[gentoo-dev] [PATCH 00/10] GLEP 72 (arches.desc) revival

[gentoo-dev] [PATCH 00/10] GLEP 72 (arches.desc) revival

[Michał Górny]

Hi,

I've just found another use case for arches.desc, so I'd like to revive
dilfridge's GLEP.  I tried to keep the changes to minimum, while
addressing earlier concerns.

Besides wording changes and clarifications, the following significant
changes were made:

1. 'Broken' status was removed, as it is redundant to profile status.

2. State names were changed from 'testing' and 'unstable' to 'degraded'
   (broken stable tree) and 'testing' (pure ~arch) to avoid confusion.

3. Overlays (slave repos) now override parent's arch state rather than
   using 'the strictest setting'.


Michał Górny (10):
  glep-0072: Use 'profiles subdirectory' to clarify
  glep-0072: Remove redundant 'broken' status
  glep-0072: Rename bad depgraph state to 'degraded'
  glep-0072: Use 'testing' for pure ~arch
  glep-0072: Remove weird third column from example
  glep-0072: Update initial values
  glep-0072: Combine and amend description of states
  glep-0072: Explicitly cover file not existing case
  glep-0072: Move 'overlays' to spec, and change behavior
  glep-0072: Update metadata

 glep-0072.rst | 121 ++++++++++++++++++++++----------------------------
 1 file changed, 54 insertions(+), 67 deletions(-)

-- 
2.26.0

[gentoo-dev] [PATCH 01/10] glep-0072: Use 'profiles subdirectory' to clarify

[Michał Górny]

Technically speaking, the 'profiles' directory is used for two distinct
purposes: to hold profiles, and to hold some global repository
configuration.  Make it clear that we're talking about the latter
to avoid confusion that arches.desc could be used in actual profiles.

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index 61f9c16..04f7f2d 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -82,9 +82,9 @@ Specifications for profiles/arches.desc
 File and format
 ---------------
 
-In the main profiles directory, a file ``arches.desc`` is added. Name
-and location are chosen in analogy to the existing ``profiles.desc`` file.
-The format of ``arches.desc`` is as follows:
+In the ``profiles`` subdirectory of the repository, a file ``arches.desc``
+is added. Name and location are chosen in analogy to the existing
+``profiles.desc`` file.  The format of ``arches.desc`` is as follows:
 
 Every ``#`` starts a comment; the character and the rest of the line
 are ignored.  Every blank line is ignored. Otherwise the file consists of two
-- 
2.26.0

Re: [gentoo-dev] [PATCH 01/10] glep-0072: Use 'profiles subdirectory' to clarify

[Ulrich Mueller]

[-- Attachment #1: Type: text/plain, Size: 654 bytes --]

>>>>> On Fri, 10 Apr 2020, Michał Górny wrote:
 
> -In the main profiles directory, a file ``arches.desc`` is added. Name
> -and location are chosen in analogy to the existing ``profiles.desc`` file.
> -The format of ``arches.desc`` is as follows:
> +In the ``profiles`` subdirectory of the repository, a file ``arches.desc``

PMS calls it the "profiles directory" and I would suggest to stay with
that terminology, rather than inventing yet another one. IMHO the
previous wording was just fine.

> +is added. Name and location are chosen in analogy to the existing
> +``profiles.desc`` file.  The format of ``arches.desc`` is as follows:

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 507 bytes --]

Re: [gentoo-dev] [PATCH 01/10] glep-0072: Use 'profiles subdirectory' to clarify

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 811 bytes --]

On Fri, 2020-04-10 at 12:14 +0200, Ulrich Mueller wrote:
> > > > > > On Fri, 10 Apr 2020, Michał Górny wrote:
>  
> > -In the main profiles directory, a file ``arches.desc`` is added. Name
> > -and location are chosen in analogy to the existing ``profiles.desc`` file.
> > -The format of ``arches.desc`` is as follows:
> > +In the ``profiles`` subdirectory of the repository, a file ``arches.desc``
> 
> PMS calls it the "profiles directory" and I would suggest to stay with
> that terminology, rather than inventing yet another one. IMHO the
> previous wording was just fine.

