Age | Commit message (Collapse) | Author |
|
This patch updates QEMU version in master branch. I have tested building
all affected variants locally by running respective ./build.sh.
We intend on updating the version for master branch and then using this
with buildfarm-cmp job to compare against tcwd-tested branch.
Change-Id: I538dc01f8f4b931e018af1ec987059b8465b6dba
|
|
|
|
|
|
We need to first install software-properties-common, which contains the
add-apt-repository command. Then we can use that to add the toolchain test
builds PPA which contains gcc and g++ 7. Installing the g++-7 package is
enough to get both packages.
We also update run.sh to explicitly invoke gcc-7 and respectively g++-7 in
the ccache wrappers.
Change-Id: I6624cf31b0a7a135264eebf21230a6d0e7633a1b
|
|
The x2go images take a long time to build, and, with Ubuntu Bionic out,
there's not much use for the xenial one.
Also remove custom stanza to install software-properties-common.
Change-Id: I1ae7a0c2c783e481db706adf9026894700513d1c
|
|
This should make it available in tcwg-host, so we can run
generate-dockerfiles.sh there.
Change-Id: I4633b3afe9299370ac85bcd2a349ded8451baf91
|
|
Change-Id: I41ed38dde47a460bca83f4cd11f9ec84297439d7
|
|
... to allow building GCC without re-building these components inside ABE.
Change-Id: I1561fca965235001c9d4696e2018a4bfda54b531
|
|
... so that ccache volumes are mounted with correct permissions.
Change-Id: I858968cc8a825fed16e55acebd4d2b64747ade37
|
|
... so that we can throttle LLVM builds on Packet.net machines in
tcwg-build containers (e.g., in tcwg-rr-gnu-master job).
Change-Id: I94ff9adf9c17ec66d2d1dd715d38b541ebe0eb79
|
|
... which we use in tcwg-cleanup-stale-container.sh to kill stale
ssh-agent processes.
Change-Id: I2dfb1ccd9301ce284479e72011ceed07cabb277c
|
|
Change-Id: I8c994d79fac6df3f7930cc2404465ea9342859f6
|
|
... to save build time when changing new-user.sh script.
Change-Id: I1ce110494227c217241857aa590784e4f58dc2d9
|
|
Change-Id: Ibb7e338bcd22ae736a22bda617f66f08fcdfd04b
|
|
We now create directories for mount-points in start-container-docker.sh.
Change-Id: I4996616399ed87420dbd64eba36895f632f6af3f
|
|
Move passwd file to home-data/, and update tcwg-build to use it.
Change-Id: Ie409103ae4e47a8f9cf393a4052ddec92509a8af
|
|
Replace "new-user.sh --key KEY" interface with "new-user.sh --home-data DIR"
interface. Use the new interface in both tcwg-host and tcwg-build images.
Change-Id: I58e7e6506daa70bd10b4f29deebcf85ca0c5bc4f
|
|
Change-Id: Idd91d3b71b1980958895bb3e5a0c4a814bb1ed09
|
|
.. in preparation to move it to tcwg-base/
Change-Id: I4263bb1b5cde191b85118b5aff4b1773b7ffc1ab
|
|
... and use it to create users inside containers.
Change-Id: Ic00d13c30ad6e8a4a613040fa76537eec7e1ad97
|
|
ABE no longer requires "root" ssh access, it uses "sudo" now.
Change-Id: I62bd6b3d0bbce37a942d4328a70c0e0293708da0
|
|
... for the sake of jenkins-scripts.git:jenkins-helpers.sh.
Change-Id: I6418837d9101f2f39f4f38e60eefa1e94287bca7
|
|
Change-Id: I9bdda54997cc8a042a7af37a21af2c89ca1d6458
|
|
Turns out "dpkg --add-architecture i386" in amd64's tcwg-base images
was necessary to install wine. Since wine with its i386 dependencies
increase image size by ~600MB, install it only in tcwg-build images for
the purposes of automated release testing.
Hopefully, we will be able to ditch testing of mingw releases.
Change-Id: I0c177409c720cb07070d769d6ed693d1bfdc31df
|
|
Change-Id: Ia4fb466289f0087839f0d39a786be527ab9572e7
|
|
Change-Id: I2156b5bfd8bb9195b6b8aeeb995fbe2952436274
|
|
The configuration needs to be patched at runtime with the right
user/password, provided by Jenkins master.
Change-Id: I3b1d312f65be5c1b30bc9b5d9c42e051a6c19d36
|
|
... and use system's clang 3.8 for all other bots.
Change-Id: I427f8ce9d44e8d2322025734f37271fc0745ee56
|
|
Change-Id: Ie60ece7017af20c750c65e6c8bdb757cfdf79b86
|
|
Change-Id: I2c3e8268616363b16273cc1fbb6def9180806541
|
|
We need it in pre-commit upstream patch testing jobs to submit patches
to gerrit.
