kernel-package (11.001) unstable; urgency=low
* Bug fix: "kernel-package: Typo "call" -> "cause" in make-kpkg(1)
manpage", thanks to Ted Percival (Closes: #418496).
* Bug fix: "[kernel-package] please fail more verbose in preinst",
thanks to Carsten Hey (Closes: #404871).
* Bug fix: "kernel-package: make-kpkg should not limit choice of config
method", thanks to Alex Hermann. Added silenoldconfig, but did not
make it the default, since older kernel versions do not have this
target, and make-kpkg should not fail by default for those versions.
(Closes: #413098).
* Bug fix: "french manpage is out of date, and doesn't include the
ramdisk option.", thanks to Sven Luther (Closes: #404855).
* Bug fix: "kernel-package: make-kpkg debian fails if .config not
present", thanks to Sami Liedes. (Closes: #416090).
* Bug fix: "kernel/pkg/postinst says it's symlinking when hardlinking
and vice versa", thanks to Simon Huggins (Closes: #417814).
* Bug fix: "kernel-package: Option --cross-compile=- dont't work whith
target modules*", thanks to Jose Calhariz (Closes: #417948).
* Make kernel-package handle a non-existent .config file again (tested
on linux-2.2.26, linux-2.4.34, and linux-2.6.21.1). We figure out
which kernel needs a .config (>= 2.6.18), and for those kernels, call
make defconfig as needed.
* Added support for CONFIG_LOCALVERSION_AUTO and git source trees. If
CONFIG_LOCALVERSION_AUTO is set, see if we are running in a git
tree. If so, append -dirty to the git version string (since we do edit
the scripts Makefile to ensure ./debian does not go missing during
dpkg-buildpackage). Now that we can handle the appending of -dirty, we
no longer move setlocalversion out of the way. This has not, however,
been tested with a git tree.
* Added XS-VCS-Arch and XS-VCS-Browse to debian/control
-- Ubuntu Archive Auto-Sync <email address hidden> Sun, 06 May 2007 08:08:06 +0100
kernel-package (10.068) unstable; urgency=high
* Bug fix: "/sbin/update-grub is run by default by kernel postrm if it
exists, and complains", thanks to Joey Hess. Don't semi-silently
ignore missing hooks. This means: Don't gracefully handle missing hook
scripts, the diagnostic might be lost in a flurry of messages, and
the user may never be aware that the hook script was never run.
Depending on what the hook script is meant to do, this may cause a
security problem or data loss. (Closes: #412937).
* Bug fix: "kernel-package: version suffix added when building in git
trees ", thanks to Marcus Better. When run in a git sourced tree,
scripts/setlocalversion checks to see if the tree is the trunk, if
not, it adds version information to the kernel version. It also checks
to see if there are uncommited changes, if so, it adds the string
"-dirty" to the version. Since we change a couple of files in
./scripts/, this falls afoul of the make-kpkg checks, since the
version number is now different from what make-kpkg knows it to
be. Also, there is an additional truncated commit ID added by the
script scripts/setlocalversion whenever it detects that we are not
building from a tagged commit of the official tree, so it might be
just simpler to get rid of scripts/setlocalversion altogether, so we
now move it out of the way in the config phase, and restore during
clean. This allows make-kpkg to be useful for building from got trees
again. (Closes: #406180).
* Bug fix: "powerpc arch problems with 2.4 kernel", thanks to Brendan
O'Dea. For powerpc, handle circular dependency of version number and
KERNEL_ARCH. For powerpc, we need to set the (KERNEL_ARCH) based on
the version numbers -- which we can't get to unless we know the kernel
architecture. Yossarian would have liked this. Use grep to get the
most significant digits of the version number. Use these to calculate
significant digits of the version number. Use these to calculate the
KERNEL_ARCH. Use KERNEL_ARCH to get to the full version. This change
allows us not to drop support for 2.4 kernels on powerpc machines. It
does not impact anything but powerpc, and should not cause a change of
behavior on 2.6.16+ kernels even on powerpc. (Closes: #411038).
* Bug fix: "kernel-package: 'applied kernel patches' is missing from
buildinfo.gz", thanks to Norbert Buchmuller. Since kernel-source-* and
kernel-patch-* packages were renamed to linux-source-* and
linux-patch-*, the "applied kernel patches:" is missing from
buildinfo.gz. (It still tries to find kernel-patch-<whatever> package
for the <whatever> patch. (Closes: #412463).