Ok, removed this patch.

> 
> > +is added. Name and location are chosen in analogy to the existing
> > +``profiles.desc`` file.  The format of ``arches.desc`` is as follows:

-- 
Best regards,
Michał Górny


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 618 bytes --]

[gentoo-dev] [PATCH 02/10] glep-0072: Remove redundant 'broken' status

[Michał Górny]

This is really no different from marking the profiles exp, and there
seems no value in having this controlled in two places.

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index 04f7f2d..a0aa5d5 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -91,8 +91,7 @@ are ignored.  Every blank line is ignored. Otherwise the file consists of two
 whitespace-separated columns:
 
 - first column: architecture name (keyword)
-- second column: one of the four values ``stable``, ``testing``, ``unstable``,
-  ``broken``
+- second column: one of the three values ``stable``, ``testing``, ``unstable``
 
 Additional columns are ignored to allow for future revisions of this document.
 
@@ -146,11 +145,6 @@ as an error and aborts. Consistency is only tested for ``~arch``.
 Which profiles of the arch are tested is still controlled by profiles.desc
 (and ``-d`` / ``-e`` switches).
 
-broken
-~~~~~~
-Repoman is not testing any profiles of this architecture, irrespective
-of the settings in profiles.desc.
-
 Meaning of the values for other tools
 -------------------------------------
 
-- 
2.26.0

[gentoo-dev] [PATCH 03/10] glep-0072: Rename bad depgraph state to 'degraded'

[Michał Górny]

In Gentoo terms, 'testing' and 'unstable' are mostly synonymous,
so using the two names for different purposes is confusing.  Use
'degraded' instead.

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index a0aa5d5..6fdee6b 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -56,7 +56,7 @@ a) An architecture loses its stable status (imagine c128), but
    about a broken stable dependency tree. If we do that, repoman does however
    also not check ~c128 consistency, meaning that the ~c128 dependency tree
    will soon be broken as well due to negligence.  Given arches.conf as
-   described below, one could set the architecture c128 to "testing" status
+   described below, one could set the architecture c128 to "degraded" status
    and keep stable profiles. This results in stable keywords being ignored,
    but consistency of the ~c128 dependency tree is still enforced.
 
@@ -66,7 +66,7 @@ b) An architecture prepares for becoming a stable architecture (think arm64).
    as the stable dependency tree is not complete yet, the profiles need to be
    set to dev/exp, and again this brings the danger of the ~arm64 dependency
    tree getting inadvertently broken. Again the combination of setting the
-   architecture to "testing" in arches.desc and profiles to stable helps.
+   architecture to "degraded" in arches.desc and profiles to stable helps.
 
 Finally, at the moment the "semi-official" algorithm to figure out if an
 architecture is stable in the colloquial sense (e.g., requires stabilization
@@ -91,7 +91,7 @@ are ignored.  Every blank line is ignored. Otherwise the file consists of two
 whitespace-separated columns:
 
 - first column: architecture name (keyword)
-- second column: one of the three values ``stable``, ``testing``, ``unstable``
+- second column: one of the three values ``stable``, ``degraded``, ``unstable``
 
 Additional columns are ignored to allow for future revisions of this document.
 
@@ -101,14 +101,14 @@ An example arches.desc file might look as follows::
     amd64   stable
     x86     stable     # not for long
 
-    mips    testing
+    sparc   degraded
     m68k    unstable   outdated
 
 Initial value in the gentoo repository
 --------------------------------------
 
 On introduction, the setting will be ``stable`` for all stable architectures,
-``testing`` for all architectures where "inofficial" stable keywords are
+``degraded`` for all architectures where "inofficial" stable keywords are
 maintained and are present in the repository by the arch teams (sh, s390,
 ...), and ``unstable`` everywhere else.
 
