Gentoo News

The Gentoo Council held a meeting on April 9, 2009. The items put up for discussion were:

Technical issues:

• Migration of KEYWORDS out of ebuilds: There was initial discussion on this idea. The council requested that Ned Ludd post a draft proposal of his idea to one of the lists for discussion, since the idea interested the council and they wish to get an idea of its pros and cons.
• EAPI 3 features block: To make sure that there is enough discussion on all EAPI 3 features a block on new features for consideration into EAPI 3 was considered. A vote was taken to:
  A. block features to those already discussed
  B. 'A' with mtime preservation
  C. no block for features
  Choice 'A' won with 4 votes. This block doesn't affect discussion of the implementation of the features, only new features.
• EAPI 3 updates: Zac Medico commented that while he hadn't worked on [use(+/-)] yet it shouldn't take more than a week or two to complete.

The Gentoo Council held a meeting on March 12, 2009. The items discussed were:

EAPI-3 Proposals:

Note: The following two proposals were discussed before it was realized that there was not sufficient time to discuss all of them. At that point a call for objections to any of the proposals was asked for, and none were made. A full list of proposals for EAPI 3 follow.

• New phase (pkg_pretend): This phase is most useful for displaying conflicting USE flags at dependency resolution time (pretend), though it has various other uses so that errors about installing the package can be displayed before installation of packages begin.
  Conclusion: Approved for the draft.
• [flag(+/-)] USE dependencies: This may be needed when one wishes to depend on a package with a certain USE flag but if the USE flag is not present in that package assume it is on or off (+ and - respectively).
  Conclusion: Approved for the draft.
• Multislot Dependency Specifications: This allows ebuilds to tell the package manager that runtime dependencies are not swappable (:1 installed at runtime can't be removed even though :2 'satisfies' the dependency).
• PROPERTIES mandatory in cache: Some information provided by this variable is useful at --pretend time (interactive packages).
• DEFINED_PHASES mandatory in cache: Same reasons as for PROPERTIES, but is also useful for determining the phases a package provides with just the cache.
• Provide a default src_install prototype: Get rid of the need for the src_install functions with just emake install in them. Some discussion is needed to clear up issues with a DOCS variable for extra documentation and a list of docs to automatically get installed.
• Provide a docompress function: This function serves as a replacement for prepalldocs. docompress can optionally compress files in /usr/share/doc according to a set of inclusion and exclusion lists.
• Provide an '-r' option to dodoc: Providing a way to put dodoc in recursive mode is widely accepted.
• Make doins preserve symlinks: This obsoletes the cp -R constructs frequently seen and is easy to implement.
• Limit values in $USE to those in $IUSE: Certain USE_EXPAND flags may be in USE even if they aren't specifically set in IUSE. Eliminate this.

The Council agreed to have portage implement as many of these as possible in a month and then make that EAPI 3.

Technical issues:

• GLEP 54: Thomas Anderson sent out a comparison of GLEP 54 and the liveebuild proposals. Among those discussing GLEP 54 there was a general consensus that there was nothing wrong with it as a first step to get correct ordering. Luca Barbato commented that all he was concerned about was that there was not enough 'meat' to the GLEP.
  Conclusion: Doug Goldstein and Luca Barbato intend to write a GLEP to handle the second part of the problem, making the revision available to ebuilds/package manager/users.
• GLEP 55: Petteri Räty, Zac, and Ciaran were supposed to benchmark the various proposals and report back. Zac Medico did not write the code for portage so Petteri Räty had nothing to report on this issue. Ciaran commented that the solutions other than GLEP 55 had a 50% slowdown in the valid cache situation compared to GLEP55, but did not post the raw numbers or the patches used.
  Conclusion: Zac Medico needs to benchmark the proposals in portage.

Open floor:

• Migration of KEYWORDS from ebuilds to profiles: Ned Ludd brought this up, but it came up in the middle of agenda items so was not talked about much. Some points were made that such a scheme would require a git conversion, but nothing was agreed upon.

The Gentoo Council held a meeting on February 26, 2009. The items discussed were:

Nontechnical issues:

• Open council spot: Should Mart Raudsepp fill the empty council spot? Since Mark Loeser resigned from the council there is an empty spot. Since Mart Raudsepp is ranked next from the last election, he is eligible to fill the spot.
  Conclusion: Mart Raudsepp is unanimously approved for the council.

