app-editors/vim— it does not refer to a specific version. The terms 'ebuild' or 'package version' are used when this meaning is intended. This distinction is important.
Most ebuilds specify a
KEYWORDS variable. This variable is used to
indicate the suitability and stability of both the package and the ebuild on
each given arch (
KEYWORDS entry might look like:
KEYWORDS="-ia64 ~mips ~ppc sparc x86 ~ppc-macos"
The different levels of keyword are:
- Both the package version and the ebuild are widely tested, known to work and not have any serious issues on the indicated platform.
The package version and the ebuild are believed to work and do not have any
known serious bugs, but more testing is required before the package version
is considered suitable for
- No keyword ("unkeyworded")
If a package has no keyword for a given arch, it means it is not known
whether the package will work, or that insufficient testing has occurred for
- The package version will not work on the arch. This could be caused by badly written code (for example, non-64-bit or endian clean code), relying upon particular hardware (for example, a BIOS querying tool would not work on non-BIOS architectures) or binary only packages.
-* keyword is special. It is used to indicate package versions which are
not worth trying to test on unlisted arches. For example, a binary-only package
which is only supported upstream on
x86 might use:
KEYWORDS="-* ppc x86"
This is different in implication from
"ppc x86" — the former implies
that it will not work on other arches, whereas the latter implies that it has
not been tested.
Do not use the
~* special keywords in ebuilds.
An ebuild must not depend upon any package that is of a lower keyword level
than itself. For example, if
foo-1.2 depends upon
foo-1.2 must not be marked stable on
bar-1.2 is also stabilised.
You may assume that if a user accepts
~arch for a given arch then they also
For optional dependencies, all possible dependencies must satisfy the above.
Note that certain
USE flags can be forcibly disabled on a per-profile basis
— talk to the arch teams if you require this. For either-or dependencies, at
least one of the options must be of equal or better visibility than the
package in question.
package.mask file can be used to 'hard mask' individual or groups of
ebuilds. This should be used for testing ebuilds or beta releases of software,
and may also be used if a package has serious compatibility problems. Packages
which are not hard masked must not have a dependency upon hard masked
The only time it is acceptable for a user to see the
Possibly a DEPEND
problem error message is if they have manually changed visibility levels for a
package (for example, through
/etc/portage/) and have missed a dependency.
You should never commit a change which could cause this error to appear on a
~arch("testing") only upon architectures for which the committing developer has tested. If proxying commits for a non-developer, please ensure they have some relationship with the relevant arch teams, or — better yet — ask them to file a keywording bug instead for non-
Do not assume that your package works on all architectures. Do not
assume that user submitted ebuilds will have correct
KEYWORDS — chances are
they just copied from somewhere else. Do not assume that upstream's
'supported architectures' list is correct. Do not assume that because your
code is written in Perl / Python / Java / whatever that it will run on other
arches (there is at least one case of a
vim script which only worked on
Note that most (non-
x86) arches expect you to be on the
arch team and bugzilla alias if you are committing packages with keywords for
that arch, and may have additional requirements of which you should be aware
mips, for example, there are multiple ABIs and byte orders to
consider — a package working on your
o32 box may not work on
n32). Contact the individual arch teams for details.
It's important to note that alternative arches (like
ppc*) are mainly understaffed
arches, some of them are slow, they have more basic problems and have a small
userbase. Just file bugs for these architectures when a package is going to be
a dependency of a package already keyworded.
Do not commit straight to
When upgrading, drop all existing keywords from
("testing"), and leave any existing
~arch keywords intact. This must be
done even if you think you're just making a trivial fix — there have
been several examples of the stable tree getting broken this way.
This also applies to revision bumps, not just to upstream version changes.
If a new version introduces new dependencies which are not available on some
architectures, then you should file a bug or ask on IRC before you upgrade the
ebuild. If you really need to get the ebuild added in a hurry, for example,
for a security fix, then you should drop any
KEYWORDS which are causing
problems and CC the relevant architectures on the bug — you must file
new bug to the architecture in question regarding this if no bug is
If there are no new dependencies, do not remove keywords if your commit fails
pkgdev — instead please commit with
repoman -I and file a bug to the broken architecture, noting it in your
git commit message.
Stabilization, i.e., moving an ebuild from
~arch ("testing") to
arch ("stable"), is done by the relevant architecture teams. If you have
access to exotic hardware but are not on the arch teams, you may wish to make
individual arrangements — the arch teams are happy for help, so long as they
know what is going on.