@@ -125,8 +125,8 @@ by profiles.desc (and ``-d`` / ``-e`` switches).
 This is the current behaviour and shall be the default if nothing is specified
 for an architecture.
 
-testing
-~~~~~~~
+degraded
+~~~~~~~~
 When a profile of an architecture is tested, then repoman treats ``arch``
 in ebuilds as ``~arch``, and tests consistency only for ``~arch``.
 
@@ -134,7 +134,7 @@ Which profiles of the arch are tested is still controlled by profiles.desc
 (and ``-d`` / ``-e`` switches).
 
 A new switch for repoman may be provided to temporarily upgrade
-an architecture from ``testing`` to ``stable`` status (for architecture team
+an architecture from ``degraded`` to ``stable`` status (for architecture team
 work).
 
 unstable
@@ -170,7 +170,7 @@ arches.desc present and old system
 Utilities ignore the unknown file.
 
 Repoman and other tools may emit surplus dependency errors when profiles are
-checked on arches that are ``testing`` (they check the consistency
+checked on arches that are ``degraded`` (they check the consistency
 of the stable tree alone, which may fail, since ``arch`` is supposed to be
 treated like ``~arch``). This affects only development work and can be fixed
 by updating repoman.
-- 
2.26.0

Re: [gentoo-dev] [PATCH 03/10] glep-0072: Rename bad depgraph state to 'degraded'

[Ulrich Mueller]

[-- Attachment #1: Type: text/plain, Size: 708 bytes --]

>>>>> On Fri, 10 Apr 2020, Michał Górny wrote:

> In Gentoo terms, 'testing' and 'unstable' are mostly synonymous,
> so using the two names for different purposes is confusing.  Use
> 'degraded' instead.

"Degraded" has a negative ring to it, so can you find another term?
How about "lenient" or "relaxed"?

> -``testing`` for all architectures where "inofficial" stable keywords are
> +``degraded`` for all architectures where "inofficial" stable keywords are

I am aware that it isn't your change, but while at it, could you change
"inofficial" to "unofficial"? "Inofficial" is a false friend for German
writers and (as I've been told by a native speaker) isn't really a word
in English.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 507 bytes --]

Re: [gentoo-dev] [PATCH 03/10] glep-0072: Rename bad depgraph state to 'degraded'

[Ulrich Mueller]

[-- Attachment #1: Type: text/plain, Size: 508 bytes --]

>>>>> On Fri, 10 Apr 2020, Ulrich Mueller wrote:

>> -``testing`` for all architectures where "inofficial" stable keywords are
>> +``degraded`` for all architectures where "inofficial" stable keywords are

> I am aware that it isn't your change, but while at it, could you change
> "inofficial" to "unofficial"? "Inofficial" is a false friend for German
> writers and (as I've been told by a native speaker) isn't really a word
> in English.

Never mind, I just see that patch 06/10 drops the word entirely.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 507 bytes --]

Re: [gentoo-dev] [PATCH 03/10] glep-0072: Rename bad depgraph state to 'degraded'

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 757 bytes --]

On Fri, 2020-04-10 at 12:19 +0200, Ulrich Mueller wrote:
> > > > > > On Fri, 10 Apr 2020, Michał Górny wrote:
> > In Gentoo terms, 'testing' and 'unstable' are mostly synonymous,
> > so using the two names for different purposes is confusing.  Use
> > 'degraded' instead.
> 
> "Degraded" has a negative ring to it

Isn't that the truth, though?  I dare say it's a negative situation if
stable depgraph is broken.

> , so can you find another term?
> How about "lenient" or "relaxed"?

Thinking about it, all these terms seem too generic.  Would be nice to
find one that clearly suggests it's between testing and stable, and not
'lenient' in ~arch.  How about 'transitional' or 'incomplete-stable'?

