Age | Commit message (Collapse) | Author |
|
The jcstress testsuite generates a results file with a new file name.
The cp command to copy this file was failing.
Compress the jcstress tests results. Before this change, the output was
suffixed with gz, but wasn't actually gzipped.
Change-Id: Ifd034150a6dbd38ff5294694c4b94cfc6a23c4d1
|
|
With the exception of the benchmarks, the job configurations should be
viewable.
Change-Id: Id1784ae8cb9a54f77b80c1844fdc78509f4ad163
|
|
According to the author, JCStresss mustn't have any additional options
passed to it. Remove them, we don't need to distinguish between C1 and
C2, the framework will do that.
Change-Id: I8908cbabde12aed7e552494419f30d162d502d31
|
|
default mode of jcstress takes well over 1 day to run. Reducing to
quick.
Change-Id: Ie043a19af7af5c954796ca3362a3a3cfb9590b54
|
|
Before we were running with "sanity" mode, which is a very short run. As
the tests are stochastic in nature, the longer they run, the better.
Change-Id: I5ade5945f59d0f3e99d99381febf48256a0a9c24
|
|
This reverts commit fae653c9a8d564f1d67283a8bfe57f315040d0b3.
Change-Id: I26472c56cf567b88671e7f0c7d928d1672b18c93
|
|
Perhaps re-enable this as C1 concurrency is important too.
Remove the restriction in jdk?-build-and-test for only server variants
to run, or else this job will fail because of the missing results for
client builds.
Change-Id: I0a7877e7e8f8c630ae75ecfb4c5458f13e53e151
|
|
Rather than using patches to re-enable the client VM (i.e. C1 compiler),
use the -XX:TieredStopAtLevel=1 option to prevent higher-levels of
compilation. This requires the scripting to be aware of the JVM_VARIANT envvar
and alter the behaviour accordingly, rather than just changing the execution
path.
Change-Id: I9955db0a959788685e050c060d80f0303a917e62
|
|
For testing purposes and re-runs.
Change-Id: Id39700071e0c01684e7da996b99240cb02605153
|
|
And pass JDK_VERSION variable to scripts. This make splitting the jobs
more easy to split between jdk8u and jdk9.
Change-Id: I8d0d2ff61cbab955d7d144e18bdeab6b3d016a50
|
|
There were warnings using the logrotate property. This replaces it with
a build-discarder property.
Change-Id: I6b34938eca86c050d2edb5aa70ab91a9c1b76d11
|
|
This will reduce the execution time from 22 hours to about 2 minutes.
Restore this to "quick" once we've validated the build system.
Change-Id: I50e3125442f619c110ccea15f8e1d4c8fff845b9
|
|
Sometime after October last year, JCStress breaks if its jar file
is appended to the bootclasspath. I can find no reference to
instructions that require the bootclasspath to be set.
This will allow the tests to run. This might be revisited.
Change-Id: I48d36817d96dd0ff0819d857f4c328e584e35459
|
|
* restrict where the build job is run
* get rid of combination filter comment
* don't flatten jcstress-build artifacts
* copy jdk from local filesystem
Change-Id: If1d3b6aa20b8dd3faa809e79c1e2116f7f91e61c
Signed-off-by: Fathi Boudra <fathi.boudra@linaro.org>
|
|
Change-Id: I21c918bb514312c9f7ec821f934b9ca200e8b36b
Signed-off-by: Fathi Boudra <fathi.boudra@linaro.org>
|
|
Change-Id: Ia242cb747b65167a203c729b7e0f842640247cff
Signed-off-by: Fathi Boudra <fathi.boudra@linaro.org>
|