Date: Fri, 26 Aug 2016 18:29:46 +0000 (UTC) From: Jung-uk Kim <jkim@FreeBSD.org> To: ports-committers@freebsd.org, svn-ports-all@freebsd.org, svn-ports-head@freebsd.org Subject: svn commit: r420940 - in head/java/openjdk8: . files Message-ID: <201608261829.u7QITkD3036503@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: jkim Date: Fri Aug 26 18:29:46 2016 New Revision: 420940 URL: https://svnweb.freebsd.org/changeset/ports/420940 Log: Make the FPU hack for ARM optional. Note this hack is not necessary for VFPv3 or later, i.e., only VFPv1 and VFPv2 require this hack. https://svnweb.freebsd.org/changeset/base/288983 Added: head/java/openjdk8/files/fpuhack.patch - copied, changed from r420939, head/java/openjdk8/files/armv6.patch Deleted: head/java/openjdk8/files/armv6.patch Modified: head/java/openjdk8/Makefile Modified: head/java/openjdk8/Makefile ============================================================================== --- head/java/openjdk8/Makefile Fri Aug 26 18:14:36 2016 (r420939) +++ head/java/openjdk8/Makefile Fri Aug 26 18:29:46 2016 (r420940) @@ -234,10 +234,14 @@ CONFIGURE_SCRIPT= ../../configure CONFIGURE_WRKSRC= ${WRKSRC}/common/autoconf OPTIONS_DEFINE= POLICY TZUPDATE +OPTIONS_DEFINE_armv6= FPUHACK OPTIONS_DEFAULT= POLICY TZUPDATE +OPTIONS_DEFAULT_armv6= FPUHACK +FPUHACK_DESC= Hack denormalized numbers in flush-to-zero mode POLICY_DESC= Install the Unlimited Strength Policy Files TZUPDATE_DESC= Update the time zone data +FPUHACK_EXTRA_PATCHES= ${PATCHDIR}/fpuhack.patch POLICY_CONFIGURE_ENABLE= unlimited-crypto TZUPDATE_RUN_DEPENDS= java-zoneinfo>0:java/java-zoneinfo @@ -293,16 +297,6 @@ CONFIGURE_ARGS+= --with-boot-jdk-jvmargs MAKE_ENV+= JAVADOC_CMD_MEM="-Xmx768m" .endif -.if ${ARCH} == "armv6" -# From NetBSD http://cvsweb.netbsd.org/bsdweb.cgi/pkgsrc/lang/openjdk7/hacks.mk -# Workaround incorrect constant folding of subnormals in javac when the FPU -# does not handle subnormal arithmetic, like on ARM in Flush-to-zero mode. -# These workarounds avoid underflow conditions during the bootstrap so the -# JDK can correctly build itself. Compiling or running programs other than -# openjdk itself on such hardware may still cause unexpected behaviour. -EXTRA_PATCHES= ${PATCHDIR}/armv6.patch -.endif - .if ${ARCH} == "amd64" || ${ARCH} == "i386" JDK_BUILD_JVM= server .else Copied and modified: head/java/openjdk8/files/fpuhack.patch (from r420939, head/java/openjdk8/files/armv6.patch) ============================================================================== --- head/java/openjdk8/files/armv6.patch Fri Aug 26 18:14:36 2016 (r420939, copy source) +++ head/java/openjdk8/files/fpuhack.patch Fri Aug 26 18:29:46 2016 (r420940) @@ -1,3 +1,10 @@ +From NetBSD http://cvsweb.netbsd.org/bsdweb.cgi/pkgsrc/lang/openjdk7/hacks.mk +Work around incorrect constant folding of subnormals in javac when the FPU +does not handle subnormal arithmetic, like on ARM in flush-to-zero mode. +These workarounds avoid underflow conditions during the bootstrap so the JDK +can correctly build itself. Compiling or running programs other than OpenJDK +itself on such hardware may still cause unexpected behaviour. + --- jdk/src/share/classes/java/lang/Double.java.orig 2014-03-04 02:57:59 UTC +++ jdk/src/share/classes/java/lang/Double.java @@ -86,7 +86,7 @@ public final class Double extends Number
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201608261829.u7QITkD3036503>