-- 
Best regards,
Michał Górny


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 618 bytes --]

Re: [gentoo-dev] [PATCH 03/10] glep-0072: Rename bad depgraph state to 'degraded'

[Ulrich Mueller]

[-- Attachment #1: Type: text/plain, Size: 288 bytes --]

>>>>> On Sat, 11 Apr 2020, Michał Górny wrote:

> Thinking about it, all these terms seem too generic.  Would be nice to
> find one that clearly suggests it's between testing and stable, and not
> 'lenient' in ~arch.  How about 'transitional' or 'incomplete-stable'?

"interim"?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 507 bytes --]

Re: [gentoo-dev] [PATCH 03/10] glep-0072: Rename bad depgraph state to 'degraded'

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 425 bytes --]

On Sat, 2020-04-11 at 12:08 +0200, Ulrich Mueller wrote:
> > > > > > On Sat, 11 Apr 2020, Michał Górny wrote:
> > Thinking about it, all these terms seem too generic.  Would be nice to
> > find one that clearly suggests it's between testing and stable, and not
> > 'lenient' in ~arch.  How about 'transitional' or 'incomplete-stable'?
> 
> "interim"?

half-ass-stable?  ;-)

-- 
Best regards,
Michał Górny


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 618 bytes --]

Re: [gentoo-dev] [PATCH 03/10] glep-0072: Rename bad depgraph state to 'degraded'

[Andreas K. Hüttel]

[-- Attachment #1: Type: text/plain, Size: 620 bytes --]

Am Samstag, 11. April 2020, 14:23:48 EEST schrieb Michał Górny:
> On Sat, 2020-04-11 at 12:08 +0200, Ulrich Mueller wrote:
> > > > > > > On Sat, 11 Apr 2020, Michał Górny wrote:
> > > Thinking about it, all these terms seem too generic.  Would be nice to
> > > find one that clearly suggests it's between testing and stable, and not
> > > 'lenient' in ~arch.  How about 'transitional' or 'incomplete-stable'?
> > 
> > "interim"?
> 
> half-ass-stable?  ;-)

transcendent ...

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer 
(council, qa, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[gentoo-dev] [PATCH 04/10] glep-0072: Use 'testing' for pure ~arch

[Michał Górny]

'Testing' has generally nicer meaning than 'unstable'.

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index 6fdee6b..56a5bc4 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -91,7 +91,7 @@ are ignored.  Every blank line is ignored. Otherwise the file consists of two
 whitespace-separated columns:
 
 - first column: architecture name (keyword)
-- second column: one of the three values ``stable``, ``degraded``, ``unstable``
+- second column: one of the three values ``stable``, ``degraded``, ``testing``
 
 Additional columns are ignored to allow for future revisions of this document.
 
@@ -102,7 +102,7 @@ An example arches.desc file might look as follows::
     x86     stable     # not for long
 
     sparc   degraded
-    m68k    unstable   outdated
+    m68k    testing    outdated
 
 Initial value in the gentoo repository
 --------------------------------------
@@ -110,7 +110,7 @@ Initial value in the gentoo repository
 On introduction, the setting will be ``stable`` for all stable architectures,
 ``degraded`` for all architectures where "inofficial" stable keywords are
 maintained and are present in the repository by the arch teams (sh, s390,
-...), and ``unstable`` everywhere else.
+...), and ``testing`` everywhere else.
 
 Meaning of the values for repoman
 ---------------------------------
@@ -137,8 +137,8 @@ A new switch for repoman may be provided to temporarily upgrade
 an architecture from ``degraded`` to ``stable`` status (for architecture team
 work).
 
-unstable
-~~~~~~~~
+testing
+~~~~~~~
 When a profile of an architecture is tested, then repoman treats ``arch``
 as an error and aborts. Consistency is only tested for ``~arch``.
 
-- 
2.26.0

[gentoo-dev] [PATCH 05/10] glep-0072: Remove weird third column from example

[Michał Górny]

While it should technically be ignored, I don't think it's a good idea
to encourage developers using it for their own purposes.

