[Android-tools-devel] Bug#935308: buster-pu: package android-sdk-meta/25.0.0+11

[Adam D. Barratt]

Control: tags -1 - moreinfo

On Tue, 2019-08-27 at 09:30 +0200, Hans-Christoph Steiner wrote:
> Adam D. Barratt:
> > Control: tags -1 + moreinfo
> > 
> > On Wed, 2019-08-21 at 15:44 +0200, Hans-Christoph Steiner wrote:
[..]
> This is my first try to start a process to include new device IDs in
> Debian/stable so that these devices work with Debian/stable without
> extra tricks.  There are always new Android devices, which will have
> new device IDs.  I think that if Debian is going to provide

I think you missed the end of that thought.

[...]
> +android-sdk-meta (25.0.0+11~deb10u1) buster; urgency=medium
> > 
> > Such a version should include the changelog for the +11 upload to
> > unstable, and then have a short "rebuild for buster" or similar
> > stanza for ~deb10u1, in the same fashion as one would for
> > backports. The style
> > that you've used is generally used for +10+deb10u1 versioning.
> 
> Thanks for this explanation!  I was looking all over for this.  It
> was not in the manual.

Which manual? Policy doesn't mention stable versioning at all, and
arguably shouldn't get into such specific detail. DevRef's coverage is
somewhat disjointed and tends to get outdated - and doesn't even
mention +debXuY versioning in the stable uploads section, only in
discussions of uploads to the security archive, t-p-u uploads and NMUs.

> So I looked through other packages tagged with
> release.debian.org, and copied what I saw there.  From what I read,
> +deb10u1 is for security updates that are not also in testing/sid,
> and ~deb10u1 is for things also in testing/sid.
> 
> So right now, the changelog entry for 25.0.0+11 in testing and
> 25.0.0+11~deb10u1 in proposed-updated are the exact same, except for
> the date and ~deb10u1.  I figured that 25.0.0+11 would never be in
> buster.

Sort of. The distinction is more whether the update package for stable
is created by starting with the package already in stable and adding
desired changes to that (+debXuY), or takes the package from unstable
and rebuilds that on stable with only changes required to do so applied
(~debXuY). The latter case should necessarily include the changelog
stanza from the unstable upload(s) "as-is", as it represents a fork
from that package.

In some cases those methods may produce packages that are identical
other than the version and which stanzas are included in the changelog.

> Should I make ~deb10u2 to do the changes you describe?

I'm not sure it's worth an upload cycle now that the package is already
in stable-new, just bearing the discussion in mind for the future.

> Seems like the FAQ on this page could be a good place for this to be
> explained:
> https://wiki.debian.org/StableProposedUpdates
> 
> I can edit this discussion into a FAQ there, if you agree.

To be entirely honest, I would rather get the information from the
"bits from SRM" posts, together with some from our "how to be an SRM"
manualette, turned into more formal documentation on release.d.o

We already have too many places that contain outdated, conflicting or
plain incorrect information and I'm not sure that trying to make a wiki
page that hasn't been written or reviewed by the Release Team to be
canonical is a useful approach.

> > >    [ Jeff Muizelaar ]
> > >    * Add some more Amazon devices
> > >  .
> > >    [ Hans-Christoph Steiner ]
> > >    * fix DEB_REVISION parsing to work with more than one digit
> > > 
> > 
> > I assume that's this change:
> > 
> > -DEB_REVISION = $(shell echo $(DEB_VERSION) | sed 's,.*\+\([0-9][0-
> > 9]*\).*,\1,')
> > +DEB_REVISION = $(shell echo $(DEB_VERSION) | sed 's,.*+\([0-9][0-
> > 9]*\).*,\1,')
> > 
> > However, I'm confused as to how this corresponds to the
> > description. It removes a match for a literal "+" symbol, but does
> > nothing related to the number of digits involved so far as I can
> > tell.
> 
> sed != PCRE.   By default, sed does not treat + as a special
> character.
>  'sed -E' makes it PCRE.   This was fixed by testing.  The original
> regex was only ever returning a single digit, this returns one or
> more.

Ugh, yeah.

>  I've been writing regexs for over 20 years, and this one took me a
> while to see...

Similarly. This is one of the reasons that I tend to end up writing
horrible RE when using sed, because trying to get it to do what I
expect is more annoying than just hacking something that works
together. But I digress...

[...]
> >  debian/android-sdk.metainfo.xml                  |  109 ++++++
> > 
> > +debian/android-sdk.metainfo.xml  usr/share/metainfo
> > 
> > Is that related to one of the items mentioned in the changelog?
> 
> * document udev rule for providing device access to the console user

OK. "metainfo" didn't sound very documentationey. :-)

Regards,

Adam