package/pkg-kconfig: quote HOSTCC_NOCCACHE

HOSTCC may contain spaces, so needs to be quoted.

Most of the places where it is already quoted use double-quotes, so we
use that.

Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

package/menu-cache: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/a97825f3c3e6245f8d1c2eb0cdb079f5dd6f1b47

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

package/graphite2: security bump to version 1.3.14

- Switch site to github, here is an extract of
  https://sourceforge.net/projects/silgraphite:
  "This project has been deprecated. Graphite2, a new version of the
  Graphite engine, is available at: https://github.com/silnrsi/graphite
  with its own bug tracker."
- graphite2 can be built statically since version 1.3.11 and
  https://github.com/silnrsi/graphite/commit/2f143c04da5caa43ddf4dba437b2f2bc26bf4238
- Update indentation in hash file (two spaces)

Extract from ChangeLog:

1.3.14
    . Bug fixes
    . Allow features to be hidden (for aliases)
    . Move to python3
    . Rename doc files from .txt to .asc

1.3.13
    . Resolve minor spacing issue in rtl non-overlap kerning
    . python3 for graphite.py
    . Better fuzzing
    . Better building on windows

1.3.12
    . Graphite no longer does dumb rendering for fonts with no smarts
    . Segment caching code removed. Anything attempting to use the segment cache gets given a regular face instead
    . Add libfuzzer support
    . Builds now require C++11
    . Improvements to Windows 64 bit builds
    . Support different versions of python including 32 bit and python 3
    . Various minor bug fixes

1.3.11
    . Fixes due to security review
    . Minor collision avoidance fixes
    . Fix LZ4 decompressor against high compression

The fixes due to security review are a little bit vague, a quick search
on github seems to indicate that those issues could be related to
segcache which has been removed since version 1.3.12:
https://github.com/silnrsi/graphite/search?q=security&type=Issues
https://github.com/silnrsi/graphite/commit/b0f77e4a9dc50a888f74e904000a2486b2fc5527

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

package/uclibc: use HOSTCC_NOCCACHE as kconfig HOSTCC

uclibc is part of the toolchain, and as such does not have a dependency
on it. As a consequence, it does not have a dependency on host-ccache,
when this is needed.

Usually, host-ccache is built before uclibc, as part of the dependency
of gcc-initial, host-binutils, and a few other host packages that are
built before uclibc.

However, during top-level parallel builds, this ordering is only ever
guaranteed at the beginning of the configure step, and not before.

But for kconfig-packages, the moment we apply the configuration to
prepare the .config file is a pseudo step that happens somewhere in
limbo between the patch step and the configure step. As such, the
build ordering that is otherwise guaranteed by the _DEPENDENCIES is not
applicable yet.

And so, with top-level parallel builds with ccache enabled, there is
nothing that guarantees host-ccache to be built and installed by the
time we are trying to generate uclibc's .config file, which can be quite
early in the build process, and thus the build fails:

    /home/raphael/github/ftcommunity-TXT/buildroot-rootfs/output/per-package/uclibc/host/bin/ccache /usr/bin/gcc /home/raphael/github/ftcommunity-TXT/buildroot-rootfs/output/build/uclibc-1.0.34/extra/config/conf.c  -c -o ../../extra/config/conf.o -Os -I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"  -DNCURSES_WIDECHAR=1 -DLOCALE -DKBUILD_NO_NLS -DCONFIG_='""'   -I/usr/include/ncursesw -DCURSES_LOC="<curses.h>"  -DNCURSES_WIDECHAR=1 -DLOCALE -DKBUILD_NO_NLS -DCONFIG_='""'
    /bin/sh: 1: /home/raphael/github/ftcommunity-TXT/buildroot-rootfs/output/per-package/uclibc/host/bin/ccache: not found
    make[2]: *** [Makefile:64: ../../extra/config/conf.o] Error 127
    make[1]: *** [Makefile.in:475: extra/config/conf] Error 2
    make[1]: Leaving directory '/home/raphael/github/ftcommunity-TXT/buildroot-rootfs/output/build/uclibc-1.0.34'
    make: *** [package/uclibc/uclibc.mk:458: /home/raphael/github/ftcommunity-TXT/buildroot-rootfs/output/build/uclibc-1.0.34/.stamp_dotconfig] Error 2
    make: *** Waiting for unfinished jobs....

The root cause is that uclibc sets;

    UCLIBC_KCONFIG_OPTS = $(UCLIBC_MAKE_FLAGS) [...]

with:

    UCLIBC_MAKE_FLAGS = [...] HOSTCC="$(HOSTCC)"

And then the kconfig-package infra calls to the configurators,
menuconfig, xconfig et al, but also olddefconfig et al.. with:

    [...] $($(1)_MAKE) [...] $(PKG_KCONFIG_COMMON_OPTS) $($(1)_KCONFIG_OPTS) [...]

with (note a latent bug in there, will be fixed in another patch):

    PKG_KCONFIG_COMMON_OPTS = HOSTCC=$(HOSTCC_NOCCACHE)

So, a HOSTCC as set by a package will always win onver the one set by
the infra, which is exactly what we want.

But in this case, uclibc sets HOSTCC so that it can build its host tools
needed during the build, and in doing so uses the ccache-enabled host c
compiler. Which might not yet be available for the kconfig-package infra
to generate the .config file.

We had a similar (non-)issue for the linux package, which was fixed in
commit 71a31b2357 (linux: use HOSTCC_NOCCACHE as kconfig HOSTCC).

But here, uclibc does not have the toolchain in its dependencies (as said
earlier, uclibc *is* part of the toolchain).

Since the host compiler is only used to build very few files to generate
the simple executable needed to generate the .config file, doing without
the ccache-enabled host compiler will be amply enough.

So, we override HOSTCC in UCLIBC_KCONFIG_OPTS, to use the non-cached
host compiler.

Note that, in a first approximation, one would be tempted to change the
ordering in the kconfig-package infra:

        $($(1)_KCONFIG_OPTS) $(PKG_KCONFIG_COMMON_OPTS)

so that the non-cached HOSTCC always wins over the cached one. But this
would be incorrect, in cases where the package really needs to override
HOSTCC; indeed we want the package-provided values to always win over
the default ones providing by the infra.

Reported-by: Raphael Jacob <r.jacob2002@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Acked-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

docs/website: update for 2020.02.5

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Update for 2020.02.5

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 6ec5f4eb7121a2dd8cf08c4ea805aa3c9a586b84)
[Peter: drop Makefile changes]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

CHANGES: fix typos in 2020.08-rcX notes

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

docs/website: update for 2020.05.2

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Update for 2020.05.2

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 1549e0b60751eab41e2d51899981d43d602275af)
[Peter: drop Makefile changes]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/domoticz: drop SYNC4 from comment

Commit 8f5a9f597e35e6bc89dc938edbe753004d0201d1 forgot to drop SYNC4
from comment

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/mosquitto: security bump to v1.6.12

Mosquitto 1.6.11 is a bugfix release, read the whole announcement on
http://mosquitto.org/blog/2020/08/version-1-6-11-released/

Mosquitto 1.6.12 is a security and bugfix release, read
http://mosquitto.org/blog/2020/08/version-1-6-12-released/

>From the 1.6.11 changelog of the client library:
mosquitto_loop_start() now sets a thread name on Linux, FreeBSD, NetBSD,
and OpenBSD. Closes #1777.
This is done with pthread_setname_np; so mosquitto now requires
BR2_TOOLCHAIN_HAS_THREADS_NPTL when built with threading support.

