Age | Commit message (Collapse) | Author |
|
Run llvm-release job weekly to catch release issues due to changes in
LLVM upstream code. Parameter candidate is given the default of trunk so
that the weekly build tries to build trunk.
Change-Id: I531ffd111122834bcbed410cca3b80cb7a015f18
Depends-On: Ia01a877b62457b462e718ec8019bc8374dad6e5a
|
|
This reverts commit 54ea52ee9700f8e188507e05bd64ea818c4e6822.
Change-Id: I6e637741d45fd395ff505b0477ae93f54eb4e48c
|
|
This reverts commit f04dc8eca53ec3f42fb21c47aa1f13499ac29ef8.
Change-Id: I24619fc1f89b7bf5d50186a9620aa129d7f3ed98
|
|
Where start-container-docker.sh is called with hard-code "--distro trusty"
it is better to remove this setting, since it is default anyway.
Change-Id: Ia33e01e4b4b48ae221ab36e85a4c96c2b3f62865
|
|
The shell script nodename2hostname.sh is not successfully being
kept up to date, so use print_host_for_node() in jenkins-helpers.sh
so that nodename2hostname.sh can be removed.
Change-Id: Ice0f71a829d8c6001ad88d663346debfeb35b5a3
|
|
This is our first attempt at stabilising the production environment in
Jenkins. For now, we're only fixing jenkins-script, but in the future,
we'll need to make sure the dockerfiles and the LLVM scripts are all
locked up in the same production branch.
By making it into a parameter, we also allow for master/staging
validation, as well as snapshot production+patch to cherry pick fixes if
needed.
Change-Id: Ida36dc7dab510a41227fb793be8e5fa0c5ed8fb0
|
|
For purposes of double check as well as to have the releases available
for cross-compilation, benchmarks and other jobs, we want the release
process to run on all three arches by default, while still allowing
users to override it.
Change-Id: Ie120c244edfc1aa4bb538c2116082d7838c87020
|
|
This change adds a "compiler" option to override the system compiler if
no toolchain is provided by the parameters. The default is clang because
that's known to work on all architectures, while GCC 5.4 (Xenial) is
known to fail on ARM32 and ARM64 in some configurations.
The scripts need to be changed accordingly to set CC/CXX on their own
based in the new argument.
Change-Id: I0632d3a7d2c53650ac74ba5f2576774c6c7f7cf1
|
|
Change-Id: Ie5ba20da643b8282c70c5a555dd9177040b92602
|
|
Change-Id: Ibcd38b591333c6f1c6fffb4a24a2848b3c75cfaa
|
|
Change-Id: I05ea7e306bbe8a6e1d0ae536cb08047a1780665c
|
|
Change-Id: Idc670497bd5dcc4c8356019e2c30dfc02d6d9dbe
|
|
This change makes all TCWG matrix and multijob "coordination" jobs
to run on tcwg-coordinator label. The "coordination" jobs do
not consume significant resources and just wait for other jobs
to finish. In case of matrix jobs these are "flyweight" jobs,
and all TCWG multijobs behave similarly.
Another benefit from this change is that git checkouts
for "coordination" jobs will no longer be mixed with workspaces of
matrix sub-builds of the same jobs, if sub-builds happen to start
on the same node as coordination job.
This has been auto-generated with the following script
$ cat > tcwg-coordinator.sh <<EOF
for f in tcwg-*.yaml; do
type=$(grep "project-type" $f | sed -e "s/.*: //")
case $f in
tcwg-bkk16-buildfarm-benchmark.yaml) continue ;;
esac
case $type in
freestyle) continue ;;
matrix|multijob)
sed -e "s/ node:.*/ node: tcwg-coordinator/" -i $f
;;
*) echo "ERROR: Wrong job type: $type"; exit 1 ;;
esac
done
EOF
Change-Id: I216abbf903537da2c2bfd3f6d5a1fbf6b8fe4877
|
|
Change-Id: I73a714393dfdff7b60db0e42601e9e0c2c7e0457
|
|
Change-Id: I62d63ee294f678184ed0b18b15b89677a54fd099
|
|
Re-enable running the release script, and remove the random
result setting.
Change-Id: Ic08f9f1a695df746fa428bf48ae30194f5ebcb30
|
|
We want to trigger the test job, even if the build fails.
This patch disables the build, so that the job runs quickly
to quickly test the mechanism.
Change-Id: Icad243da7149ff22210e4faf0aa953b0d2168372
|
|
Change-Id: Ic0207141fcf1916e989c6e2a8f76579b1cd7dc57
|
|
Change-Id: Iec14449c41db58b6969451bb65eb1dc7364121b3
|
|
Change-Id: I179466ee042a9de9caa2390eb02b46b29e699ed9
|
|
Change-Id: If044c22ff132976f58a67a2d1d4b0b8b1568207a
|
|
Change-Id: I7ad12dcc22ff3afa884437ffa9a93097f4622313
|
|
variable
Change-Id: I0dfc7977184e8bb82e7b74d703344beab5711544
|
|
Change-Id: I2f22e3a493e1d30ec67c806cf33d42e75a9f1cb9
|
|
This is an attempt to chain a test-suite job, like the staged-build,
after the release. IIUC, the chaining will not happen if the container
script fails (any test failure), which is *not* what I want.
I'm looking for ideas on how to do that. If we need to use the same
trick as the release.sh script, I'll update the patch to do that.
Change-Id: I1ab593c27edcbe118bffaafa9c2498c382fade4b
|
|
Change-Id: If7115e42bd55a0f6bb05d64596a4be9c8f13ffa5
|
|
Change-Id: I6eefabccbc56a35d4b41cbcdeb9af6c55b33dd34
|
|
Change-Id: Ia54509e922fcf067897833133f40eb4a8c75d0ae
|
|
Change-Id: Iefa5bbf5a2aa471818312c8d709d4801254671e0
|
|
Change-Id: I91467075ac997d126a9e421620bec1162181bd4d
|
|
This reverts commit 59ea006406f841837040f94ab604b579e5f16f9a.
Our trial didn't work, so I'm reverting this change and we'll try
something different.
Change-Id: I5ebbd6c89796b858f903512dc930f3fa9428538d
|
|
For now, just commenting out the script, so we can focus on the matrix
job scheduling and make sure that we can filter per target. Later on,
re-enable the release script, so we can actually test it.
Change-Id: I866adfb49d8f215aabd0c03bfc5d0f5f1785f9f0
|
|
Copied from a different job, now it's the same as the other LLVM
scripts.
Change-Id: I925bb8e16241cb5085fe81ecf1dcd41682dff215
|
|
Not all builders have three slots available, and this would force too
high concurrent jobs for all but the x86 builder. Hard-coding weight to
1 until we have a better way of doing it. Also, restricting the jobs to
4 on the TK1s, since they only have that many cores.
Change-Id: I5bb0dd21dac487798fcfa2f88f03efc88a9ddc74
|
|
This adds a new LLVM release matrix job, which will run native bootstrap
builds on x86, ARM32 and ARM64 machines and push the release to our file
repository. This release tarball can be used by test-suite or further
staged builds.
Change-Id: I77d75517d30c225ba5a7da974ff57e62af543661
|