> On Jan 9, 2016, at 11:23 PM, Alan Bateman <alan.bate...@oracle.com> wrote: > > > > On 09/01/2016 19:35, James Laskey wrote: >> Alan questioned why this is showing up after the sjavac changes. There was >> concern there was some lurking issue. >> > No issue removing the test from the exclude list, that should have been done > as part of JDK-8146712. > > I think we need to establish whether the META-INF/jvmci.providers > configuration files are meant to be in the run-time image. I assume they are > but they have not been present until the recent changes (looks like the > sjavac change). Their presence confuses the test, a minor oversight in the > test that is fixed now. The issue is of course nothing to do with the jimage > code. If you look at a build from a few days ago then the > META-INF/jvmci.providers aren't in the exploded build, pull in the sjavac > change and they are in the exploded build. Maybe the changes to the > exclude/include patterns causes this but needs digging into.
The @ServiceProvider mechanism will be moved from the JVMCI to the JVMCI compiler to make the JVMCI simpler: http://hg.openjdk.java.net/graal/graal-jvmci-9/hotspot/rev/d1326c9f3cfb http://hg.openjdk.java.net/graal/graal-compiler/rev/56359eb3abfa It’s a pending change that hasn’t reached hs-comp yet but should in the next weeks. Do whatever you need to do to make the tests work. > > -Alan