2 reverse dependencies use the threaded API, but they already
depend on BR2_TOOLCHAIN_HAS_THREADS_NPTL:
* domoticz [1] (we add a comment for mosquitto)
* shairport-sync [2]

[1] https://github.com/domoticz/domoticz/blob/2020.1/main/mosquitto_helper.cpp#L344
[2] https://github.com/mikebrady/shairport-sync/blob/3.3.6/mqtt.c#L227-L229

Signed-off-by: Titouan Christophe <titouan.christophe@railnova.eu>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/{collectd, domoticz}: fix outdated dependencies for mosquitto

In 4fc62e1eb6b3adbfc3d3eb7f841275ae8cd1b424, we removed arch/toolchain
dependencies from the mosquitto library (MMU, !STATIC, SYNC4), and moved
them to the mosquitto broker only.

All the packages modified here only need the mosquitto library, so they
shouldn't have those depends anymore; but this was never done before.

Signed-off-by: Titouan Christophe <titouan.christophe@railnova.eu>
[Peter: leave mmu/!static dependency for domoticz as it uses fork()/looks
for libmosquitto.so]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/wolfssl: fix build with big endian

Fixes:
 - http://autobuild.buildroot.org/results/21098180d386890025ed5cdd243bf5a9b444c5cf

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/haproxy: bump to version 2.2.2

Drop patch (already in version)

http://www.haproxy.org/download/2.2/src/CHANGELOG

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/libressl: bump to version 3.1.4

https://ftp.openbsd.org/pub/OpenBSD/LibreSSL/libressl-3.1.4-relnotes.txt

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/zbar: fix NLS build with musl

Fixes:
 - http://autobuild.buildroot.org/results/b93ce5430bf22ddda94ee30882a883348617f5b1

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/systemd: disable audit for host package

Disable audit for host package to avoid getting the following error if
it is found on host:

[84/662] Generating audit_type-list.txt with a meson_exe.py custom command
In file included from <command-line>:32:
./../src/basic/missing_audit.h:7:10: fatal error: libaudit.h: No such file or directory
    7 | #include <libaudit.h>
      |          ^~~~~~~~~~~~

Fixes:
 - http://autobuild.buildroot.org/results/67782c225c08387c1bbcbea9eee3ca12bc6577cd

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/systemd: disable cryptsetup for host package

Build with cryptsetup and without libblkid will fail on:

../src/shared/dissect-image.c:1336:34: error: 'N_DEVICE_NODE_LIST_ATTEMPTS' undeclared (first use in this function)
 1336 |         for (unsigned i = 0; i < N_DEVICE_NODE_LIST_ATTEMPTS; i++) {
      |                                  ^~~~~~~~~~~~~~~~~~~~~~~~~~~

This bug has been reported upstream:
https://github.com/systemd/systemd/pull/16901
and is not an issue for the target variant as libblkid is select by
BR2_PACKAGE_UTIL_LINUX_MOUNT

As cryptsetup does not seem needed for host-systemd, just disable it

Fixes:
 - http://autobuild.buildroot.org/results/67782c225c08387c1bbcbea9eee3ca12bc6577cd

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Makefile: use $(Q) instead of @ to silence target-finalize commands

As 18f6c26118 just did to silence the file lists commands, switch to
using $(Q) instead of a plain @, to silence the commands.

Using $(Q) will allow to debug the commands with V=1.

We keep @ for the calls to MESSAGE, though.

The commands that are not currently silenced are left as-is, and they
can be converted to being silent in a followup patch, if need be,

Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

linux: run depmod only if modules directory exists

If the modules directory that corresponds to the version of the kernel
being built has been deleted, don't try to run depmod, which will
obviously fail.

This can happen for instance when the modules are stripped from the main
root filesystem, and placed into a separate filesystem image, so that
the root filesystem and the kernel can be updated separately.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

fs/cpio/init: unbreak ttyname_r() on glibc after dropping /dev/console exec

Commit 98a6f1fc02e41 (fs/cpio: make initramfs init script survive 'console='
kernel argument) dropped the explicit /dev/console execs for fd 0,1,2, as
they fail when booted with console= and aren't really needed as the kernel
will setup fd 0,1,2 from /dev/console before executing the initramfs anyway.

Not doing this unfortunately confuses glibc's ttyname_r(3) implementation
(used by E.G.  busybox/coreutils 'tty'), causing it to fail with ENOENT as
it does a fstat on fd 0 and tries to match up st_ino / st_dev against the
entries in /dev (since glibc 2.26):

 commit 15e9a4f378c8607c2ae1aa465436af4321db0e23
 Author: Christian Brauner <christian.brauner@canonical.com>
 Date:   Fri Jan 27 15:59:59 2017 +0100

    linux ttyname and ttyname_r: do not return wrong results

    If a link (say /proc/self/fd/0) pointing to a device, say /dev/pts/2, in a
    parent mount namespace is passed to ttyname, and a /dev/pts/2 exists (in a
    different devpts) in the current namespace, then it returns /dev/pts/2.
    But /dev/pts/2 is NOT the current tty, it is a different file and device.

    Detect this case and return ENODEV.  Userspace can choose to take this as a hint
    that the fd points to a tty device but to act on the fd rather than the link.

Signed-off-by: Serge Hallyn <serge@hallyn.com>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
The reason it fails is that we manually mount devtmpfs on /dev in /init, so
the /dev/console used by the kernel (in rootfs) is not the same file as
/dev/console at runtime (in devtmpfs).

Notice: Once logged in, tty does work correctly.  Presumably login reopens
stdin/stdout/stderr.

To fix this, re-add the exec of /dev/console for fd 0,1,2, but only do so if
possible.  Because of the above mentioned shell behaviour (specified by
POSIX [0]), perform this check in a subshell.

