetch/mirror restriction
> - IDEPEND
> - Empty working directory in pkg_* phase functions
> - Different src_prepare implementation
> - PROPERTIES and RESTRICT accumulated across eclasses
> - useq banned
> - hasv and hasq banned
> - econf adds --datarootdir
> - econf adds --disable-static
> - dosym can create relative paths
> - insopts no longer affects doconfd, doenvd and doheader
> - exeopts no longer affects doinitd
> - usev supports an optional second argument
> - unpack no longer supports .7z, .rar, .lha

> The rendered version of the spec can be found:
> PDF:  https://dev.gentoo.org/~ulm/pms/8-draft/pms.pdf
> HTML: https://dev.gentoo.org/~ulm/pms/8-draft/pms.html

> Status of implementation in Portage and Pkgcore can be traced here:
> https://wiki.gentoo.org/wiki/Future_EAPI/EAPI_8_tentative_features

> Thanks to Michał Górny for contributing patches for some of the more
> complicated features.

> Ulrich




signature.asc
Description: PGP signature


Re: [gentoo-dev] News item: sys-libs/db old SLOT removal

2021-05-27 Thread David Seifert
On Wed, 2021-05-26 at 16:09 -0700, Patrick McLean wrote:
> On Thu, 27 May 2021 00:41:23 +0200
> David Seifert  wrote:
> 
> > The old Berkeley DB slots need to go at this point. The Base Project
> > has
> > decided to consider BDB a deprecated database backend, and we'll
> > slowly
> > be working towards a (possibly) BDB-free ::gentoo some time in the
> > long-
> > term future.
> 
> I think we should keep at least one non AGPLv3 berkdb in the tree as
> long
> as we have any packages that unconditionally depend on it. June 1st is
> too short a time frame for masking pre AGPLv3 berkdb versions. I think
> it
> is reasonable to fix packages that either force berkdb USE flags on in
> their
> deps, or have a hard dep (either by updating/fixing or last-rite).

Have you looked at which versions will remain in the tree? Slots 4.8 and
5.3 are non-AGPL, and given the number of projects unlikely to support
>=6, it's unlikely that 5.3 will ever go away.

> > Other distros such as Fedora have started a gradual phase-out of
> > Berkeley DB too, given Oracle's strong-armed approach to community
> > input and their arguably hostile switch to the AGPLv3
> > (  https://fedoraproject.org/wiki/Changes/Libdb_deprecated).
> > Furthermore,
> > Oracle is known to remove critical features from BDB in patch
> > releases,
> > such as the removal of the client-server architecture and the SQL
> > API
> > between 18.1.32 and 18.1.40.
> 
> Gradual phase-out is also the approach we should take. Dropping non
> AGPLv3 version sort of immediately forces the issue for users that
> can't or won't accept that license.

Again, the premise of this argument is wrong (see above), so I won't go
into more detail here.

> > To this end, we will also be removing USE="berkdb" from
> > profiles/default/linux/make.defaults. If you implicitly depend on
> > profiles enabling optional use of sys-libs/db, you will need to
> > enable
> > this USE flag yourself, beginning 1st June.
> > 
> > From here on, you should be working under the assumption that the
> > sys-libs/db package will be gone from the Gentoo repository within
> > **two years** from the time of this news item. If you depend on BDB
> > in
> > a production environment, we strongly suggest you move to one of the
> > modern replacements, such as GDBM, SQLite or LMDB.
> 
> This makes sense for end users, but we should fix ::gentoo before we
> force it on our users.

This is a message to users, obviously we won't just randomly delete the
ebuilds unless ::gentoo is fixed. This is about managing expectations
for users, not an internal TODO list for gentoo developers.