pkgsrc-Users archive

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

Re: packages w/introspection option out-of-order update



"John D. Baker" <jdbaker%mylinuxisp.com@localhost> writes:

> While building stuff from pkgsrc-HEAD on i386-current, I ran into a
> couple of packages that failed to build because the "Foo-X.Y.gir" file
> it was including from another package couldn't be found.
>
> Perhaps my 'pkg_rolling-replace' was getting ahead of itself and replacing
> something due to another dependency update before updating the packages
> that provided the various "Foo-X.Y.gir" files.
>
> So far, I ran into:
>
>   clutter-gtk <- clutter <- {cogl,json-glib}
>   gst-plugins1-base <- gstreamer1
>
> After explicitly replacing the dependencies, I was able to build the
> package originally failing.
>
> Just a heads-up in case anyone else's 'pkg_rr' goes off into the weeds.

pkg_rr is not known to have any getting-dependencies-wrong issues.  It
looks at DEPENDS and BUILD_DEPENDS, and then does a tsort.   Almost all
reported pkg_rr problems are:

  some package doesn't build, and people don't understand that pkg_rr
  choosing to replace a package and the package not building is not a
  pkg_rr problem (despite me pointing this out almost every time :-)

  PLIST changes such that A depends on B, but B' has something in the
  PLIST that conflicts with A, even though it wouldn't conflict with
  A',n so you have to remove both and build both, which
  replace-one-at-a-time doesn't do.  This used to happen a lot with tex.
  It is typical with foo and foo-lib where a file is moved from foo to
  foo-lib during an update.

  odd stuff about packages whose source has been removed, not so much recently.

So it could be that there is an actual dependency which is not expressed
in the DEPENDS/bl3 logic.  And of course there could actually be a
pkg_rr bug.

Attachment: signature.asc
Description: PGP signature



Home | Main Index | Thread Index | Old Index