[0] https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_20_01

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/lxc: bump to version 4.0.4

- Bug fix release: https://linuxcontainers.org/fr/lxc/news
- Drop patch (already in version)

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/postgresql: security bump to version 12.4

- Fix CVE-2020-14349: It was found that PostgreSQL versions before 12.4,
  before 11.9 and before 10.14 did not properly sanitize the search_path
  during logical replication. An authenticated attacker could use this
  flaw in an attack similar to CVE-2018-1058, in order to execute
  arbitrary SQL command in the context of the user used for replication.
- Fix CVE-2020-14350: It was found that some PostgreSQL extensions did
  not use search_path safely in their installation script. An attacker
  with sufficient privileges could use this flaw to trick an
  administrator into executing a specially crafted script, during the
  installation or update of such extension. This affects PostgreSQL
  versions before 12.4, before 11.9, before 10.14, before 9.6.19, and
  before 9.5.23.

https://www.postgresql.org/docs/12/release-12-4.html

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/mongodb: security bump to version 4.2.9

SERVER-47733 SymmetricEncryptorWindows shouldn’t pad when update is
called

https://docs.mongodb.com/manual/release-notes/4.2-changelog/#id1

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Makefile: hide commands that build the package file lists at end of build

Since commit 0e2be4db8ab01d479177a3a187c22525752195ae
("package/pkg-generic: make file list logic parallel build
compatible"), the commands executed at the every end of the build
to assemble the list of files installed by the different packages
are visible in the make output. They are quite noisy, and clutter
the output.

The other commands in target-finalize are also hidden using "@",
so we should also do the same for those commands. But that hurts
debuggability, so we use $(Q) (the existing '@'s can be changed
in a followup patch).

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
[yann.morin.1998@free.fr: use '$(Q)', not '@']
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/squid: security bump to version 4.13

Fixes the following security issues:

CVE-2020-15810: HTTP(S) Request Smuggling
Due to incorrect data validation Squid is vulnerable to HTTP Request
Smuggling attacks against HTTP and HTTPS traffic.  This leads to cache
poisoning.
https://github.com/squid-cache/squid/security/advisories/GHSA-3365-q9qx-f98m

CVE-2020-15811: HTTP(S) Request Splitting
Due to incorrect data validation Squid is vulnerable to HTTP Request
Splitting attacks against HTTP and HTTPS traffic.  This leads to cache
poisoning.
https://github.com/squid-cache/squid/security/advisories/GHSA-c7p8-xqhm-49wv

CVE-2020-24606: Denial of Service processing Cache Digest Response
Due to Improper Input Validation Squid is vulnerable to a Denial of Service
attack against the machine operating Squid.
https://github.com/squid-cache/squid/security/advisories/GHSA-vvj7-xjgq-g2jg

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/wolfssl: security bump to version 4.5.0

wolfSSL version 4.5.0 contains 6 vulnerability fixes: 2 fixes for TLS 1.3,
2 side channel attack mitigations, 1 fix for a potential private key leak
in a specific use case, 1 fix for DTLS including those 3 CVEs:

- Fix CVE-2020-12457: An issue was discovered in wolfSSL before 4.5.0.
  It mishandles the change_cipher_spec (CCS) message processing logic
  for TLS 1.3. If an attacker sends ChangeCipherSpec messages in a
  crafted way involving more than one in a row, the server becomes stuck
  in the ProcessReply() loop, i.e., a denial of service.
- Fix CVE-2020-15309: An issue was discovered in wolfSSL before 4.5.0,
  when single precision is not employed. Local attackers can conduct a
  cache-timing attack against public key operations. These attackers may
  already have obtained sensitive information if the affected system has
  been used for private key operations (e.g., signing with a private
  key).
- Fix CVE-2020-24585: An issue was discovered in the DTLS handshake
  implementation in wolfSSL before 4.5.0. Clear DTLS application_data
  messages in epoch 0 do not produce an out-of-order error. Instead,
  these messages are returned to the application.

Also update hash of LICENSING as well as WOLF_LICENSE due to later
verbage update with
https://github.com/wolfSSL/wolfssl/commit/970391319beb023680eccd0e447e76834dbb4808

https://www.wolfssl.com/docs/security-vulnerabilities/

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/wireshark: security bump to version 3.2.6

Fix CVE-2020-17498: In Wireshark 3.2.0 to 3.2.5, the Kafka protocol
dissector could crash. This was addressed in
epan/dissectors/packet-kafka.c by avoiding a double free during LZ4
decompression.

https://www.wireshark.org/security/wnpa-sec-2020-10.html

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/chocolate-doom: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/02828f2d9956d1e3727774b5045790aa3611428d

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/bluez-tools: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/2641cea0483c5f6b65ece8016d546ee9bea0d7d1

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

Update for 2020.08-rc3

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/glibc: security bump for additional post-2.31.x fixes

Fixes the following security issue:

  CVE-2016-10228: An infinite loop has been fixed in the iconv program when
  invoked with the -c option and when processing invalid multi-byte input
  sequences.  Reported by Jan Engelhardt.

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/openal: needs gcc >= 4.9

openal uses std::max_align_t since version 1.20.0 and
https://github.com/kcat/openal-soft/commit/585b0cf3bed7d1c5720633eb7e5358a9fca865f6

As a result, it is affected by
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=56019
and the build with gcc <= 4.8 will fail on:

/home/buildroot/autobuild/instance-2/output-1/build/openal-1.20.1/common/almalloc.cpp: In function 'void* al_malloc(size_t, size_t)':
/home/buildroot/autobuild/instance-2/output-1/build/openal-1.20.1/common/almalloc.cpp:20:45: error: 'max_align_t' is not a member of 'std'
     alignment = std::max(alignment, alignof(std::max_align_t));
                                             ^

Fixes:
 - http://autobuild.buildroot.org/results/589c7853ce334c7502f7cd4cdbcaaf3c6840f43b

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/mender: Fix incorrectly named service file

Many of the mender CLI commands use systemctl commands to get information about
the daemon, such as the PID (IE: systemctl show -p MainPID mender-client).
As seen above, these commands expect the service file to be named
"mender-client" instead of "mender."

As such, in the current state, running a forced update check in the CLI will
result in the following error:
failed to force updateCheck: could not find the PID of the mender daemon.

Changing the name of mender.service to mender-client.service fixes the issue.

Signed-off-by: Adam Duskett <Aduskett@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

{linux, linux-headers}: bump 4.{4, 9, 14, 19}.x / 5.{4, 7}.x series

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

linux: workaround make-4.1 bug

On Ubuntu 18.04, make-4.1 emits spurious, incorrect "entering/leaving"
messages, which end up in the LINUX_VERSION_PROBED variable:

    printf 'probed linux version: "%s"\n' "$(LINUX_VERSION_PROBED)"
    probed linux version: "make[1]: Entering directory '/home/buildroot'
    4.19.78-linux4sam-6.2
    make[1]: Leaving directory '/home/buildroot/output/build/linux-linux4sam_6.2'"

First, the messages are displayed even though we do explicitly pass
--no-print-directory -s.

Second, the entering and leaving messages are not about the same
directory!

This *only* occurs in the following conditions:

  - the user has the correct 0022 umask,
  - top-level parallel is used (with or without PPD),
  - initial -C is specified as well.

    $ umask 0022
    $ make -j16 -C $(pwd)
    [...]
    depmod: ERROR: Bad version passed make[1]:
    [...]

(yes, 'make[1]:' is the string depmod is trying, and fails, to parse as
a version string).