Technical issues:

• GLEP 55: There had been quite a bit of discussion on this topic recently. Within hours of the council meeting new proposals were being proposed and discussion was ongoing.
  Conclusion: No decision as of yet. Ciaran McCreesh (ciaranm) and Zac Medico volunteered to benchmark the various proposals on the package managers they maintain (paludis and portage respectively). Petteri Räty will assist with the portage benchmarks. Tiziano Müller and Alec Warner will write up a comparison of the various proposals and their various advantages and disadvantages within a week.
• GLEP 54: There had been some discussion on gentoo-dev since last meeting, though no consensus or agreement had been reached.
  Conclusion: Thomas Anderson and Luca Barbato will write up a comparison of the advantages and disadvantages of the two proposals (-scm and _live). This will be completed within a week.
• Overlay Masking in Repositories: Brian Harring (ferringb) asked for discussion for when overlays attempted to unmask packages provided by the master repository (gentoo-x86). Because this is only available in portage (it is contrary to PMS), Brian thought it should not be allowed. Numerous suggestions were made to the effect that if a standardized set format was agreed upon for repositories and package.unmask was allowed to contain sets, then this problem would be fixed.
  Conclusion: No decision, as only discussion was requested. Mart Raudsepp will follow up on this with discussion on gentoo-dev.

The Gentoo Council held a meeting on March 26, 2009. The items discussed were:

Technical issues:

• GLEP 55: Petteri Räty noted that portage had recently gotten support for both GLEP 55 and the parse-eapi proposal. Petteri will have benchmarks done by the next meeting.
• EAPI-3 Proposals: A call for objections to/questions about any of the various proposals was asked for. What follows is a list of proposals to which objections were raised or for which there are open questions as well as who raised the points.

• slot operator support: leio, open questions, position pending on answers
• default_src_install: Mart Raudsepp, Donnie Berkholz, Tobias Scherbaum, open questions
• doinclude: Donnie Berkholz, Mart Raudsepp
• dosed: Donnie Berkholz
• unpack failing on unknown types: Donnie Berkholz
• docompress: Mart Raudsepp, needs to review proposal and prepalldocs. Tiziano Müller, thinks it's useless
• doexample: Tiziano Müller, thinks it should have -r if we have it at all
• dohard being deprecated: Mart Raudsepp, thinks it should remain and have its bugs fixed.
• disable-dependency-tracking: Luca Barbato, possible breakage of configure scripts (mplayer & ffmpeg mentioned)
• utility commands should die by default: Mart Raudsepp, open questions
• ban || ( foo? ( . ) . ): Mart Raudsepp, sees no reason to ban something that might have some valid use cases

One part of the EAPI-3 discussion is whether to have variables that behind-the-scenes control the default functions. The DOCS variable was created so that a list of documentation to install can be passed to default_src_install. A 4-2 vote approved the DOCS variable for use in src_install. Specific details have not yet been worked out.

Open floor:

Ned Ludd requested that the council discuss a migration of KEYWORDS out of ebuilds to be discussed at the next meeting.

The Gentoo Council held a meeting on December 11, 2008. The items discussed were:

Technical issues:

• Label profiles with EAPI for compatibility checks: Should there be labels in the profiles telling package managers what EAPI the profile uses. This proposal raised some concerns that developers would modify current profiles and bump the EAPI which would harm users' systems.
  Conclusion: Profile EAPI files are approved for use in gentoo-x86 profiles. The file for use in profiles is eapi. All current profiles are EAPI="0" and only new EAPIs can be marked with the profile EAPI markers. Any developer profiles can be marked with a new EAPI.
• Retroactive EAPI change: Call ebuild functions from trusted working directory. Donnie (dberkholz) commented that it may not be needed to add something to EAPI=0 that is in all the package managers.
  Conclusion: Approved. This change applies to all current EAPIs (0,1,2).
• Metadata variable for DEFINED_PHASES:
  Conclusion: Approved. Infra will do a full regen of the metadata cache once Portage has support for it.

The Gentoo Council held a meeting on November 13, 2008. The items discussed were:

Open bugs:

• Slacker Arches: Mark sent a proposal to gentoo-dev about this. No other discussion or decision.
• GLEP 55: .ebuild-$eapi extension.
  Conclusion: This bug has been RESO LATER'ed pending concrete need and consensus (GLEP 54 being the main one).
• Code of Conduct Update: Proctors no longer exist.
  Conclusion: Jorge (jmbsvicetto) will talk to other members of devrel about this bug.
• Extent of Code of Conduct enforcement: Banning people entirely from Gentoo.
  Conclusion: Userrel is responsible for establishing these policies. Since each developer is also a user at some point, these policies will apply to developers as well as users.
• GLEP 54: -scm package suffix.
  Conclusion: No decision. David Leverton commented that portage's current method for identifying 'live' ebuilds was hackish, as it depends on the list of inherited eclasses, which could change. Also, it was pointed out that some ebuilds use scm eclasses to check out specific revisions (mythtv). PROPERTIES=live was considered as another option.