Similar to issue on bug #382465. I think we should add all the sandbox exceptions in java-pkg_setup-vm in java-utils-2 eclass, but not sure.
Created attachment 287591 [details] build.log.tar.gz
not a bug in gcc
(In reply to comment #2) > not a bug in gcc But it has to be worked around in gcc ebuild.
I think the problem is due to this check: "checking for jar... jar" Bug 378319 is also related. According to bug 196643, gcc should be fine with either jar or zip and zip was already added as a dep. Unless that didn't change, it would be easiest to prevent this check from happening at all.
Created attachment 287771 [details, diff] gcc-ebuild.patch
I attached a pacth against 4.6.0 ebuild that forces internal jar script which uses zip. I'm not sure whether this should be fixed in the ebuild or the eclass though.
Created attachment 287791 [details, diff] toolchain.eclass.patch Attaching a patch against toolchain.eclass as requested by Mike.
(In reply to comment #3) no it doesn't. if running the jdk causes sandbox violations, then it needs a sandbox whitelisting just like the other jdk ebuilds install. this isn't specific to gcc. (In reply to comment #7) we'll need to depend on zip/unzip when USE=gcj since the bundled jar script needs it. i'll take care of this.
as you said, we already have the ebuilds depend on zip/unzip, so i'll tweak the DEPEND stuff independently http://sources.gentoo.org/eclass/toolchain.eclass?r1=1.470&r2=1.471