If any of the three conditions above is removed, the problem no longer
occurs. Here's a table of the MAKEFLAGS:

                |                   0002                         |          0022            |
    ----+-------+------------------------------------------------+--------------------------+
        | no-j  | --no-print-directory --                        |                          |
    noC |       +------------------------------------------------+--------------------------+
        | -j16  | -j --jobserver-fds=3,4 --no-print-directory -- | -j --jobserver-fds=3,4   |
    ----+-------+------------------------------------------------+--------------------------+
        | no-j  | --no-print-directory --                        | w                        |
    -C  |       +------------------------------------------------+--------------------------+
        | -j16  | -j --jobserver-fds=3,4 --no-print-directory -- | w -j --jobserver-fds=3,4 |
    ----+-------+------------------------------------------------+--------------------------+

    0002: umask == 0002
    0022: umask == 0022

    no-j: no -j flag
    -j16: -j16 flag

    noC: no -C flag
    -C : -C /path/of/buildroot/

Only the bottom-right-most case fails...

This behaviour goes against what is documented:

    https://www.gnu.org/software/make/manual/make.html#g_t_002dw-Option

    5.7.4 The ‘--print-directory’ Option
    [...]
    you do not need to specify this option because ‘make’ does it for
    you: ‘-w’ is turned on automatically when you use the ‘-C’ option,
    and in sub-makes. make will not automatically turn on ‘-w’ if you
    also use ‘-s’, which says to be silent, or if you use
    ‘--no-print-directory’ to explicitly disable it.

So this exactly describes our situation; yet 'w' is added to MAKEFLAGS.

Getting rid of the 'w' flag makes the build succeed again, so that's
what we do here (bleark, icky)...

Furthermore, the documented way to override MAKEFLAGS is to do so as a
make parameter:

    https://www.gnu.org/software/make/manual/make.html#Options_002fRecursion

    5.7.3 Communicating Options to a Sub-make
    [...]
    If you do not want to pass the other flags down, you must change the
    value of MAKEFLAGS, like this:

        subsystem:
            cd subdir && $(MAKE) MAKEFLAGS=

However, doing so does not fix the issue. So we resort to pass the
modified MAKEFLAGS via the environment (bleark, icky)...

Fixes: #13141
Reported-by: Laurent <laurent@neko-labs.eu>
Reported-by: Asaf Kahlon <asafka7@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/trousers: add upstream security fix

Fixes the following security issues:

CVE-2020-24332
If the tcsd daemon is started with root privileges,
the creation of the system.data file is prone to symlink attacks

CVE-2020-24330
If the tcsd daemon is started with root privileges,
it fails to drop the root gid after it is no longer needed

CVE-2020-24331
If the tcsd daemon is started with root privileges,
the tss user has read and write access to the /etc/tcsd.conf file

For details, see the advisory:
https://www.openwall.com/lists/oss-security/2020/05/20/3

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/x11r7/xlib_libX11: security bump version to 1.6.12

Fixes CVE-2020-14363:
https://lists.x.org/archives/xorg-announce/2020-August/003056.html

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/x11r7/xserver_xorg-server: security bump version to 1.20.9

Fixes CVE-2020-14345, CVE-2020-14346, CVE-2020-14361 & CVE-2020-1436:
https://lists.x.org/archives/xorg-announce/2020-August/003058.html

Removed patch 0002, not needed anymore due to upstream commit
https://cgit.freedesktop.org/xorg/xserver/commit/configure.ac?h=server-1.20-branch&id=c601c8faf54ff9e3bcbc653421828d71042deef7

Build-tested with wayland:
checking for a useful monotonic clock ......
checking whether CLOCK_MONOTONIC is declared... yes
guessing yes

Removed patch 0007, included in upstream release.

Rebased and renumbered remaining patches.

Reformatted license hashes.

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/shadowsocks-libev: security bump to version 3.3.4

- Fix CVE-2019-5163: An exploitable denial-of-service vulnerability
  exists in the UDPRelay functionality of Shadowsocks-libev 3.3.2. When
  utilizing a Stream Cipher and a local_address, arbitrary UDP packets
  can cause a FATAL error code path and exit. An attacker can send
  arbitrary UDP packets to trigger this vulnerability.
- Fix CVE-2019-5164: An exploitable code execution vulnerability exists
  in the ss-manager binary of Shadowsocks-libev 3.3.2. Specially crafted
  network packets sent to ss-manager can cause an arbitrary binary to
  run, resulting in code execution and privilege escalation. An attacker
  can send network packets to trigger this vulnerability.

Also update indentation in hash file (two spaces)

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/openjpeg: add CVE-2020-15389 entry

Commit b006cc373f96ec86c027779e113c8f70bc40d1c3 forgot to add
the OPENJPEG_IGNORE_CVES entry

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/python-matplotlib: simplify version checks

Hopefully, this should fix the following error on one of the
autobuilders:

png: no  [The C/C++ header for libpng (png.h) could not
     be found.  You may need to install the development
     package.]

Fixes:
 - http://autobuild.buildroot.org/results/afddcc44b2fb7983244f24542bfae921869e4ab8

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/dillo: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/9c777af97fe50143c6a68f0170fc86c87d8ead3f

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/dillo: renumber patches

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/gnuradio: backport patch to fix INTERFACE_INCLUDE_DIRECTORIES

gnuradio-runtimeTargets.cmake and gnuradio-pmtTargets.cmake are filled
using CMAKE_INSTALL_PREFIX for INSTALL_INTERFACE.

Since CMAKE_INSTALL_PREFIX, in buildroot, is set to /usr, these files contains
path to host system.

With BR2_COMPILER_PARANOID_UNSAFE_PATH package using gnuradio fails with:
arm-linux-gnueabihf-g++: ERROR: unsafe header/library path used in cross-compilation: '-isystem' '/usr/include'

By simply providing 'include', produced .cmake contains:
INTERFACE_INCLUDE_DIRECTORIES "${_IMPORT_PREFIX}/include"
instead of
INTERFACE_INCLUDE_DIRECTORIES "/usr/include"

[Upstream status: https://github.com/gnuradio/gnuradio/pull/3737]

Fix (many) gr-osmosdr build failure:
- http://autobuild.buildroot.net/results/66b76c07f15bb3e6db697c47796ae3dd15ecf4b9/

Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

DEVELOPERS: add myself as contact for linuxptp+ipmitool

Signed-off-by: Heiko Thiery <heiko.thiery@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/openjpeg: fix CVE-2020-15389

Fix CVE-2020-15389: jp2/opj_decompress.c in OpenJPEG through 2.3.1 has a
use-after-free that can be triggered if there is a mix of valid and
invalid files in a directory operated on by the decompressor. Triggering
a double-free may also be possible. This is related to calling
opj_image_destroy twice.

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/json-c: security bump to version 0.15

Fix CVE-2020-12762: json-c through 0.14 has an integer overflow and
out-of-bounds write via a large JSON file, as demonstrated by
printbuf_memappend.