* Bug fix: "kernel-package: 'kernel source package used:' info in
build.gz is inaccurate", thanks to Norbert Buchmuller. The information
that make-kpkg puts into the buildinfo.gz file with the "kernel source
package used:" description represents the version of the corresponding
{kernel,linux}-source-* package _installed_ on the system, but not the
version of the package that the current source tree was _untarred_
from. (Eg. one untars the source tree from linux-source-2.6.18-6, then
(s)he updates the installed package to linux-source-2.6.18-7, then
builds a kernel in the previously untarred source tree. In this case
the source really used is 2.6.18-6, but it writes 2.6.18-7 to
buildinfo.gz, which is misleading.) (Closes: #412589).
kernel-package (10.067) unstable; urgency=high
* Bug fix: "kernel-package: using obsolete package 'ldso'", thanks to
<email address hidden>. Well, we were not exactly "using" the package, but
just recording the version number. This is a low risk change.
(Closes: #407793).
* Bug fix: "kernel-package needs some additions to @files_to_remove",
thanks to Stephen Gran. Added modules.seriomap as a file to be
removed in the postrm, allowing /lib/modules/$version to be removed.
(Closes: #408900).
* Bug fix: "kernel-package: linux-headers for UML do not include various
Makefiles", thanks to Marcus Better. This should be a low impact
change. (Closes: #408369).
* Bug fix: "Kernel and initrd image symlinks aren't updated properly",
thanks to Dennis Vshivkov. This bug needs to be fixed for Etch, since
it causes the symbolic links to the image and initrd to not be
updated in some cases. (Closes: #409240).
* Bug fix: "kernel-package support for armel", thanks to Joey Hess. This
should be a low risk change for other architectures, and armel is not
yet in consideration for etch, so any support in kernel-package is
better than none. (Closes: #410459).
kernel-package (10.066) unstable; urgency=high
* Bug fix: "kernel-package: Enable -source and -headers packages for Xen
kernels", thanks to Ian Campbell. This is a simple change, and should
have minimal impact on any other package, or on current user built
packages. The upside is that the previously disabled packages are
enabled for Xen and UML variants, allowing people to build third party
modules for Xen adn UML variations -- and even if these were to break
(they do not, in my testing), there is no regression. In the future,
the image and modules packages will be separated, to allow for
host/guest installs. Official kernel images do not use make-kpkg to
build kernel or uml images, so the official images or debian-installer
are not impacted either. (Closes: #390881).
* Bug fix: "kernel-package: postinst link_in_boot value set by
image_in_boot", thanks to Niall Walsh. This was an out-and-out cut and
paste error; and the fix is very simple (remove offending line). The
downside is minimal, and current implementation is just wrong.
(Closes: #405081).
* Update spanish manual pages, thanks to Rudy Godoy. He also provided
patches for spelling errors and style flaws in the English man pages.
* Bug fix: "Typos in manpage", thanks to Goswin Brederlow
(Closes: #402155).
* Bug fix: "kernel-package: default stem inconsistent with
documentation", thanks to Lionel Elie Mamane (Closes: #400697).
* Bug fix: "'man kernel-img' typo: "messgaes"", thanks to
A. Costa (Closes: #397379).
* Bug fix: "'man kernel-pkg' typos: "Mutualy" and "theversion"", thanks
to A. Costa (Closes: #397378).
* Bug fix: "make-kpkg clean: leaves scripts/package/Makefile.kpkg-dist
behind", thanks to Yann Dirson. This was a simple typo, which
prevented the naming back scripts/package/Makefile.kpkg-dist to
scripts/package/Makefile -- making clean not restore the package to
original state after running build/clean. There is little downside to
this typo fix; we rename the Makefile, and we should be restoring it
in clean. (Closes: #397518).
* Bug fix: "kernel-package: linux-headers doesn't inclue arch-specific
scripts", thanks to dann frazier. When trying to compile external
modules on on IA64/McKinley ItaniumII, the Makefile in arch/ia64/
needs some scripts from arch/ia64/scripts/ directory. These scripts
were not so far being shipped. On non ia64, this has no impact on the
headers package built; on ia64 the changes are required for the
headers package to have any value. (Closes: #405494).
* Bug fix: "kernel-package: kernel image cant be installed when vmlinuz
link is already", thanks to Kai Sassmannshausen. Actually, the problem
was not depending on a new enough version of coreutils. See Below.
(Closes: #407128).
* Bug fix: "kernel-package: dependency problem: kernel package
installation uses a invalid option of readlink", thanks to Kai
Sassmannshausen. In release version 10.063, I cleaned up the postinst
as while I was at it, going with readlink -q -m instead of the built
in version. Since -m was a flag not present in Sarge's coreutils, this
means the generated image files should have a versioned dependency on
coreutils. The previous versioned dependency was not strict enough,
updated to a version that works. (Closes: #407130).
-- Ben Collins <email address hidden> Thu, 26 Apr 2007 07:23:10 +0100