In order to request stabilization of an ebuild, file a bug to the package's
maintainer (which may be yourself) in the Stabilization component, and list
all secondary maintainers in the bug's CC. When the maintainers consider the
ebuild to be ready for stabilization, they will add the relevant architecture
teams to the CC list. They can fill the package
list field, add the
CC-ARCHES keyword, and let
automatically add arch teams to CC (preferred), or do this manually.
That way, teams can remove themselves from the list when they are done, giving
a clear indication of which teams still remain to stabilize a package.
It is strongly
NATTkA be used to file keywording or
stabilization bugs to ensure that arches are not accidentally forgot about;
the automation removes an often error-prone step in the arch-testing process.
Ask in #gentoo-dev or on the gentoo-dev mailing list if you need help.
NATTkA will set the
sanity-check field on Bugzilla to either
+ depending on whether the package list results in a
consistent dependency graph. You may need to add more packages or adjust the
arches listed for each package based on the output the bot posts as a comment
on the bug. Please note that arch teams may not process or notice bugs with a
- sanity-check, so please fix this if it occurs — or ask for
help to do so.
nattka sanity-check BUGwhere
BUGis the bug number you're having an issue with in order to debug the issue to receive verbose output.
You may wish to read NATTkA's
documentation for help on package list syntax or other information about
the tool. The documentation details syntax which can be used, such as
to represent "all previously stable arches", or
^ to copy the arches
from the line(s) above.
For an ebuild to move to stable, the following guidelines must be met (see GLEP 40 for further details):
The ebuild has spent a reasonable amount of time in
~archfirst. Thirty days is the usual figure, although this is clearly only a guideline. For critical packages, a much longer duration is expected. For small packages that have only minor changes between versions, a shorter period is sometimes appropriate.
The ebuild must not have any non-
- The package version (and the ebuild) must not have any severe outstanding bugs or issues.
- The package version must be widely tested.
- If the package is a library, it should be known not to break any package which depends upon it.
For security fixes, the "reasonable amount of time" guideline may be relaxed. See the Vulnerability Treatment Policy.
x86: If you are the maintainer of a package and own the respective
x86hardware, you can do your own testing (stabilization and keywording) of your packages; as long as it is not a core system set dependency. Note that it is acceptable to test
x86using a specialized environment on
arm64: These teams have a strict policy on passing testsuites. You must run test suites for packages with e.g.
FEATURES=testfor Portage. Keywording or stabilizing packages which fail these tests is avoided but acceptable if the test suite is fragile or failures are known and reported. It is preferred to have to keyword or stabilize more packages on these arches if needed for e.g. test dependencies rather than masking or disabling tests here.
sparc: You must have prior permission from the arch lead. Usually we expect you to be on the sparc alias for QA reasons, although other arrangements can be made if you will only be working with a small group of packages.
alpha: Maintainers may keyword their own packages but are reminded to inform the Alpha team if they can help out with testing and keywording packages so the team can keep an eye out for possible keywording mistakes.
Exotic architectures (like
are short on help, so it's best if you avoid opening bugs for stabilization
of new packages for them, unless it is absolutely necessary (e.g., a reverse
dependency for your package).
Some architectures (like
riscv) do not maintain a stable
keyword, so packages are not to be marked stable for one of these architectures.
If you maintain an architecture-independent package (data files, icons, pure
Python, ...), then you may request that your package be stabilized on all arches
at once. To do this — when you are filing the stabilization bug — please
add the keyword
ALLARCHES and CC the arches that you would like to
If your package is architecture-independent, you should add the
<stabilize-allarches/> tag to metadata.xml. This allows consistency
in future stabilizations and saves arch teams considerable work.
The arch teams, when encountering the
ALLARCHES keyword, should perform
their usual set of tests on a single convenient architecture. Then, if
everything works, stabilize not only the arch that was used during testing,
but also all of the other arches in CC on the bug. Afterwards, the CC field can
be cleared and the bug closed if appropriate.
Note that first-time
ALLARCHES stabilizations are done "as normal" —
i.e. all arch teams test individually as if it was not set. This nuance in the
procedure is part of why developers should not manually set
bugs, but instead let
NATTkA set it automatically based on metadata.xml.
When removing ebuild, ensure that you do not remove the most recent version at any given keyword level on any profile. The aim here is:
Never to force a downgrade. (Exception: occasionally you really do want to
force a downgrade, for example if the newly committed
foo-1.3turns out to be badly broken and that making everyone downgrade to
foo-1.2is the better option. This is rare.)
- Do not break any existing dependencies.
If you would like a particular package version moved to stable on certain arches so that you can tidy up, file a bug.