Change-Id: I5f29a179d41c8b92260466bcf6a30afabee42cc7
|
|
... to throttle new jobs when system RAM utilization nears 50%.
Also place all of /home on a scratch volume to allow fast access
to developer's files for investigations.
Change-Id: Ief8648a93793a2ffa99b561d983796cead3187f6
|
|
Change-Id: I509a1b89d2ac751efd57ecb5b7a48dbe5822d0a3
|
|
... to allow re-building of TCWG image on non-master branches.
See https://review.linaro.org/#/c/20116/ .
Change-Id: I23cdbaf3a6ff8b5ded6892938ab85c46ac7c3b38
|
|
... which turn out to be symlinks themselves for now, but we will now
have a dedicated build.sh script for every image type.
Change-Id: Id64c469d1aaf6e792d8e6e5b80c13af3c25732f6
|
|
Current tcwg-base/*/build.sh files could be used only by hand and,
therefore, are pretty useless. Get rid of them.
Change-Id: Ibd35529f733729345c504704684c7e696a03ff92
|
|
This patch adds tcwg-llvmbot images for Xenial arm64 and armhf.
The images are setup using llvm-config.git scripts. We need these
scripts to configure bare-metal LLVM bots, so it's not worthwhile
to port their functionality to Dockerfile.
Change-Id: I574cc4d2b49fb882042d81116653748fc1b7c0a8
|
|
The fact that tcwg-benchmark user is present only in amd64 Trusty tcwg-build
image is historical artifact. When we added the tcwg-benchmark we didn't
want to destabilize other images. The change has been extensively tested
since then, so enabling for all tcwg-build images.
Change-Id: Ic341a98405cc39ea6e6f36a5d24dd96b239eaa33
|
|
In order to avoid accidental modification of auto-generated
tcwg Dockerfiles, add checksum validation to those files.
This patch:
. adds comments to generated Dockerfiles indicating the
source file and the file's md5sum.
. regenerates all *-tcwg-* Dockerfiles (to add those comments)
. adds validation of the md5sum to the build-image.sh
. restricts search for Dockerfile.in to the tcwg-base hierarchy
Change-Id: I691ef4f1d93820bcdd7d50e19c11542a1db52875
|
|
Change-Id: Iccc264951f9c6979be3f23afe5770b9df04454d0
|
|
The harness implementing running QEMU VMs in docker containers is in
toolchain/jenkins-scripts/start-container-qemu.sh.
Rootfs of the container is shared via NFS with the VM. For NFS to work inside
container NFS kernel module needs to be loaded, which is easiest achieved
by installing nfs-kernel-server package on the host.
For KVM to work container needs to be started in --privileged mode
to provide /dev/kvm inside the container.
Change-Id: I7cba73e350dfbee6e4dde0797eb4bed90fe0c664
|
|
Add linux-tools-generic to get "perf" and binutils-aarch64-linux-gnu with
binutils-arm-linux-gnueabihf to enable cross perf-annotate on x86_64 boxes.
The latter is done by passing --objdump <objdump> option to perf.
Change-Id: Ib33aae197bc0b04f1c6de69bb3060a98e8de5854
|
|
TCWG now starts docker container using jenkins-scripts, and volumes
are handled there.
Change-Id: I97916b3f7b2b7c71c365013e93341831c2eb0dea
|
|
On Xenial, even python3-virtualenv isn't enough.
The virtualenv package provide the virtualenv binary.
Change-Id: Ie35bd08fdab3c5629c03633bf71309be097009e4
Signed-off-by: Fathi Boudra <fathi.boudra@linaro.org>
|
|
As demonstrated by the error in our deployment:
tcwg-llvm-testsuite.sh: line 76: virtualenv: command not found
On xenial, virtualenv depends on python3-virtualenv.
Change-Id: I1487c3b49a3be15be5de1c82fffa380d70f74238
|
|
virtualenv exists only in xenial and depends on python3-virtualenv.
Use python-virtualenv instead, which exists in both trusty and xenial.
Change-Id: Icbcbdeb7250b2286d72751491ccd145a6fea9f94
Signed-off-by: Fathi Boudra <fathi.boudra@linaro.org>
|
|
... to make room for build.sh that will re-build all affected images.
Change-Id: I765526adf18007adf738f804137f745eb6d7946e
|
|
Change-Id: I652eb0077a716cf1204aa952f388abbdaad51f58
|
|
Change-Id: Iadc5a62d0809268565e889053a230a63e9439594
|
|
These are copies of respective arm64 images with minimal changes.
Change-Id: I6f732221282bca666570f16282f19d3054efcf86
|