A lot of changes upstream since the last upgrade, the highlights:
* MQTT v5 protocol support
* Performance improvements
* New mosquitto_rr (request/response) client
* TLS enhancements: OCSP stapling, TLS Engine support, explicit TLS v1.3
support, removed TLS v1.0 support
For further details, see https://mosquitto.org/blog/
Additionally, SUMMARY and DESCRIPTION were tidied up as needed.
Based upon an earlier upgrade patch by Oleksandr Kravchuk <open.source@oleksandr-kravchuk.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
There have been a couple of releases since we last upgraded, including
a number of additional features. Changelog is here:
https://github.com/npat-efault/picocom/releases
Also tweak SUMMARY - I don't think this has been 20k in a while; with
this upgrade the binary alone went from 60k to 70k.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Notable changes since 0.7.3:
* Fixed CVE-2018-10933 authentication bypass
* Removed support for deprecated SSHv1 protocol
* Added support for OpenSSL 1.1
* Added support for chacha20-poly1305 cipher
* Added ECDSA support with gcrypt backend
* Improved threading support (note: libssh_threads is now gone)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
A dependency on python-protobuf was added in commit
5f6fcfd362 for the sole purpose of
providing a ptest, however python-protobuf is in meta-python and thus
this means that meta-oe would depend on meta-python by default (assuming
your distro enables ptest by default), and we don't want that - meta-oe
isn't supposed to depend upon any layer other than openembedded-core.
Luckily we can still have a ptest even without the python support, so
add a PACKAGECONFIG and leave it disabled by default.
Note: the PACKAGECONFIG here is not particularly useful since it's only
about what goes into the -ptest package and thus also the dependency. I
contemplated just using LANG_SUPPORT instead, but PACKAGECONFIG does
have the advantage that it's introspectable and fairly well understood
so in the end I went with it.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This includes one security fix and numerous other bugfixes. MQTT version
3.1.1 is now the default protocol.
See the following for details:
https://mosquitto.org/blog/2018/08/version-151-released/https://mosquitto.org/blog/2018/05/version-1-5-released/
Changes to the recipe:
* Drop explicit installation of the libmosquitto.a static library
because this no longer gets built by default, and the normal
"make install" which we are running will install it anyway if it were
to be enabled.
* Drop our service file since there is one supplied in the source tree
* Enable systemd notify support when systemd is being used
* Update comments in the init script
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Commercial usage requires a license according to the homepage, so set
LICENSE_FLAGS = "commercial" to indicate that.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Commercial usage requires licensing according to the homepage, so set
LICENSE_FLAGS = "commercial" to indicate that.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The following improvements have been made over the recipe that was in
meta-intel-iot-middleware (a layer which is no longer actively
maintained):
* Upgrade to latest upstream version (765.50.9)
* Fix compilation failures by passing CC and LD
* Fix "no GNU hash in the ELF binary" issue
* Point S at the correct source subdirectory so that "make clean" works
* Fix lack of soname on libdns_sd.so leading to missing RDEPENDS QA error
* Add SUMMARY
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Joe MacDonald <joe_macdonald@mentor.com>
Make this recipe a little more widely visible (has previously appeared
in meta-refkit-core and meta-intel-iot-middleware). Also make couple of
minor tweaks:
* Fix BINDINGS example to use the correct recipe name (copy-paste error)
* Set HOMEPAGE
* Drop AUTHOR as this is field is only really intended for older
software where there is no real point of contact for upstream other
than an author's email address.
* Set same COMPATIBLE_HOST as mraa since upm depends upon it
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Make this recipe a little more widely visible (has previously appeared
in meta-refkit-core and meta-intel-iot-middleware). Also make couple of
minor tweaks:
* Make SUMMARY more accurate by borrowing the first part of the short
description now in the mraa github repo
* Set HOMEPAGE
* Drop AUTHOR as this is field is only really intended for older
software where there is no real point of contact for upstream other
than an author's email address.
* Set COMPATIBLE_HOST to match what is checked by mraa's CMakeLists.txt
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The following improvements have been made over the recipe that was in
meta-intel-iot-middleware (a layer which is no longer actively
maintained):
* Upgrade to 1.0.3
* Inherit from setuptools instead of distutils since that changed with
the upgrade
* Inherit pypi and also provide Python 3 version
* Use the correct LICENSE (changed BSD -> MIT back at version 0.9.5)
* Set short description in SUMMARY instead of DESCRIPTION
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The following improvements have been made over the recipe that was in
meta-intel-iot-middleware (a layer which is no longer actively
maintained):
* Upgrade to 1.2.0. The version number being used previously (3.1) was
bogus - I guess someone saw that it supported the MQTT *protocol*
version 3.1 and assumed that was the library version. Since we are
renaming it to paho-mqtt-c at the same time to distinguish it from
other Eclipse Paho MQTT library flavours, we don't need to bump PE.
* Fix SRC_URI since the repository has been moved
* Fix "No GNU_HASH in the elf binary" QA error
* Add notice.html and about.html to LIC_FILES_CHKSUM since they cover
licensing
* Add description, Signed-off-by and Upstream-Status to makefile.patch
* Set short description in SUMMARY
* Drop unnecessary custom do_compile - the default implementation
basically does the same thing
* Drop RDEPENDS on openssl - linking already ensures runtime
dependencies get set correctly
* Move DEPENDS further up
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The following improvements have been made over the recipe that was in
meta-intel-iot-middleware (a layer which is no longer actively
maintained):
* Upgrade to 1.4.14
* Use correct LICENSE value (license changed back in version 1.4)
* Add files containing the actual license terms to LIC_FILES_CHKSUM
* Make optional dependencies optional through PACKAGECONFIG (c-ares
dependency now defaults to disabled)
* Use ${prefix} instead of /usr
* Drop python package since the python client was removed in 1.4
* SUMMARY and DESCRIPTION now reflect that it also supports MQTT 3.1.1
* Add brief description and Upstream-Status to build.patch
* Drop PR = "r0"
* Drop unnecessary setting of FILES_${PN}-staticdev since the default
already includes this file
* Move SRC_URI checksums next to SRC_URI
* Move inherit of systemd further up
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
During do_compile if setup.py finds that pytest-runner - which is
declared as a setup dependency in setup.py - isn't available it was
attempting to download it from pypi, which is problematic if you are
behind a proxy because the environment for do_compile isn't set up for
that. Add an appropriate native dependency to prevent this situation.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* Make LICENSE more accurate - it's dual licensed
* Include the actual license text files in LIC_FILES_CHKSUM
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Updated version of the recipe in meta-openstack. Also made LICENSE more
accurate and set SUMMARY instead of DESCRIPTION since the value is
short.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
We rely on the host sed all over the place; we don't need sed-native
here just because it's being used in do_configure.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Among other things, includes a fix for CVE-2014-8132.
(There are a couple of useful-looking fixes after 0.6.4 in the 0.6
branch, hence I have renamed the recipe to _git and set PV to the
standard format.)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Allow imap support to be easily disabled; default it to enabled based on
previous configuration.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Add a pam PACKAGECONFIG option defaulted from the pam DISTRO_FEATURES
item, so that if pam isn't in DISTRO_FEATURES, it is explicitly
disabled (and conversely if it is in DISTRO_FEATURES that we have an
explicit dependency on libpam).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Fix several issues with how the phar package is installed and packaged:
* The symlink was pointing to an absolute path on the host, which
naturally does not work on the target
* The script itself had a shebang pointing to a path on the host
* A file-rdeps warning was occurring due to rpmdeps detecting that the
script requires the php binary; in order to prevent the main php
package depending on php-cli which provides the php binary, move the
phar script to its own package (I suspect most people won't need it
anyway).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Among many fixes since 5.5.15, the following security issues were fixed:
CVE-2014-8142
CVE-2014-3710
CVE-2014-3669
CVE-2014-3670
CVE-2014-3668
CVE-2014-3538
CVE-2014-3587
CVE-2014-2497
CVE-2014-5120
CVE-2014-3597
For more details please see the upstream changelog:
http://php.net/ChangeLog-5.php
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Build the mod_php Apache module within the main php recipe; this avoids
files in the sysroot from stepping on eachother and therefore avoids the
resulting build failure.
Changes made:
* Add an "apache2" PACKAGECONFIG, default disabled. This means that if
you previously used the modphp recipe you now need to set
PACKAGECONFIG for php to include "apache2".
* Bring in most patches and configure options from the modphp recipe.
* Specify ${libdir}/php5 as libdir. This avoids the PHP extensions
directory being /usr/lib/extensions by default which is ugly.
* Tidy up some superfluous/malformed entries in FILES_${PN}-pear
* Drop obsolete PHP_MAJVER
* Replace virtclass- override with class-
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
* Upgrade to 2007f
* Rename to uw-imap; this is the name used by Fedora and is much more
sensible than libc-client
* Fix errors when linking the library against code compiled with -fPIC
such as mod-php (patch from Fedora).
* Add a security patch from Fedora
* Rename files directory to uw-imap
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Based on the corresponding changes to the modphp recipe in
meta-webserver. From the commit message for the modphp upgrade:
* LIC_FILES_CHKSUM needed to change because the copyright year changed
* Enabled mcrypt since upstream recommend this for acceptable
performance
* Disabled the opcache; this is a new feature in 5.5 and the configure
check for it currently breaks when cross-compiling.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
* License changed to GPLv2+
* Trim SUMMARY
* gz file is definitely being provided for download upstream, so drop
comment saying it isn't
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* Newer automake compatibility fixed upstream, so drop patches.
* LIC_FILES_CHKSUM changed due to a trailing space being removed, no
actual change to the license text.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Joe MacDonald <joe_macdonald@mentor.com>
The ntp-utils package contains at least one perl-using script as well as
a supporting perl module, therefore we need a dependency on perl.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Joe MacDonald <joe_macdonald@mentor.com>
After the upgrade to 4.2.8, ntp's configure process now uses a custom
script which looks at the host to determine what install locations it
should use. This resulted in the recipe working on some people's
machines and failing during do_install on others. Force it to use the
"redhat" configuration as this seems closest to what we used to be
using prior to the upgrade (this means that binaries are now back in
sbindir as they used to be).
