Build fails with "shifting a negative signed value is undefined" on Mac

[Tobias Hartmann]

Hi Ben,

thanks for the bug report! This issue has been fixed in JDK 9 by JDK-8152856 [1] but it seems that we accidentally 
reintroduced it in the MVT repo. I've filed JDK-8189219 [2].

You can use the following as a quick workaround:

diff -r 7dd1c51108a3 src/hotspot/share/opto/library_call.cpp
--- a/src/hotspot/share/opto/library_call.cpp	Mon Oct 09 13:23:10 2017 +0200
+++ b/src/hotspot/share/opto/library_call.cpp	Thu Oct 12 09:21:03 2017 +0200
@@ -3896,7 +3896,7 @@
    }
    // Now test the correct condition.
    jint  nval = (obj_array
-                ? (jint)(Klass::_lh_array_tag_type_value
+                ? (jint)((juint)Klass::_lh_array_tag_type_value
                     <<    Klass::_lh_array_tag_shift)
                  : Klass::_lh_neutral_value);
    Node* cmp = _gvn.transform(new CmpINode(layout_val, intcon(nval)));

Thanks,
Tobias

[1] https://bugs.openjdk.java.net/browse/JDK-8152856
[2] https://bugs.openjdk.java.net/browse/JDK-8189219

On 12.10.2017 08:54, Ben Evans wrote:
> Hi,
> 
> At changeset:   47322:aa3ca9702304 the build on Mac 10.11.1 configured with
> 
> ./configure --with-freetype-include=/usr/local/include/freetype2
> --with-freetype-lib=/usr/local/lib/
> 
> fails with:
> 
> === Output from failing command(s) repeated here ===
> * For target hotspot_variant-server_libjvm_objs_library_call.o:
> /Users/boxcat/projects/openjdk/valhalla-mvt/src/hotspot/share/opto/library_call.cpp:3900:20:
> error: shifting a negative signed value is undefined
> [-Werror,-Wshift-negative-value]
>                     <<    Klass::_lh_array_tag_shift)
>                     ^
> 1 error generated.
> 
> * All command lines available in
> /Users/boxcat/projects/openjdk/valhalla-mvt/build/macosx-x86_64-normal-server-release/make-support/failure-logs.
> === End of repeated output ===
> 
> I couldn't find a description of how much detail is required for a bug
> report - hopefully the above is enough, please let me know if more is
> required or these should be reported in a different way.
> 
> Thanks,
> 
> Ben
>