How to get KDE in: total analysis
I believe that the following solve (in one way or another) all the
packages broken by the imagemagick or KDE hints on i386 as of last
night, *not* including those already dealt with in vorlon's hint file.
Everything here deals with some actual specific package broken by one of
those two hints. I went through every such package, and I checked
reverse depends on everything proposed to be removed or rendered uninstallable,
recursively and in detail. I assume that the imagemagick hint goes through
This is quite long but there's less in it than one might imagine; I
commented quite heavily and provided alternate solutions (remove or force)
for a large number of things.
I expect that gcc-4.0 blockages are only going to increase as
time goes on, because some maintainers can't seem to stop making
uploads, so the faster this can be done, the better.
Now, it's true that no matter what we do, a bunch of stuff will be stuck
behind binutils/glibc/gcc-4.0, which is going to take a while, probably
a month or more. (It's not reasonable to force them through given the
build failures and grave bugs in binutils and glibc.)
I really think we need to get the KDE/JACK/etc hint through before
new gcc-4.0/glibc/binutils. If we do so:
* the packages clogged behind them will be primarily C++ packages,
as opposed to the current situation where the JACK transition,
the new PHP4 API, the new wxwindows, the new wireless-tools,
and various other things are sucked in
* little or no hinting will be necessary to let them in (unlike the
state of the current clog)
* the vast majority of the C++ transition will be present in testing
* many, many more recent packages will be in 'testing',
affording substantially more testing for those packages
And if we don't do so, it will take long enough that the current hinting
data will all be out of date and we'll have to start fresh.
So anyway, here's the results of my research.
# -- for the imagemagick hint -- #
# For the imagemagick hint, these are too young
# For the imagemagick hint, this is missing builds
# hasn't undergone the libmagick transition (bug 332455)
# For the imagemagick hint, the following packages' versions in
# unstable depend on gcc-4.0 on all or most architectures, including i386,
# and new imagemagick breaks the old versions
# (pstoedit will come out after autotrace is removed)
# For the imagemagick hint, xastir (FTBFS on hppa)
# needs to be removed as well.
# It should be removed on an *earlier line* than the imagemagick hint,
# I believe (currently it's being removed in the kde hint grouping,
# lower down)
# I also suggest throwing vips/7.10.15-1 into the imagemagick hint,
# because nip2 depends on it, and britney might check them in the wrong order.
# --- Now for the KDE hint, first the non-gcc-4.0-related ones --- #
# otherwise ready
# failing build on hppa, m68k, arm (ICE)
# (depends on sword, but you got that)
# Alternately, remove bibletime/1.4.1-2.1 bibletime-i18n/1.4.1-1
# FTBFS on alpha, fine everywhere else
# Alternately, remove mypasswordsafe/0.0.20041004-2
# FTBFS on s390, fine everywhere else
# Alternately, remove ecasound2.2/2.4.1-1
# non-free, out of date on ia64. Who cares?
# Alternately, remove lincvs/1.4.2-2
# pre-existing license bug
# Alternately, remove php-html-template-it/1.10.3
# Provided by php-pear, so maybe totally obsolete
# not in unstable, perhaps gone from testing now too?
# if needed, remove bookcase/???
# Depends on new openoffice.org, really seriously need to break the linkage
# both old & new busted by C++/KDE 3.4 transition, unsolved (#329231)
# FTBFS everywhere in unstable (& bug 324649)
# but old version broken by Qt C++ transition
# libqttestrunner1 is listed as made uninstallable by the KDE upgrade,
# but it's part of old cppunit and new cppunit should go in, so --
# perhaps cppunit/1.10.2-4 needs to be added to the KDE hint?
#--- the rest of this is at least tangentially gcc-4.0 related ---#
# kig from kdeedu depends on libboost-python1.33.0
# (and kdeedu build-depends on boost).
# Boost has difficult-to-solve FTBFSes on m68k and hppa.
# Hence the missing kdeedu binaries for m68k and hppa.
# It's not safe to rebuild either of them, anyway, as
# they're all C++ and gcc-4.0 would get dragged in.
# (Plus kdeedu is too young.)
# junior-kde and kde-amusements (from meta-kde) have dependencies on
# parts of kdeedu -- which is most likely not going to be available on
# m68k or hppa, see above -- so they will need to be forcibly broken on
# those architectures by use of a force-hint for the KDE hint when it's
# finally ready. But a force-hint is a good idea for gcc-4.0 related
# reasons, too, noted below.
# langdrill needs to be removed or rebuilt (for rekall, which is
# in substantially better condition and shouldn't be penalized),
# but a rebuild would drag in gcc-4.0.
# kdebindings doesn't build on m68k
# needs hppa, arm builds uploaded (or old binaries removed? something
# simple like that)
# is too young
# and can't have a new source upload without dragging in gcc-4.0
# stupid FTBFS which only happens when switching to new openssl
# it's a C++ package so it can't be rebuilt without dragging in gcc-4.0
# Alternatively, remove anon-proxy/00.02.39-7
# FTBFS on three arches, planned to use gcc-3.4 for them
# it's a C++ package so it can't have a new source without dragging in gcc-4.0
# Alternatively, force kxmleditor/1.1.4-2
# Depends on new gcc-4.0. Packages in unstable seem confused, with
# hplip being arch:all in one version and arch:any in the next, plus
# which the collection of built packages are changing,
# so I didn't try to work out whether forcing it would be an improvement,
# but I doubt it because it's brand new and is going to build on every
# arch with the new gcc-4.0, most likely.
# Unstable has brand new uploads which are probably going to depend on new
# gcc-4.0 everywhere, or at least too many places.
# not-yet-uploaded alpha, i386, mips, mipsel, s390, sparc builds,
# and probably FTBFSed arm build, will depend on gcc-4.0
# Alternately, force rosegarden4/1.0-1.2
# ------ #
# The remaining items should be added to the KDE hint (as a force-hint) when
# appropriate (during or after the KDE force-hint).
# They depend on new gcc-4.0 but only on relatively few architectures (noted), so
# for the greater good we should probably break them on those arches and let them in.
# Alternately, I think msot of these could be removed, as noted.
# force-hint kdeaddons/4:3.4.2-2.1
# -- ia64 (and unbuilt sparc, m68k)
# removal not a good alternative
# force-hint mapserver/4.6.1-3
# -- ia64
# Alternately, remove mapserver/4.4.1-2
# force-hint droidbattles/1.0.6-4
# -- powerpc
# Alternately, remove droidbattles/1.0.6-3
# force-hint keybled/0.65-4
# -- alpha build which hasn't been uploaded yet
# Alternately, remove keybled/0.65-2
# force-hint kxdocker/0.35-2.2
# -- ia64 build which hasn't been uploaded yet
# Alternately, remove kxdocker/0.26-2 kxdocker-data/0.12-1
# force-hint libqt-perl/3.008-1.4
# -- i386 (ugh), powerpc (and unbuilt sparc, m68k)
# also too young
# Alternately, remove libqt-perl/3.008-1.3
# force-hint xmms-kde/3.1-3
# -- i386 (ugh), powerpc, alpha, sparc (and unbuilt m68k)
# also too young
# Alternately, remove xmms-kde/3.1-1
Nathanael Nerode <email@example.com>
[Insert famous quote here]