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index 56a5bc4..b60d269 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -102,7 +102,7 @@ An example arches.desc file might look as follows::
     x86     stable     # not for long
 
     sparc   degraded
-    m68k    testing    outdated
+    m68k    testing
 
 Initial value in the gentoo repository
 --------------------------------------
-- 
2.26.0

[gentoo-dev] [PATCH 06/10] glep-0072: Update initial values

[Michał Górny]

I'm not aware of any profiles that should be set to 'degraded', so let's
focus on the immediate problem of stable/testing.  It will also probably
make sense to wait before we start using the third state.

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index b60d269..d6e96d5 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -107,10 +107,9 @@ An example arches.desc file might look as follows::
 Initial value in the gentoo repository
 --------------------------------------
 
-On introduction, the setting will be ``stable`` for all stable architectures,
-``degraded`` for all architectures where "inofficial" stable keywords are
-maintained and are present in the repository by the arch teams (sh, s390,
-...), and ``testing`` everywhere else.
+On introduction, the setting will be ``stable`` for all architectures using
+stable keywords, and ``testing`` for those that do not (``alpha``, ``mips``,
+``riscv``, Prefix profiles at the moment).
 
 Meaning of the values for repoman
 ---------------------------------
-- 
2.26.0

[gentoo-dev] [PATCH 07/10] glep-0072: Combine and amend description of states

[Michał Górny]

Provide a combined description for every status that explains what it
means, how it's used by linting tools and how it affects stabilization
requests.

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 46 +++++++++++++++++++---------------------------
 1 file changed, 19 insertions(+), 27 deletions(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index d6e96d5..fe3e0e5 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -111,49 +111,41 @@ On introduction, the setting will be ``stable`` for all architectures using
 stable keywords, and ``testing`` for those that do not (``alpha``, ``mips``,
 ``riscv``, Prefix profiles at the moment).
 
-Meaning of the values for repoman
----------------------------------
+Meaning of the values
+---------------------
 stable
 ~~~~~~
-When a profile of an architecture arch is tested, then repoman checks
-consistency of the dependency tree for ``arch`` and for ``~arch`` separately.
+Stable means that the architecture is actively maintaning stable keywords.
+When dependency graphs of packages with stable keywords are tested, they
+are tested separately for ``arch`` and ``~arch`` systems.
 
-Which profiles of the architecture are tested is still controlled
-by profiles.desc (and ``-d`` / ``-e`` switches).
+Stable architectures are listed first in keyword-relevant contexts (``eshowkw``,
+Bugzilla) and developers are expected to file stabilization requests on these
+arches.
 
 This is the current behaviour and shall be the default if nothing is specified
 for an architecture.
 
 degraded
 ~~~~~~~~
-When a profile of an architecture is tested, then repoman treats ``arch``
-in ebuilds as ``~arch``, and tests consistency only for ``~arch``.
+Degraded means that the architecture does not maintain a consistent stable
+dependency graph but uses stable keywords on some packages.  When dependency
+graphs of packages with stable keywords are tested, they are tested only
+for ``~arch`` systems, i.e. stable keywords are ignored.
 
-Which profiles of the arch are tested is still controlled by profiles.desc
-(and ``-d`` / ``-e`` switches).
+Degraded architectures are generally listed after stable architectures,
+possibly mixed with testing.  Developers are not expected to file stabilization
+requests.
 
-A new switch for repoman may be provided to temporarily upgrade
+A new switch for linting tools may be provided to temporarily upgrade
 an architecture from ``degraded`` to ``stable`` status (for architecture team
 work).
 
