Thanks for your testing! OK, I'd like to do 2 things: add the workaround to uimaj code, and also build under Java 8 in case there are other issues.
I'm wondering about the workaround you mentioned earlier. From the previously posted stack trace, it seems the issue is in DKPro. In looking at CasIOUtils, I don't see any use of ByteBuffer. Can you post a stack trace or some other evidence, that helps me figure out where to put the workaround? Thank you very much :-) -Marshall On 11/7/2018 6:21 PM, Richard Eckart de Castilho wrote: > On 7. Nov 2018, at 23:19, Richard Eckart de Castilho <[email protected]> wrote: >> I didn't yet test against the artifacts that you staged (which were built >> with >> JDK 11 if I remember correctly). If they do not exhibit the problem, I'd say >> we can go forward checking the RC. However, if they do exhibit the problem >> and if they cannot be used in a Java 8 context, then I'd say we have to >> re-do the RC. >> >> I'll keep you posted. > Built now against the artifacts in the staging area. Again the setup: > > 1. UIMA v3.0.1 artifacts from RC1 staging Maven repo accessed by adding > corresponding repository/pluginRepository entries to uimaFIT and DKPro Core > in the two steps below > > 2. locally built version of uimaFIT branch "release/3.0.0" (which is release > in preparation, not actually released) using Oracle Java 1.8.0_181 > > 3. locally built DKPro Core branch "2.0.x" (largely compatible with UIMAv3) > using Oracle Java 1.8.0_181 > > In step 3, the build fails again with > > java.lang.NoSuchMethodError: > java.nio.ByteBuffer.position(I)Ljava/nio/ByteBuffer; > > Note that DKPro Core has an extensive test for the different (binary) file > formats supported by UIMA which in this form does not exist in uimaFIT, so > the problem is not triggered in step 2, only in step 3. > > Looks to me that people who would use the CasIOUtils from the current RC1 in > a Java 8 environment would have trouble. > > Cheers, > > -- Richard > >
