[8u] RFR(S) 8187045: [linux] Not all libraries in the VM are linked with -z,noexecstack

David Holmes david.holmes at oracle.com
Fri Jan 12 06:35:14 UTC 2018


That all looks reasonable to me.

Thanks,
David H.

On 12/01/2018 3:24 PM, David Buck wrote:
> Hi!
> 
> May I please get a review of my backport of this small change:
> 
> bug report:
> https://bugs.openjdk.java.net/browse/JDK-8187045
> 
> JDK 10 review thread:
> http://mail.openjdk.java.net/pipermail/build-dev/2017-September/019723.html
> 
> JDK 10 changeset:
> http://hg.openjdk.java.net/jdk10/master/rev/8b5ed0358b3a
> 
> webrevs of JDK 8 backport for review:
> http://cr.openjdk.java.net/~dbuck/8187045/
> 
> As there have been significant changes (refactoring) to the build files between JDK 8 and JDK 10, I had to modify the change slightly. For each of the two source trees modified, hotspot and root, I tried to follow the preexisting style as much as possible. So for hotspot (really just libjsig.so and the SA agent), I introduced a new variable, LDFLAGS_NO_EXEC_STACK to the make files. (Note that libjvm.so was already built with noexecstack before this fix). However, for the forest root changes, I directly added the needed option to LDFLAGS_JDK. Ultimately these are merely syntax differences, the actual change in resulting builds is exactly the same as in JDK 10.
> 
> The test case only required very minor modification (test library refactoring, minor change to how ElfFile::specifies_noexecstack is called).
> 
> I have manually confirmed, on arm64 (an impacted platform), that the the included test case now passes after this fix. JPRT hotspot testset run and passed.
> 
> Cheers,
> -Buck
> 


More information about the build-dev mailing list