Thanks to Philip Balister for reporting this.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Joe MacDonald <joe_macdonald@mentor.com>
* Upgrade to 4.2.8 which fixes several security issues, including
CVE-2014-9293, CVE-2014-9294, CVE-2014-9295, and CVE-2014-9296. For
more details please see:
https://ics-cert.us-cert.gov/advisories/ICSA-14-353-01A
* LIC_FILES_CHKSUM changed due to a number of copyright year and patch
list changes; nothing material about the license text changed.
* This version moves a number of binaries from sbindir to bindir;
there's supposed to be a configure option --with-locfile=legacy to use
the old layout but it does not seem to work. I guess we'll just have
to live with the change.
* Drop patches which are no longer applicable.
* Merge inc file into recipe; there were too many changes required to
the inc file in this version and it's unlikely it was much use split
out in any case.
* Move remaining files in files/ to ntp/
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Joe MacDonald <joe_macdonald@mentor.com>
systemd support has already been added to the OE-Core at recipe based on
this file (in the dizzy branch).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
We already patch configure.ac and we're not bypassing autoreconf,
so we don't need to patch configure as well.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
* Upstream switched from autotools to cmake
* Separate build dir and parallel make now work
* Library versioning no longer used for plugins; other libtool cruft
gone
* Proper upstream initscript
* Plugins moved from client library package to mariadb-server package
(matches how Fedora packages these)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Introducing the aufs-util target recipe. This would be used in
conjuction with the aufs kernel feature to provide union mounts suitable
for live boot applications.
Only the target binaries are currently packages, with man pages,
libraries and other artifacts skipped in this initial integration.
(Integration work done by Bruce Ashfield <bruce.ashfield@windriver.com>
and Chen Qi <Qi.Chen@windriver.com> .)
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* Use correct LICENSE value to match up with common license file (and
avoid corresponding warning)
* Set SUMMARY instead of DESCRIPTION
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Add the Xdebug license file to avoid a missing generic license file
warning during building.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
* LIC_FILES_CHKSUM changed because of the introduction of an extra blank
line in the LICENSE file (!)
* Refreshed TLS Next Protocol Negotiation support patch for conflict
with 2.4.7. Thanks to Hongxu Jia for doing this work.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
In recent versions, upstream has decided to place additional
restrictions on commercial use beyond a standard open source license
(LGPLv3) [1]. This makes it hard to set a LICENSE value that is easily
understood. Of course, as the authors, they have the right to decide
what licensing terms they wish to distribute their project under, and we
could always set LICENSE_FLAGS to denote the extra terms, but this is
somewhat messy and personally I feel less inclined to continue
maintaining this recipe in meta-webserver now, especially since I
originally put it together on my own time. At the moment due to a
branch/commit mismatch it is no longer fetching in any case.
(If someone wants to resurrect this recipe in another layer, they are
more than welcome to do so.)
[1] http://support.ajenti.org/topic/351265-clarify-licensing/
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
We don't have a proper help2man so building the manpages fails; just
disable building them to fix this as we do elsewhere.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Joe MacDonald <joe@deserted.net>
Originally based on the recipe from OE-Classic. Changes from that
version:
* Remove packaging of gstreamer plugin that was removed in 1.09
* Inherit pkgconfig since pkg-config is being used
* Set more appropriate LICENSE
* Add LIC_FILES_CHKSUM
* Add SUMMARY
* Optimise out inc file since it doesn't seem particularly useful to
have one for this recipe
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Originally based on the recipe from OE-Classic. Changes from that
version:
* Update to 1.28
* Add a patch from Debian to fix compilation errors
* Split out staticdev files into their own package
* Set reasonable SUMMARY instead of DESCRIPTION
* Add LIC_FILES_CHKSUM
* Drop PRIORITY and default value of S
* Rearrange recipe into logical order
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The 0.3 release was a year ago and git HEAD contains fixes to make it
work with contiki, so build from git instead of backporting all of the
patches.
Based on a patch by Koen Kooi <koen@dominion.thruhere.net>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Joe MacDonald <joe@deserted.net>
This bbappend has effectively been merged into OE-Core, although the
log buffer size is the busybox default rather than 64K - layers may
change this either by providing their own /etc/default/busybox-syslog
file (when using systemd) or modifying the
CONFIG_FEATURE_IPC_SYSLOG_BUFFER_SIZE option in busybox's build time
config.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
This is now in OE-Core as of ad2851cf0abc2ab35e0f60c96d3142c29a07c8fc.
ca-certificates-cross is no longer used (effectively replaced by a native
BBCLASSEXTEND in the main recipe).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
This is now in OE-Core as of f72c0ff46d6895978e7e5230cb882d8eee4005ba
(libnotify3 remains).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Paul Eggleton