[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#898494: dgit: conventions for uploading to suites with semantics like *-backports



Package: dgit
Version: 4.4
Severity: wishlist
X-debbugs-cc: mattia@debian.org, debian-backports@lists.debian.org

Hello,

Here are two ways of managing uploads to -backports with dgit:

1. maintain a stretch-bpo branch, merge the debian/ tags for uploads
   that have migrated to testing, `dch --bpo` and `dgit push-source` to
   backports

2. start a new, temporary branch for each backport, based on the debian/
   tags for uploads that have migrated to testing, `git cherry-pick` any
   needed changes for the backport from dgit/dgit/stretch-backports,
   `dch --bpo`, and then either `git merge -s ours
   dgit/dgit/stretch-backports`, or equivalently, `dgit push-source
   --overwrite`

I have been using (1) because it seems wrong to have a workflow based
around --overwrite, and the cherry-picking stage of (2) is a little
cumbersome.  However, I think that maybe (2) fits better with the
semantics of our -backports suites -- in a sense, backports are
continually rebased on the contents of testing.

In particular, the changelog entry for the upload to backports is meant
to contain a description of all the changes from the version in testing
that were needed in order to backport.  If you use (1), this means that
you have multiple entries in d/changelog with the same text, which
doesn't really make sense in a changelog.[1]

Is this strange debian/changelog reason enough to recommend (2) over
(1), are there other reasons to so recommend, or is the choice between
(2) and (1) just personal preference?  Maybe we need
dgit-maint-backports(7), or maybe we can just leave it to users.

I thought I'd file a bug CCed to relevant people, to see whether this
needs documenting or can just be left as it is.

[1]  Example, from one of my packages:

    ocrmypdf (6.2.0-1~bpo9+1) stretch-backports; urgency=medium

      * Rebuild for stretch-backports.
      * Tighten dependency on qpdf to ensure the backport gets installed.
        With qpdf 7.0.0 or newer, OCRmyPDF produces better output.  So what we
        want to achieve is the semantic meaning of:
            Depends: qpdf
            Recommends: qpdf (>= 7.0.0)
        However, these lines won't cause apt to install qpdf from
        stretch-backports.  So we add a hard dependency on qpdf (>= 7.0.0) to
        ensure that happens, which is most useful to users.

     -- Sean Whitton <spwhitton@spwhitton.name>  Fri, 11 May 2018 20:01:12 -0700

    ocrmypdf (6.2.0-1) unstable; urgency=medium

      * New upstream release.
      * [elided]

     -- Sean Whitton <spwhitton@spwhitton.name>  Sat, 05 May 2018 12:12:22 -0700

    ocrmypdf (6.1.2-1~bpo9+1) stretch-backports; urgency=medium

      * Rebuild for stretch-backports.
      * Tighten dependency on qpdf to ensure the backport gets installed.
        With qpdf 7.0.0 or newer, OCRmyPDF produces better output.  So what we
        want to achieve is the semantic meaning of:
            Depends: qpdf
            Recommends: qpdf (>= 7.0.0)
        However, these lines won't cause apt to install qpdf from
        stretch-backports.  So we add a hard dependency on qpdf (>= 7.0.0) to
        ensure that happens, which is most useful to users.

     -- Sean Whitton <spwhitton@spwhitton.name>  Wed, 11 Apr 2018 14:40:26 -0700

    ocrmypdf (6.1.2-1) unstable; urgency=low

      * New upstream release (Closes: #888917).
      * [elided]

     -- Sean Whitton <spwhitton@spwhitton.name>  Sat, 31 Mar 2018 11:30:50 -0700

-- 
Sean Whitton

Attachment: signature.asc
Description: PGP signature


Reply to: