It's not actually being built by any job, and its contents are anyway
wrong since it basically pulls in the same as the sysmocom-nitb-image.
Initially it made sense to have these 2 images because we either
connected the BTS to osmo-bsc or osmo-nitb. This is no longer the case
since we deprecated osmo-nitb, and now it always connects to osmo-bsc.
The initiall idea after osmo-nitb split was to have a sysmo-bsc-image
containing only osmo-bsc, and then a sysmocom-nitb-image superseeding
it by installing all other core network components such as hlr, msc, etc.
Since there's no real need for that now, and in any case we can just
duse the nitb one and disable the extra services, let's drop the bsc
one.
This way we have it too in sysmocom-nitb-image, but only when needed,
that is: when we build for a machine inheritinh from sysmocom-bsc, which
is a machine not containing any BTS HW and thus requires a BTS connected
to it, so it needs a DNS server to serve it.
According to MACHINEOVERRIDES, alix and apu2 are subclass machines of
sysmocom-bsc (abstract superclass). However, only alix requires
sysmocom-bsc.conf. Make both alix and apu2 requires sysmocom-bsc.conf,
and move alix specific stuff to sysmocom-alix.conf.
In the process, linux-firmware-rtl-nic is removed from RDEPENDS for
apu2, since it doesn't use RTL nic but e100e/igb ones.
As a result of these movements, osmo-bsc.conf is actually empty, but
it's left there for future additions which have to affect both machines:
ie. adding stuff which is expected for "bsc" role on machines containing
no BTS.
After upgrading to poky pyro, generation of grub2 menuentry in grub2
became broken. Our init-install.sh script, origianlly coming from an old
poky version, was modified in poky
c2e63957d5abaa4bb10507da5ba16f01112d0bf5 to adapt to the fact that
/etc/grub.d/40_custom was not there anymore. We need to do the same,
otherwise the menyentry is not appended to the grub config and the
installed image won't boot correctly.
poky script ./meta/recipes-core/udev/udev-extraconf/mount.sh is called
by systemd-udev rules every time a storage device is found, and it is
responisble to automount them. The init script waits until the device is
automounted to finish the boot process.
The mount.sh script sues "stat", and right now our busybox doesn't have
it enabled. As a result, stat is not found, the script fails, the
device is not mounted, and the init script hangs forever.
Since osmo-msc a2353c69cc3d2ccc7caee4571b3b8d2a98961150 and
osmo-bsc 3561bd48976dbee8dbd4659dad15be85a3e79ace osmo-bsc-mgcp support
is dropped in favour of osmo-mgw and one instance of osmo-mgw can be
shared between osmo-msc and osmo-bsc.
This patch is not longer used since the complete declaration of the
function was removed from the header file in osmo-bts
085569df55ceb3737b79f53b015b0c064698ddc6.
osmo-mgw recipe in meta-telephony has been updated to generate several
packages. Right now, osmo-bsc requires osmo-mgw service, but osmo-msc
still requires osmo-bsc-mgcp, so let's add both to the image.
After switch to linux 4.9.14, make oldconfig run by OE was disabling
CONFIG_IP_NF_TARGET_MASQUERADE even if original defconfig had it enabled
as a module. This target is required for osmo-ggsn.
This new package group contains what used to be the "task-sysmocom"
group, with an osmo-nitb. In turn, task-sysmocom.bb will be changed to
use new git repositories.
According to documentation, they are deprecated by
KERNEL_MODULE_AUTOLOAD:
WARNING: linux-sysmocom-4.9.14+gitAUTOINC+8d5d275254-r1 do_package: module_autoload_leds-gpio was replaced by KERNEL_MODULE_AUTOLOAD for cases where basename == module name, please drop it
After switch to linux 4.9.14, make oldconfig run by OE was disabling
CONFIG_IP_NF_TARGET_MASQUERADE even if original defconfig had it enabled
as a module. This target is required for osmo-ggsn.
WARNING: osmo-pcu-0.4+gitAUTOINC+0cd8e4eade-r5.18 do_configure: QA Issue: osmo-pcu: configure was passed unrecognised options: --enable-sysmocom-bts [unknown-configure-option]
Looking at ./configure --help in osmo-pcu, that option doesn't seem to
be available. The recipe is already using --enable-sysmocom-dsp too,
which is correct.
Commit d32cbbb13087eae2617920d4ea980f0e8869a89e changed --enable-sysmocom-bts to --enable-osmo-bts.
Later on, commit 08e5d604d3fce75b955549244b36fde62f20894b removed --enable-osmo-bts option.
In the past I copy and pasted the compiler-gcc4.h to have gcc-5.h and
gcc-6.h suffix. I didn't want to do the same for gcc7 and cherry picked
the original commit. Not compile tested though.