Re: [gentoo-dev] Re: please stop using foo-${PV}-bar.patch in other ebuild versions than ${PV}

[Sebastian Pipping]

Ryan Hill wrote:
> Alin Năstac <mrness@g.o> wrote:
>
>> I suppose what everyone does in their part of the tree is their
>> business, but a small subset of packages I maintain have other
>> maintainers as well. It is annoying to see rules you assume being
>> respected on your ebuilds being broken at every bump made by others.
>
> I'm sure they're just as annoyed by your bizarre take on patch version
> numbering as you are with theirs. ;)

Let me try summarizing and dissecting this issue.
Please correct and extend where necessary.


Bike shedding versus "real issue"

  - Issue itself might not be earth-shaking

  - Repeated frustration can become a problem

  - Frustration with this is present (me included)

  - Happy developers stay much longer


People split into three groups:

  - Friends of  ${P}-fix-issue.patch  naming

  - Friends of  ${PN}-fix-issue.patch  naming

  - Friends of  ${PN}-1.2.3-fix-issue.patch  naming


Qualities

  - ${P} i.e. ${PN}-${PV}
    - On version bump either ..
      - Constant on version bump
      - Several same-content patch files across ebuilds
    - .. or ..
      - Change to ${PN}-<pre-bump-version>
      - Single patch file across ebuilds
    - Maybe helps reminding most patches should
      move from downstream to upstream?
    - 1:1 patch-ebuild relation (see below)

  - ${PN}
    - Constant on version bump
    - Single patch file across ebuilds
    - Several ebuilds need to be inspected
      to find out if a patch file is still used
    - 1:1 patch-ebuild relation (see below)

  - ${PN}-1.2.3
    - Constant on version bump
    - Single patch file across ebuilds
    - Several ebuilds need to be inspected
      to find out if a patch file is still used
    - <=${PV} values indicate things: either
      - the patch is downstream only
      - upstream has not applied it
    - 1:n patch-ebuild relation
      (inverse qualities of 1:1 patch-ebuild
      relation, see below)


  - 1:1 patch-ebuild relation
    - Finding out if a patch is still needed
      is trivial
    - In-place patch updates possible without
      affecting other ebuilds
    - 1:1 benefits only apply if ${P} is used
      consistently in the whole tree


Possible solutions

  - *Communicating* your likes to all co-maintainers
    in hope the will respect and remember your agreement

  - Add a related local comment (*documenting*) to ebuilds
    and expect other developers to act accordingly on a bump

  - Making a GLEP *enforcing* on of these and make people
    vote on which



Sebastian