 testing
 ~~~~~~~
-When a profile of an architecture is tested, then repoman treats ``arch``
-as an error and aborts. Consistency is only tested for ``~arch``.
-
-Which profiles of the arch are tested is still controlled by profiles.desc
-(and ``-d`` / ``-e`` switches).
-
-Meaning of the values for other tools
--------------------------------------
-
-All architectures with the value ``stable`` are considered as stable
-architectures, in the sense that
-
-- they require stabilization requests on bugzilla, which are handled
-  by an arch team
-- they may, e.g., be listed first by tools such as eshowkw
-- and similar, to the discretion of tool authors
+Testing means that the architecture does not use stable keywords at all.
+Presence of such keywords is considered an error.  Consistency is tested
+only for ``~arch``.
 
 
 Backwards Compatibility
-- 
2.26.0

Re: [gentoo-dev] [PATCH 07/10] glep-0072: Combine and amend description of states

[Andreas K. Hüttel]

[-- Attachment #1: Type: text/plain, Size: 609 bytes --]

Am Freitag, 10. April 2020, 09:58:44 EEST schrieb Michał Górny:

>  Developers are not expected to file
> stabilization +requests.

This kinda changed in usage in the meantime (for, say, stuff like sparc and 
s390). The request was to CC them in the stabilization bugs if relevant.

How about
"Developers should file stabilization requests, however, pending stabilization 
on these arches alone cannot block any further steps (as, e.g., cleanup of old 
versions)."

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer 
(council, qa, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [gentoo-dev] [PATCH 07/10] glep-0072: Combine and amend description of states

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 671 bytes --]

On Fri, 2020-04-10 at 13:23 +0300, Andreas K. Hüttel wrote:
> Am Freitag, 10. April 2020, 09:58:44 EEST schrieb Michał Górny:
> 
> >  Developers are not expected to file
> > stabilization +requests.
> 
> This kinda changed in usage in the meantime (for, say, stuff like sparc and 
> s390). The request was to CC them in the stabilization bugs if relevant.
> 
> How about
> "Developers should file stabilization requests, however, pending stabilization 
> on these arches alone cannot block any further steps (as, e.g., cleanup of old 
> versions)."
> 

Isn't that implied by exp status, i.e. separate from this?

-- 
Best regards,
Michał Górny


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 618 bytes --]

Re: [gentoo-dev] [PATCH 07/10] glep-0072: Combine and amend description of states

[Andreas K. Hüttel]

[-- Attachment #1: Type: text/plain, Size: 468 bytes --]

> > "Developers should file stabilization requests, however, pending
> > stabilization on these arches alone cannot block any further steps (as,
> > e.g., cleanup of old versions)."
> 
> Isn't that implied by exp status, i.e. separate from this?

Hmm... do all "degraded" profiles have to be "exp"?
(Need to get the coffee first.)

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer 
(council, qa, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [gentoo-dev] [PATCH 07/10] glep-0072: Combine and amend description of states

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 649 bytes --]

On Fri, 2020-04-10 at 13:42 +0300, Andreas K. Hüttel wrote:
> > > "Developers should file stabilization requests, however, pending
> > > stabilization on these arches alone cannot block any further steps (as,
> > > e.g., cleanup of old versions)."
> > 
> > Isn't that implied by exp status, i.e. separate from this?
> 
> Hmm... do all "degraded" profiles have to be "exp"?
> (Need to get the coffee first.)

To be honest, I don't think they are really required to be bound
together.

Degraded just means 'ARCH is ~ARCH'.  stable/dev/exp determines whether
~ARCH-level integrity we care about.

-- 
Best regards,
Michał Górny


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 618 bytes --]

[gentoo-dev] [PATCH 08/10] glep-0072: Explicitly cover file not existing case

[Michał Górny]

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/glep-0072.rst b/glep-0072.rst
index fe3e0e5..763be77 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -95,6 +95,8 @@ whitespace-separated columns:
 
 Additional columns are ignored to allow for future revisions of this document.
 
+If the file does not exist, it is treated as if it were empty.
+
 An example arches.desc file might look as follows::
 
     # Example arches.desc file
-- 
2.26.0

[gentoo-dev] [PATCH 09/10] glep-0072: Move 'overlays' to spec, and change behavior

[Michał Górny]