Also update indentation in hash file (two spaces)

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

support/scripts/pkg-stats: drop erroneous "break" in CVE.affects()

Commit 7d2779ecbb142b62f8913d30352b11058f922b2a
("support/script/pkg-stats: handle exception when version comparison
fails") erroneousy introduced a "break" within a try/expect block.

This break has the unfortunate consequence that every CVE that was
using the <= operator was skipped, and according to the current
CVE statistics, made us miss 74 CVEs out of 141 CVEs.

Here is for reference the complete list of CVEs we missed:

 - gnupg
   CVE-2006-3082
   CVE-2019-13050

 - jhead
   CVE-2020-6624
   CVE-2020-6625

 - patch
   CVE-2018-6952
   CVE-2019-20633

 - json-c
   CVE-2020-12762

 - git
   CVE-2018-1000110
   CVE-2018-1000182
   CVE-2019-1003010
   CVE-2020-2136

 - iperf2
   CVE-2016-4303

 - libtorrent
   CVE-2009-1760
   CVE-2016-5301

 - lua
   CVE-2020-15888
   CVE-2020-15889
   CVE-2020-15945
   CVE-2020-24342

 - openvpn
   CVE-2020-7224

 - smack
   CVE-2016-10027

 - bashtop
   CVE-2019-18276

 - links
   CVE-2008-3319

 - argus
   CVE-2011-3332

 - libraw
   CVE-2020-15503

 - netcat
   CVE-2008-5727
   CVE-2008-5728
   CVE-2008-5729
   CVE-2008-5730
   CVE-2008-5742
   CVE-2015-2214

 - subversion
   CVE-2017-1000085
   CVE-2018-1000111
   CVE-2020-2111

 - python
   CVE-2013-1753
   CVE-2015-5652
   CVE-2017-17522
   CVE-2017-18207
   CVE-2019-20907
   CVE-2019-9674

 - cereal
   CVE-2020-11104
   CVE-2020-11105

 - opencv
   CVE-2017-1000450
   CVE-2017-12597
   CVE-2017-12598
   CVE-2017-12599
   CVE-2017-12600
   CVE-2017-12601
   CVE-2017-12602
   CVE-2017-12603
   CVE-2017-12604
   CVE-2017-12605
   CVE-2017-12606
   CVE-2017-12862
   CVE-2017-12863
   CVE-2017-12864
   CVE-2019-15939

 - docker
   CVE-2015-1843
   CVE-2015-3627
   CVE-2015-3630
   CVE-2015-3631
   CVE-2016-3697
   CVE-2017-14992
   CVE-2019-16884

 - trousers
   CVE-2020-24330
   CVE-2020-24331
   CVE-2020-24332

 - libcroco
   CVE-2020-12825

 - libpupnp
   CVE-2020-13848

 - openjpeg
   CVE-2020-15389

 - flex
   CVE-2015-1773

 - libesmtp
   CVE-2019-19977

 - ed
   CVE-2015-2987

 - libmad
   CVE-2018-7263

 - grub
   CVE-2020-15705

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/ripgrep: fix build directories

RIPGREP_CARGO_MODE was no longer defined after 832c076f26 and caused
issues during the install step as the build directory was malformed.

This patch maintains the release/dev profile distinction, while also
assigning appropriate build folders.

Fixes:
 - http://autobuild.buildroot.net/results/a4cd7ecc6d983aa6f15d3be1e21529f17e04b825/
 - http://autobuild.buildroot.net/results/2bab8ffa590d4c4eabffe94ed27311c7f6607c98/

Signed-off-by: Sam Voss <sam.voss@gmail.com>
CC: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/libroxml: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/b6ac3664d61ad826515b57c4d057b6f001b5167d

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/assimp: also build with -mxgot on mips64(el)

Since the bump of assimp to 5.0.1, we have build failures on mips64el,
due to relocations being truncated. The issue seems to be quite
similar to the one on m68k coldfire, as both m68k and MIPS have this
-mxgot gcc option to switch to using a GOT that has no size limit (but
causes less efficient code to be produced).

Here as well, the overall relevance of assimp on mips64(el) platforms
being probably very limited, the incentive to search for a better
solution is pretty limited.

Fixes:

  http://autobuild.buildroot.net/results/7df487d5117b2ee440a07dbff9cae1b181566748/

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/assimp: workaround m68k build issues

On m68k coldfire, we already pass -mxgot, but since the bump to assimp
5.0.1, this is no longer sufficient, and we have failures such as:

/tmp/ccqmJLil.s: Assembler messages:
/tmp/ccqmJLil.s:307948: Error: value -43420 out of range
/tmp/ccqmJLil.s:307985: Error: value -38606 out of range
/tmp/ccqmJLil.s:308010: Error: value -38626 out of range
/tmp/ccqmJLil.s:308056: Error: value -33280 out of range

Since these issues only arise when building with -O2, let's disable
the optimization for this package on m68k. The very relative relevance
of assimp on m68k coldfire makes the research of a better solution not
really useful (for the record, assimp is a "library to import various
well-known 3D model formats in a uniform manner").

Fixes:

  http://autobuild.buildroot.net/results/a7d4fb2653b0f1be4d036ee46a44e72da0ed4376/

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/patchelf: keep RPATH entries even without DT_NEEDED libraries

Our patch
0003-Add-option-to-make-the-rpath-relative-under-a-specif.patch adds
an option --make-rpath-relative, which we use to tweak RPATH of target
binaries.

However, one of the effect of this option is that it drops RPATH
entries if the corresponding directory does not contain a library that
is referenced by a DT_NEEDED entry of the binary.

This unfortunately isn't correct, as RPATH entries are not only used
by the dynamic linker to resolve the location of libraries listed
through DT_NEEDED entries: RPATH entries are also used by dlopen()
when resolving the location of libraries that are loaded at runtime.

Therefore, the removal of RPATH entries that don't correspond to
directories containing libraries referenced by DT_NEEDED entries break
legitimate uses of RPATH for dlopen()ed libraries.

This issue was even pointed out during the review of the upstream pull
request:

  https://github.com/NixOS/patchelf/pull/118#discussion_r329660138

This fixes tst-origin uClibc-ng unit test:

https://github.com/wbx-github/uclibc-ng-test/blob/master/test/dlopen/Makefile.in#L25
https://github.com/wbx-github/uclibc-ng-test/blob/master/test/dlopen/tst-origin.c#L15

Without this patch:

$ gcc -o toto toto.c -Wl,-rpath,/tmp/test/bar
$ readelf -d toto | grep PATH
 0x000000000000000f (RPATH)              Library rpath: [/tmp/test/bar]
$ ./output/host/bin/patchelf --debug --make-rpath-relative /tmp/
toto
patching ELF file `toto'
Kernel page size is 4096 bytes
removing directory '/tmp/test/bar' from RPATH because it does not contain needed libs
new rpath is `'
$ readelf -d toto | grep PATH
 0x000000000000001d (RUNPATH)            Library runpath: []

With the patch applied:

$ gcc -o toto toto.c -Wl,-rpath,/tmp/test/bar
$ readelf -d toto | grep PATH
 0x000000000000000f (RPATH)              Library rpath: [/tmp/test/bar]
$ ./output/host/bin/patchelf --debug --make-rpath-relative /tmp/ toto
patching ELF file `toto'
Kernel page size is 4096 bytes
keeping relative path of /tmp/test/bar
new rpath is `test/bar'
$ readelf -d toto | grep PATH
 0x000000000000001d (RUNPATH)            Library runpath: [test/bar]

Signed-off-by: Yann Sionneau <ysionneau@kalray.eu>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

support/config-fragments/autobuild: test Bootlin x86-64 toolchain

As we recently stopped testing the x86-64 Sourcery toolchain, it means
we no longer have any x86-64 glibc based toolchain in our
autobuilders. Since this is a pretty common configuration, it makes
sense to test it, which this commit does by adding a config fragment
to use the x86-64 glibc bleeding edge Bootlin toolchain.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

toolchain/toolchain-external/toolchain-external-codesourcery-amd64: remove package

This toolchain uses an old gcc 6.2.0 compiler (not even the latest gcc
from the 6.x series), which fails to build the recent Boost
package. Since newer versions of this toolchain are no longer made
publicly available from Mentor Graphics, our only option is to drop
the toolchain.

Fixes:

  http://autobuild.buildroot.net/results/10edaed22c15b9d0f7de187085aeebc96e5ebe6c/

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

support/config-fragments/autobuild: stop testing Sourcery AMD64 toolchain

This toolchain uses an old gcc 6.2.0, and newer versions of the
toolchain are no longer publicly available. This old gcc 6.2.0 causes
build issues of Boost, which are unfixable without updating the
toolchain. As we're about to drop support for this toolchain entirely,
we must stop testing it in our autobuilder infrastructure.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/davfs2: bump to version 1.6.0

This bump is needed to fix a build failure with gcc 10:
https://savannah.nongnu.org/support/?func=detailitem&item_id=110186#options

Here is an extract of the bug report:

"It is not a bug to have variables with the same name in different source
files. The bug was the missing keyword "static".

But there was a different bug that was not tolerated by GCC 10. It was
same strange data type conversions in dav_coda.c. The resolution was to
drop coda altogether because there is still fuse which is better suited
anyway.

The new release 1.6.0 should fix all these problems. Please tell me if
there are still problems with GCC 10."

Also update indentation in hash file (two spaces)

Fixes:
 - http://autobuild.buildroot.org/results/42beafade6fd31927c8db14bc52110c0fc5b17c2

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/hostapd: add upstream 2020-1 security patches

Fixes the following security vulnerabilities:

CVE-2020-12695: The Open Connectivity Foundation UPnP specification before
2020-04-17 does not forbid the acceptance of a subscription request with a
delivery URL on a different network segment than the fully qualified
event-subscription URL, aka the CallStranger issue.

For details, see the advisory:
https://w1.fi/security/2020-1/upnp-subscribe-misbehavior-wps-ap.txt

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/ibm-sw-tpm2: re-fix build with uClibc-ng and older glibcs

In commit 26e37cef1627faa1f5ab8935482e2b2bc3465c33, we started using
__WORDSIZE to get the size of longs on the given architecture, in
order to support all CPU architectures.

Unfortunately, __WORDSIZE is not enabled in musl, so in
19bd08900448aa45b506320ad2ab912f789e6e5e, we switched to using
LONG_BIT instead of __WORDSIZE.

However, LONG_BIT is not readily available on glibc, you need
_XOPEN_SOURCE to be defined, which was done in
a34e7f88f67b77066f73894dc8e42bca3c076fa6.

However, in a34e7f88f67b77066f73894dc8e42bca3c076fa6, _XOPEN_SOURCE
was just defined, with no specific value. This caused the build to
break again on uClibc-ng and older glibcs, because clock_gettime() and
CLOCK_MONOTONIC were no longer defined. In both uClibc-ng and glibc,
CLOCK_MONOTONIC is only defined if __USE_POSIX199309 is defined. It
turns out that simply defining _XOPEN_SOURCE with no value does not
lead to __USE_POSIX199309 being defined in uClibc-ng and old glibcs,
while it is defined in newer glibcs.

The difference comes from the following snippet of code, which is
present in recent enough glibc's <feature.h> but not uClibc-ng's or
older glibc's <feature.h>:

/* If none of the ANSI/POSIX macros are defined, or if _DEFAULT_SOURCE
   is defined, use POSIX.1-2008 (or another version depending on
   _XOPEN_SOURCE).  */

So the fact that we are defining _DEFAULT_SOURCE makes it assume that
we're using POSIX 2008.09, which obviously includes POSIX 1993.09.

Due to the lack of this code snippet, uClibc-ng <features.h> only
enables:

     !defined _POSIX_SOURCE && !defined _POSIX_C_SOURCE)

but not:

So we need an _XOPEN_SOURCE level of at least 500 for POSIX 1993.09
definitions to be available.

This is confirmed by the feature_test_macros man page, which states:

                  _XOPEN_SOURCE < 500
                         _POSIX_C_SOURCE is defined with the value 2.

                  500 <= _XOPEN_SOURCE < 600
                         _POSIX_C_SOURCE is defined with the value 199506L.

When this is fixed, another issue arises with older glibc toolchains
(such as Sourcery ARM), where fd_set is no longer defined. Inded, with
POSIX-1.2001 being enabled, we need to include <sys/select.h> to
access the fd_set definition and friends (see man fd_set for details).

This commit was tested with two glibc toolchains (recent and old), one
uClibc-ng toolchain and one musl toolchain.

Fixes:

  http://autobuild.buildroot.net/results/e20f9474fc0217036faa6561df33fa983466ddfe/
  (uClibc-ng)

  http://autobuild.buildroot.net/results/b5d944389fc96ef2c5e0608fe4ac34149e5f9739/
  (glibc)

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Acked-by: Vincent Fazio <vfazio@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/netopeer2: add patch to solve issue with empty group name

When building on a host that has no name specified for the used group in
/etc/group the script in install step will fail due to missing group name.

CMake Error at CMakeLists.txt:80 (message):
  Learning server module group failed: id: cannot find name for group ID 8000

The patch was taken from upstream and modified manually because of merge
conflicts.

Fixes:
http://autobuild.buildroot.net/results/f197ca1def9dc1292e1e784757f2da9d95484431/

Signed-off-by: Heiko Thiery <heiko.thiery@gmail.com>
Cc: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Cc: Matthew Weber <matthew.weber@rockwellcollins.com>
Reviewed-by: Matthew Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/ipmitool: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/b6231d601d6051c97d3c2a0ed3065df03648c40d

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Tested-by: Heiko Thiery <heiko.thiery@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

legacy: drop options that are now forcibly enabled

As Thomas said:

> In this sort of situation, we generally don't add any legacy
> handling.  Indeed, since the feature is now mandatory... the
> default behavior will always be OK.

