Re: mercurial backport failed to build - quilt backport required?
Vincent Danjean schrieb am Tuesday, den 18. August 2009:
> Alexander Wirt wrote:
> > Vincent Danjean schrieb am Tuesday, den 11. August 2009:
> >> Gerfried Fuchs wrote:
> >>> Hi!
> >>> * Chris Butler <firstname.lastname@example.org> [2009-08-10 13:36:31 CEST]:
> >>>> I just tried to install the mercurial 1.3.1 backport on a lenny/i386 system,
> >>>> and got slightly confused when the package wasn't found (since I'd just
> >>>> installed it on amd64).. Turns out that only the amd64 package is available.
> >>>> According to the buildd log, the automated builds failed due to the required
> >>>> version of quilt not being available:
> >>>> After installing, the following source dependencies are still unsatisfied:
> >>>> quilt(inst 0.46-6 ! >= wanted 0.46-7)
> >>> Vincent, do you plan to backport quilt, or can you please revert your
> >>> change with respect to "dh --with quilt" to work with the build
> >>> environment in lenny? Makes me wonder how you managed to build and
> >>> upload the package for amd64, unless you haven't used a lenny system for
> >>> that?
> >> Inverting the "dh --with quilt" seems too invasive for me. I was planning
> >> to backport quilt before mercurial but it contradict the backports rules :
> >> quilt from testing is installable without any rebuild on lenny.
> >> To build my backport for lenny, I used a chroot where I manually installed
> >> quilt from testing (and quilt dependencies from plain lenny).
> > I'm sorry but this is total bullshit.
> I do not like at all the tone of your mail. Can I quote the contribute webpage ?
> "Reconsider if the package can be installed directly from testing without any
> recompilation and handled via pinning"
This has been removed for some time. But is a criterium for NOT uploading
packages to bpo and of course it doesn't count if a package is a build
depency for another package. Or should our autobuilder scrying where to get a
build depdency from?
> It was on the web when I backported mercurial 1.3 for lenny (never uploaded because
> mercurial 1.3.1 have been uploaded before 1.3 reaches testing). I admit I did not
> reread the whole website before preparing the 1.3.1 backport. This precision is
> important because the quoted sentence disappeared from the web page :
> Now, we can read :
> "Make sure that you have a proper build environment which only contains lenny and
> no unneeded backports. Maybe you want to consider pbuilder or cowbuilder for
> building packages."
> But this one (that show me I make a mistake when preparing the mercurial backport)
> as been recently added.
it has been changed a few months ago and it was even mentioned some time ago
via the mandatory ml:
> > We had quilt backports
> It was not there at the time I prepared the backport of mercurial (ie 1.3)
> > and we will have
> > quilt backports if requested as build dependency. Please don't do uploads to
> > bpo if you build them this way. The only rule that ever applies is: talk to
> > me if you have questions.
> I was thinking I was following the rules. The changes of the
> http://backports.org/dokuwiki/doku.php?id=contribute have been important. I do not
> know when the changes have been made but this is recent.
last year during etch backports and it wasn't a real change. You had to
fullfill your build deps since we had autobuilders which was during the whole
lifetime of etch-backports. This was discussed several times on this list.
> So, I will reupload mercurial and depend on (quilt >> 0.46-6) instead of
> (quilt >= 0.46-7).
No. Depend on the specific backport version.
> But, please, do not accuse someone the tell bullshit when it tells something
> that have been on the website a few weeks ago !
No it was just something that was there all the time.