Change the handling of slave repositories to the usual notion of 'slave
overrides master'.

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 22 ++++++++++++++--------
 1 file changed, 14 insertions(+), 8 deletions(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index 763be77..74ba9e8 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -149,6 +149,20 @@ Testing means that the architecture does not use stable keywords at all.
 Presence of such keywords is considered an error.  Consistency is tested
 only for ``~arch``.
 
+arches.desc in slave repositories
+---------------------------------
+
+If ``arches.desc`` is present in several repositories, then each file affects
+packages in the repository in question.  If the file does not specify a value
+for given arch, the value from the master repository is used.  However, using
+it in multiple repositories is discouraged.
+
+Note that the stability status override affects only packages in the slave
+repository and their direct dependencies.  If an arch is set to ``testing``,
+then master repositories are still permitted to use stable keywords.  If it is
+set to ``stable``, then missing stable keywords in dependencies from the master
+repository will cause dependency graph inconsistency.
+
 
 Backwards Compatibility
 =======================
@@ -177,14 +191,6 @@ to determine a list of stable arches shall fall back to the current method
 of determining stable arches by scanning profiles.desc for stable profiles.
 
 
-arches.desc in overlays
-=======================
-
-If arches.desc is present in several repositories, then the strictest setting
-for an architecture wins. Using arches.desc outside the gentoo (or
-alternative) master repository however is discouraged.
-
-
 Copyright
 =========
 
-- 
2.26.0

[gentoo-dev] [PATCH 10/10] glep-0072: Update metadata

[Michał Górny]

Signed-off-by: Michał Górny <mgorny@gentoo.org>
---
 glep-0072.rst | 16 +++++-----------
 1 file changed, 5 insertions(+), 11 deletions(-)

diff --git a/glep-0072.rst b/glep-0072.rst
index 74ba9e8..312525f 100644
--- a/glep-0072.rst
+++ b/glep-0072.rst
@@ -1,23 +1,17 @@
 ---
 GLEP: 72
 Title: Architecture stability status file
-Author: Andreas K. Hüttel <dilfridge@gentoo.org>
+Author: Andreas K. Hüttel <dilfridge@gentoo.org>,
+        Michał Górny <mgorny@gentoo.org>
 Type: Standards Track
-Status: Deferred
+Status: Draft
 Version: 1
 Created: 2017-05-06
-Last-Modified: 2019-06-10
-Post-History: 2017-05-06
+Last-Modified: 2020-04-10
+Post-History: 2017-05-06, 2020-04-10
 Content-Type: text/x-rst
 ---
 
-Status
-======
-
-Marked as deferred by GLEP editor Ulrich Müller on 2019-06-10, due to
-inactivity.
-
-
 Abstract
 ========
 
-- 
2.26.0

Re: [gentoo-dev] [PATCH 00/10] GLEP 72 (arches.desc) revival

[Andreas K. Hüttel]

[-- Attachment #1: Type: text/plain, Size: 814 bytes --]

Am Freitag, 10. April 2020, 09:58:37 EEST schrieb Michał Górny:
> 
> 1. 'Broken' status was removed, as it is redundant to profile status.
> 
> 2. State names were changed from 'testing' and 'unstable' to 'degraded'
>    (broken stable tree) and 'testing' (pure ~arch) to avoid confusion.
> 

Back in time there was also the idea to use this file to indicate security 
support of an arch. The suggestion was to introduce another column, but I 
found that rather horrible.

So a better idea would be to introduce an additional status "security", 
designating "stable with security support".

(Acts in every other respect exactly like "stable".)

What do you think?

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer 
(council, qa, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [gentoo-dev] [PATCH 00/10] GLEP 72 (arches.desc) revival

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1007 bytes --]

On Fri, 2020-04-10 at 13:37 +0300, Andreas K. Hüttel wrote:
> Am Freitag, 10. April 2020, 09:58:37 EEST schrieb Michał Górny:
> > 1. 'Broken' status was removed, as it is redundant to profile status.
> > 
> > 2. State names were changed from 'testing' and 'unstable' to 'degraded'
> >    (broken stable tree) and 'testing' (pure ~arch) to avoid confusion.
> > 
> 
> Back in time there was also the idea to use this file to indicate security 
> support of an arch. The suggestion was to introduce another column, but I 
> found that rather horrible.
> 
> So a better idea would be to introduce an additional status "security", 
> designating "stable with security support".
> 
> (Acts in every other respect exactly like "stable".)
> 
> What do you think?

I suppose it'd work for me.  I'm not sure if we have any use case for