> People who could be annoyed are people who had this feature
> disabled... which is now always enabled. But the legacy handling
> will anyway not help those people.

Reported-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

docs/website/news.html: correct left/right ordering of 2020.08-rc2 entry

And drop the confusing class="timeline" tag from the 2020.08-rc1 entry.

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Update for 2020.08-rc2

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/openfpgaloader: drop ftdipp

ftdipp is not needed since version 0.1 and
https://github.com/trabucayre/openFPGALoader/commit/3df577b70638c37885eef3a27b8931261b259e47

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/openfpgaloader: drop udev from comment

Commit 5714f3f81fdd85640d627f9b43490c52419650ea forgot to remove udev
from comment

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

{linux, linux-headers}: bump 4.{4, 9, 14, 19}.x / 5.{4, 7}.x series

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/ripgrep: fix debug build

There is no --debug mode for cargo resulting in the following build
failure since the addition of this package with commit
4b0d1ef6ac00c5170a3fb9d15d06b3d3172e0c97:

error: Unknown flag: '--debug'

Fixes:
 - http://autobuild.buildroot.org/results/58e74bb056ec65680ecebaa559aa14bdebbf5c85
 - http://autobuild.buildroot.org/results/28c6364a89a6044d5a036614f7a6e59815efb770

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
[yann.morin.1998@free.fr: keep the default 'dev' mode when in debug]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/openfpgaloader: bump to current master

- argp is no more used;
- UDEV dependency is now optional

Fix:
- http://autobuild.buildroot.org/results/f3f3cc216ae42bb8a8925b0df7c1a3cc79b027d7

/home/buildroot/autobuild/instance-1/output-1/build/openfpgaloader-849e5751e06d4d00f323205d5f02ee01f9f59a61/src/spiFlash.cpp:
In member function 'void SPIFlash::jtag_write_read(uint8_t, uint8_t*, uint8_t*, uint16_t)':
/home/buildroot/autobuild/instance-1/output-1/build/openfpgaloader-849e5751e06d4d00f323205d5f02ee01f9f59a61/src/spiFlash.cpp:92:43:
error: variable-sized object 'jtx' may not be initialized
  uint8_t jtx[xfer_len] = {reverseByte(cmd)};
                           ^
and

src/gowin.cpp:73:11: error: 'runtime_error' is not a member of 'std'
 throw std::runtime_error("both write-flash and write-sram can't be set");
       ^
src/gowin.cpp:81:10: error: 'runtime_error' is not a member of 'std'
  throw std::runtime_error("incompatible file format");

Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>
[yann.morin.1998@free.fr:
  - don't add a sub-option for udev; directly rely on udev being avail
  - fix conflict after 1ca0077d9141
]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

docs/manual/adding-packages-cargo.txt: drop debug profile

There is no debug profile on cargo. The available profiles are: dev
(enabled by default), release, test and bench.

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/xen: add upstream security fix for XSA-327

Fixes the following security issue:

CVE-2020-15564: Missing alignment check in VCPUOP_register_vcpu_info

For further details, see the advisory:

https://xenbits.xenproject.org/xsa/advisory-327.html

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

support/tests: add runtime test for python-rpi-gpio

Modeled after similar python packages.

However, this one is picky, and throws an exception when it
detects that it is not running on a Raspberry Pi. So we just
catch that exception and check this is what we expect.

Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Christian Stewart <christian@paral.in>
Cc: Michael Fischer <mf@go-sys.de>
Cc: Asaf Kahlon <asafka7@gmail.com>
Cc: Yegor Yefremov <yegorslists@googlemail.com>
Cc: Ian Haylock <haylocki@yahoo.co.uk>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/python-rpi-gpio: fix gcc-10 compatibility patch

Although the patch makes the package build OK, it fails at runtime
when the module is imported, because of missing symbols:

    ImportError: /usr/lib/python3.8/site-packages/RPi/_GPIO.cpython-38-aarch64-linux-gnu.so: undefined symbol: high

Fix that by making sure the symbols are declared once, but only once.

Fixes: #13166
Reported-by: Christian Stewart <christian@paral.in>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Michael Fischer <mf@go-sys.de>
Cc: Asaf Kahlon <asafka7@gmail.com>
Cc: Yegor Yefremov <yegorslists@googlemail.com>
Cc: Ian Haylock <haylocki@yahoo.co.uk>
Tested-by: Christian Stewart <christian@paral.in>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/qt5: needs host gcc >= 5.0 for full C++11

Building qmake requires full C++11, which boils down to gcc >= 5.0,
which is what upstream advertises as a requirement anyway:

    https://doc.qt.io/qt-5.15/supported-platforms.html

    Distribution   | Architecture     | Compiler
    Generic Linux  | x86 and x86_64   | GCC (5 or later), ICC 18.x

Fixes:
    http://autobuild.buildroot.org/results/c3e/c3ee971a72f268e72b69a647e8fd00a8cee7dc91/
    http://autobuild.buildroot.org/results/89c/89c9a88b4e1195e952528574263201d4fbc27570/
    [...]

Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Peter Seiderer <ps.report@gmx.net>
Cc: Julien Corjon <corjon.j@ecagroup.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/tpm2-abrmd: bump to version 2.3.3

Bugfix release with a single fix:

Fixed:
  - Fixed handle resource leak exhausting TPM resources.

https://github.com/tpm2-software/tpm2-abrmd/releases/tag/2.3.3

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/libcurl: security bump to 7.72.0

This new version fixes, amongst many other things, CVE-2020-8231
(https://curl.haxx.se/docs/CVE-2020-8231.html). See the full changelog
on https://curl.haxx.se/changes.html#7_72_0 .

Also drop the 4 patches, that have all been released upstream.

Signed-off-by: Titouan Christophe <titouan.christophe@railnova.eu>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/openfpgaloader: C++ dependency is not inherited

openfpgaloader is written in C++, so the dependency on C++ is not
inherited from libftdipp1.

Drop the confusing comment.

Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/php: drop BR2_PACKAGE_PHP_EXT_HASH

hash extension can't be disabled since version 7.4.0 and
https://github.com/php/php-src/commit/bf344425812b0f6156d0a8a54ed7bc38054f7636

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/php: drop --with-libxml-dir

--with-libxml-dir has been dropped since version 7.4.0 and
https://github.com/php/php-src/commit/29d1b7fd521af288e9f04d784f59a2d15b494a30

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/php: ffi needs dynamic library

FFI support in php has been added with commit
e16f05c0f00081d446ebee56d9f3ad3cf41da1ae, it depends on dynamic library
otherwise the build will fail on:

/home/buildroot/autobuild/instance-1/output-1/host/opt/ext-toolchain/bin/../lib/gcc/arm-buildroot-linux-uclibcgnueabi/5.5.0/../../../../arm-buildroot-linux-uclibcgnueabi/bin/ld: ext/ffi/ffi.o: in function `zim_FFI_cdef':
ffi.c:(.text+0xe78): undefined reference to `DL_LOAD'

Fixes:
 - http://autobuild.buildroot.org/results/3380b7b6777d1c5fcb53c855b003466fa3bf2079

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/bind: security bump to version 9.11.22

Fixes the following security issues:

CVE-2020-8622: A truncated TSIG response can lead to an assertion failure
https://kb.isc.org/docs/cve-2020-8622

CVE-2020-8623: A flaw in native PKCS#11 code can lead to a remotely
triggerable assertion failure in pk11.c
https://kb.isc.org/docs/cve-2020-8623

CVE-2020-8624: update-policy rules of type "subdomain" are enforced incorrectly
https://kb.isc.org/docs/cve-2020-8624

For more details, see the release notes:
https://downloads.isc.org/isc/bind9/9.11.22/RELEASE-NOTES-bind-9.11.22.html

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/chrony: security bump to version 3.5.1

Fixes the following security issues:

CVE-2020-14367: Insecure writing of pidfile
-------------------------------------------

When chronyd is configured to save the pidfile in a directory where the
chrony user has write permissions (e.g. /var/run/chrony - the default
since chrony-3.4), an attacker that compromised the chrony user account
could create a symbolic link at the location of the pidfile to make
chronyd starting with root privileges follow the symlink and write its
process ID to a file for which the chrony user doesn't have write
permissions, causing a denial of service, or data loss.

This issue was reported by Matthias Gerstner of SUSE.

For further details, see the oss-security posting:
https://www.openwall.com/lists/oss-security/2020/08/21/1

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/qt5/qt5imageformats: fix build with jasper

Fixes:
 - http://autobuild.buildroot.org/results/ea3bc0d3110cb54421ecf433317ca79ca0ff834a
 - http://autobuild.buildroot.org/results/6831dfdd1ecaaa3812de481882040c1742a72aec

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

docs/manual: fix typo

depednencies -> dependencies

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/owfs: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/7e9c33d7a1613826f85716403a00dce03dd8daf2

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/mjpg-streamer: disable opencv python and cpp example

python-numpy is an optional dependency since
https://github.com/jacksonliam/mjpg-streamer/commit/545c1a1d3abf72bb7a97c870c78e85dd0845f63a

This dependency will raise the following build failure if python-numpy
is found on the host:

aarch64_be-linux-gnu-g++: ERROR: unsafe header/library path used in cross-compilation: '-I/usr/lib/python3/dist-packages/numpy/core/include'

because mjpg-streamer uses the python executable to retrieve the include
directory:

-- Found PythonLibs: /home/peko/autobuild/instance-0/output-1/host/aarch64_be-buildroot-linux-gnu/sysroot/usr/lib/libpython3.8.so (found version "3.8.5")
-- Found PythonInterp: /usr/bin/python3.5 (found version "3.5.3")
-- Found NUMPY: /usr/lib/python3/dist-packages/numpy/core/include

So disable python filter as well as cpp example while at it

Fixes:
 - http://autobuild.buildroot.org/results/e6795fa8ed5d8514b3e10ea1135afb31eab22b7e

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/libubox: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/aef10cc43f6c34f106624588ae2a1131520ee066

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/ncftp: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/78822bc0e0039e8f8949011a256cac022863276f

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/gstreamer1/gst1-plugins-bad: fix deactivation of opencv

Build can fail if opencv3 is built before gst1-plugins-bad because
-Dopencv=disabled does not work in meson (i.e. since commit
5d6c408e9535be4b5ba0c8a49215d2d03dcb5cd5)

Fixes:
 - http://autobuild.buildroot.org/results/19605057c4956d97e9e65068680485db637282db

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/collectd: fix build with luajit

Build with luajit has been enabled with commit
911a9bf5739d94545cc482c765f876a9ff73e304 however the build fails because
collectd does not find liblua

Fixes:
 - http://autobuild.buildroot.org/results/31302b47fb70eb442cd000f4b9f27a9e12432cc3

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/igd2-for-linux: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/f296984c3851fc28341210e36ef1b55b2edac209

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/dovecot-pigeonhole: bump version to 0.5.11

Release notes:
https://dovecot.org/pipermail/dovecot-news/2020-August/000439.html

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/efl: fix build without NLS

nls is enabled by default in meson_options.txt (which is available since
version 1.22.0). NLS was explicitly disabled for host-efl in commit
6deaa3d50de166ce187e8b15c70394a8eb844c03 but nothing was done for the
target resulting in the following build failure:

../src/lib/elementary/elm_priv.h:189:25: warning: implicit declaration of function '_elm_dgettext'; did you mean 'dgettext'? [-Wimplicit-function-declaration]
 #  define E_(string)    _elm_dgettext(string)
                         ^~~~~~~~~~~~~

/nvme/rc-buildroot-test/scripts/instance-0/output-1/host/opt/ext-toolchain/bin/../lib/gcc/x86_64-buildroot-linux-uclibc/8.3.0/../../../../x86_64-buildroot-linux-uclibc/bin/ld: src/lib/elementary/libelementary.so.1.24.3.p/elc_hoversel.c.o: in function `_access_state_cb':
elc_hoversel.c:(.text+0x1210): undefined reference to `_elm_dgettext'

So enable nls dependening on BR2_SYSTEM_ENABLE_NLS and add
TARGET_NLS_DEPENDENCIES to EFL_DEPENDENCIES

Fixes:
 - http://autobuild.buildroot.org/results/5985b39b87c5c392b4f9b65c12008ec7b3a143fe

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/gstreamer1/gstreamer1-editing-services: disable introspection

Disable introspection as it raises a build failure with autotools.
It could be enabled after bumping to upcoming 1.18.x and switching to
messon. We can't switch now as version 1.16.2 don't allow to disable
examples/tools through meson

Fixes:
 - http://autobuild.buildroot.org/results/36eb875fba2847b32df05f31d8f8ca9f0ecde36f

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/c-periphery: fix build with kernel 4.6 and 4.7

Fixes:
 - http://autobuild.buildroot.org/results/c3b868c12baac9438b792ada105c0b0de0106311

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/elixir: fix host-erlang dependency

There is no target elixir package, so setting a value to
ELIXIR_DEPENDENCIES has no effect, HOST_ELIXIR_DEPENDENCIES must be
used instead.

Fixes:

  http://autobuild.buildroot.net/results/a3a37eb724ca5689f8e83c9b2af04d07afa80315/

Signed-off-by: Frank Vanbever <frank.vanbever@essensium.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/dieharder: fix build with gcc 10

Disable -fno-common with dieharder, there is more than 100 variables to
fix and upstream seems dead.

Fixes:
 - http://autobuild.buildroot.org/results/ba70d111cd1f2029a193a88af3b44daf6ef27786

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/dump1090: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/88dc97fcaa649014edb3b54a5dd4bd8ec4715bbd

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>

package/gdbm: fix build with gcc 10

Fixes:
 - http://autobuild.buildroot.org/results/39c405096908e1d15f2462b990717215bea0750f

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>