asterisk uses backtrace() from execinfo.h to print beuatiful backtraces
when it fails. But execinfo is only reliably available with glibc, while
it is optionbal in uClibc and completely missing in musl.
So, only enable backtrace support for glibc toolchains.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Asterisk: the flagship of telephony on Linux. These are the lines of
code whose continuous mission is to power small and large enterprises
telephony systems, to boldly provide IP PBX where no one has done so
before.
But it is a hell to get compiled... :-(
For starters, it needs a host tool, menuselect, to prepare its build
configuration. Unfortunately, the way it handles menuselect does not
apply very well for cross-compilation: the main ./configure calls out to
menuselect's own ./configure, and of course that runs with the same
environement, which is wrong for cross-compilation (because of variables
like CC, CFLAGS and the likes).
Furthermore, the paths to menuselect are imbricated about everywhere in
the main Makefile, so making it find menuselect in PATH is a lost cause.
Instead, we just patch-out the handling of menuselect, build it as the
host variant and copy it in place.
Now, asterisk wants to install a default set of sound files (for
answering machine stuff, I guess). They come come pre-bundled in the
official archive [0], but the buildsystem will want to download (at
install time) the sha1 files for each sound archive, to validate that
said archive is correct. However, the download is done via plain http,
so it still risks an MITM attack. And for Buildroot, it is not always
possible to download at install time, so we patch-out the sha1 check.
[0] http://downloads.asterisk.org/pub/telephony/asterisk/releases/
The official archive contains the sound archives plus a full set of
documentation. This makes it very big. Unfortunately, the hosting site
is rather slow, topping at about ~204kbps. So we get the archive from
the official mirror on Github. But that archive is missing the sound
archives, so we download them separately.
Some tests, like the crypt() one, are broken and could not have ever
possibly worked at all. Worse, the FFmpeg test is looking for headers
that FFmpeg removed more than 10 years ago and are virtually no longer
available in any distro. So, FFmpeg support is definitely not tested
by upstream and can't possibly work at all. Finally, trying to run
test-code does not work in cross-compilation.
As a final stroke of genius, asterisk checks for the re-entrant variant
of res_ninit(), and concludes that all such functions are available,
including res_nsearch(). Uclibc-ng has the former but not the latter, so
the build fails. Since there is no cache variable for that check, we
can't pre-feed that result to configure, and fixing it is a bigger
endeavour. So we make asterisk depend on glibc for now, until someone
is brave enough to fix it.
Almost all features are disabled for now. Support for additional
features will be added in subsequent patches now that we have a working
base.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Romain Naour <romain.naour@openwide.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
[Arnout:
- make libilbc a mandatory dependency instead of using the bundled one;
- add license, license files, and license file hashes;
- minor spelling corrections;
- remove redundant trailing backslash reported by check-package;
- rewrap help text to 72 columns instead of 68]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
fixup
MPD version 0.20 is affected by GCC bug 64735 and failes to build for
some archictuctures (NIOSII, ARMv4, ARMv5 and SPARCv8) with a GCC
toolchain less then version 7.
We added a version choice between 0.19 and 0.20 in Buildroot version
2017.02 as GCC 7 was not yet available at that time. This way, mpd could
still be build for those architectures.
As GCC 7 is now available in Buildroot remove the version choice for 0.19.
Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Removed patch after upstream committed a different solution:
https://sourceforge.net/p/omniorb/svn/6330/
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
[Peter: board is from Sinovoip, not Lemaker, move board files,
Use kernel headers from kernel, explicitly use 4.13.3,
drop eth0 dhcp config, cleanup post-build args,
disable tar output format, use u-boot custom format,
add host-dosfstools/mtools for genimage, drop unused
host-uboot-tools/genext2fs]
Signed-off-by: Mike Harmony <mike.harmony@snapav.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Only a static library is built (and no _INSTALL_TARGET_CMDS are defined), so
there is no need to run the target-install step.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
NS_ENABLE_SSL was renamed to MG_ENABLE_SSL in v6.0, commit e1dd3f06fe33
(Rename Mongoose constants: NS_ -> MG_, NSF_ -> MG_F_), so use the new name
instead.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
No linking is done since the package was reworked for v6.x in commit
9860746ff (mongoose: bump to version 6.1), so drop LDFLAGS and the list of
libraries to link against.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This define is not used anywhere in the mongoose sources since v6.0 (commit
8927c9d22b3f), so drop it.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Removed patch after upstream committed a similar patch:
946596172d (diff-35a172a9ca8faa1683fc747ca94681b3)
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Add upstream patch fixing CVE-2017-14062:
Integer overflow in the decode_digit function in puny_decode.c in
Libidn2 before 2.0.4 allows remote attackers to cause a denial of
service or possibly have unspecified other impact.
This issue also affects libidn.
Unfortunately, the patch also triggers reconf of the documentation
subdirectory, since lib/punycode.c is listed in GDOC_SRC that is defined
in doc/Makefile.am. Add autoreconf to handle that.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Bump U-Boot to 2017.09 version and kernel to 4.13.3.
Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes the following security issues:
CVE-2017-2862 - An exploitable heap overflow vulnerability exists in the
gdk_pixbuf__jpeg_image_load_increment functionality of Gdk-Pixbuf 2.36.6. A
specially crafted jpeg file can cause a heap overflow resulting in remote
code execution. An attacker can send a file or url to trigger this
vulnerability.
CVE-2017-2870 - An exploitable integer overflow vulnerability exists in the
tiff_image_parse functionality of Gdk-Pixbuf 2.36.6 when compiled with
Clang. A specially crafted tiff file can cause a heap-overflow resulting in
remote code execution. An attacker can send a file or a URL to trigger this
vulnerability.
CVE-2017-6311 - gdk-pixbuf-thumbnailer.c in gdk-pixbuf allows
context-dependent attackers to cause a denial of service (NULL pointer
dereference and application crash) via vectors related to printing an error
message.
The host version now needs the same workaround as we do for the target to
not pull in shared-mime-info.
Also add a hash for the license file while we're at it.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
It seems github now sometimes provides slightly changed tarballs which
produce a different sha256 hash than before. This commit fixes the
hashes of the packages when sources.buildroot.net does not store a copy
of the tarball matching the current hash.
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Install tools into /usr/bin, not /usr/local/bin, since /usr/local/bin
is not in PATH by deafault.
[Peter: set PREFIX to /usr to really install into /usr/bin]
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Tested-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Bump Python3 version to 3.6.2.
Patches dropped:
"Support PGEN_FOR_BUILD and FREEZE_IMPORTLIB_FOR_BUILD"
Rationale: With commit 9d02f562961efd12d3c8317a10916db7f77330cc, code
generation step of building CPython now became explicit (instead of
always performed as a part of 'make' invocation) and more granular. We
no longer need to use Parser/pgen at all and tricking the build system
into using different Programs/_freeze_importlib can be done as a part
of recipe.
Additional info about the build change can be found at
https://bugs.python.org/issue23404
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Tested-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
It seems github now sometimes provides slightly changed tarballs which
produce a different sha256 hash than before, this is the case for
tvheadend. Instead of fixing the hash lets bump to current HEAD and
hope that the new hash will be valid for a longer period of time ;)
Fixes
http://autobuild.buildroot.net/results/59a/59ad7940dcd498cd0aaec26c8497b21e64379a6f/
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Also add a hash for the license file while we're at it.
[Peter: use tuxfamily.org URL for annoucement]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