having this data machine-readable but I guess there's no harm in it,
and better add it now than adjust tools later.

-- 
Best regards,
Michał Górny


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 618 bytes --]

Re: [gentoo-dev] [PATCH 00/10] GLEP 72 (arches.desc) revival

[Thomas Deutschmann]

[-- Attachment #1.1: Type: text/plain, Size: 3700 bytes --]

Hi,

regarding "security supported architectures" a few words from security
project:

We don't like the differentiation. And in practice, it doesn't even
matter nor does it work:

In theory, "security supported architectures" would allow us to ignore
bugs only affecting specific architectures. But examples are rare. I can
only think about a vulnerability affecting just x86
(https://security.gentoo.org/glsa/202003-13) or arm (like some
vulnerabilities in Xen hypervisor making use of specific hardware
features) in the past 24 months. So this is not really important and in
the end we don't really have the man power to differentiate. We just
bump because if we would skip a bug fix just because we thought it
doesn't affect us, this could have serious impact for no real reason.

We also always have to cc all architectures which have stable keywords
set on any affected ebuild which should be removed (cleaned up) after
security stabilization or maintainers will be unable to do the cleanup.

In the past, "security supported architecture" was also used to
determine when security team was allowed to publish a GLSA. However, we
changed this policy ~3 years ago:

Some people don't do regular world upgrades. They only pull updates when
they want a newer version or for security reasons via glsa-check tool.
Not telling amd64 users that they are using a vulnerable package where
we already have a fix for just because slacking architecture like hppa
in these days didn't stabilize this package yet was just unacceptable.

It wasn't an easy decision even in our small team because some members
had the concern that users will get warned about a vulnerability without
a solution yet available for their architecture, resulting in bug
spam/nagging. However, this never happened and some members even believe
that this is also an opportunity: Maybe some users not knowing that
their arch team is slacking would step up and join their architecture
team and help.

For the future some members even want to go one step further and release
GLSAs more often and not just for B2 or more severe vulnerabilities.

Back to "security supported architectures:

tl;dr

Security project wants to remove "security supported architectures" for
years.

Any architecture in Gentoo which is carrying stable keywords must keep
up with stabilization or keyword requests. Security stabilization
shouldn't be treated special (Because new ebuilds often depend on recent
libs. If an arch team would only focus on security bugs, calling for
stabilization will become more difficult because we would have to add
all the missing deps which are now required but already stable
everywhere else and just ignored by this arch until today).
If an architecture can no longer keep up with stabilization/keywording
demand, the entire architecture must be dropped to ~arch. No exception.
Stop pretending that this architecture is in good shape and that those
users have the same stable experience like you have on more common
architectures.

Keep in mind: You would also need to explain a user, "Yeah, you are
using something we name 'stable' but it doesn't mean what you are
expecting and BTW, while we have said we have fixed vulnerability X in
Gentoo you heard about in the media, don't forget to check on your own
if this is also true for your architecture because in theory the
maintainer could have decided to make use of arch-depending eapply for
some reason..."

=> Keep it simple: Stable should mean the same across all architectures


-- 
Regards,
Thomas Deutschmann / Gentoo Security Team
fpr: C4DD 695F A713 8F24 2AA1 5638 5849 7EE5 1D5D 74A5


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 618 bytes --]

Re: [gentoo-dev] [PATCH 00/10] GLEP 72 (arches.desc) revival

[Andreas K. Hüttel]

[-- Attachment #1: Type: text/plain, Size: 432 bytes --]

> 
> => Keep it simple: Stable should mean the same across all architectures

OK, this is a definite statement, so stable remains stable, and we introduce no 
additionally different status. 

I'd recommend that you drop the "security-supported arches" list from the 
security team web page too.

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer 
(council, qa, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]