With virtualbox guest modules now being part of Debian's kernel, it makes sense to have
the kernel package satisfy the virtual package slot, "virtualbox-guest-modules",
that is required by the virtualbox-guest* meta package. This, notably, is what's
already done by Ubuntu's kernel as well as Arch's, and other distro
kernels are likely to follow. Note that this is a somewhat temporary
arrangement -- we only need this while we're keeping virtualbox-guest-dkms
around to support kernels that don't come with wireguard built-in.
Over time that number will dwindle, at which point we can get rid of
virtualbox-guest-dkms as well as reverting this commit. In other words, adding
this Provides: line is a temporary stop-gap solution while wireguard
transitions from being an out-of-tree module to an in-tree module.

(taken from commit 5a053251

)

Signed-off-by: Gianfranco Costamagna <locutusofborg@debian.org>

TAPRIO is used for the TSN feature, the other switches should be quite
obvious.

Drop a patch that was applied upstream.

Since 5.9~rc4-1~exp1 was already uploaded, move the previous change to
a new version entry.

Fixes a regression that causes virtio_gpu to crash pretty quickly.

With the current configuration on amd64, all of the compression
libraries are built-in and compress-modules ends up empty.

This may not be the case on all architectures and flavours, but I
don't want to have to keep track of which do and don't need a
compress-modules package.  Move the modules, if there are any, into
kernel-image, and drop the compress-modules packages entirely.

* aufs: Disable support patchset until it is updated for Linux 5.9
* Makefile: Do not check for libelf when building OOT module
  (This is an expanded version of "Revert "objtool: Fix
   CONFIG_STACK_VALIDATION=y warning for out-of-tree modules"")
* Drop/refresh remaining patches as appropriate

binder uses fd APIs that are mostly already exported, since:

commit 44d8047f1d87adc2fd7eccc88533794f6d88c15e
Author: Todd Kjos <tkjos@android.com>
Date:   Tue Aug 28 13:46:25 2018 -0700

    binder: use standard functions to allocate fds

and vma APIs that are already exported, since:

commit 880211667b203dd32724f3be224c44c0400aa0a6
Author: Todd Kjos <tkjos@android.com>
Date:   Fri Feb 8 10:35:18 2019 -0800

    binder: remove kernel vm_area for buffer space

Gbp-Dch: Ignore

Drop ~exp1 from version

Set ABI to 1

Drop "[powerpc] 32s: Disable VMAP stack which CONFIG_ADB_PMU" patch
applied in v5.8.6.

Found by inspecting the device tree provided by Armbian. I skipped all the
settings relevant for video though. This has to wait until I have the
machine on my desk.

Drop "nfsd: fix oops on mixed NFSv4/NFSv3 client access"

debian/tests/python: pycodestyle: Increase max-line-length to 100

See merge request kernel-team/linux!266

Set the maximum allowed line length to 100 as the default of 79 is too
limiting small. Raising the allowed line length to 100 seems acceptable
enough.

This addresses autopkgtest failure for the python test since the
refactoring changes for debian/bin/gencontrol.py to factor out the
siable flags for installer modules, debug info and signed code.

Fixes: 9b82c45d ("Factor out installer modules disable flag")
Fixes: 5a2d988f ("Factor out debug info disable flag")
Fixes: 710b1f5b

 ("Add flag to disable signed code")
Signed-off-by: Salvatore Bonaccorso <carnil@debian.org>

Closes: #966218

Make debian/rules.d/bpftool/Makefile propagate CFLAGS, CPPFLAGS, and
LDFLAGS to the upstream Makefile in the variables it wants them.

I don't understand why upstream is still expanding this
semi-functional parallel hierarchy of UAPI headers.  The proper fix
would be to make tools depend on headers_install, and then:

git rm -rf tools/include/uapi tools/include/arch/*/uapi

hfs is now using symbols from cdrom, which results in duplication
unless we do this.

HFS is an old MacOS thing; there's no use for it on MIPS systems.

Release linux (5.7.17-1).

Cleanup debian/changelog file

Add CVE id reference for CVE-2019-19770

Cleanup debian/changelog file