Reputation: 55999
I'm working on a configuration script for a JNI wrapper. One of the configuration parameters is the path to jni.h
. What's a good quick-and-dirty Autoconf test for whether this parameter is set correctly for C++ compilation? You can assume you're running on Linux and g++
is available.
Alternatively, is there a way to get javah
(or a supporting tool) to give me this path directly?
Upvotes: 5
Views: 2376
Reputation: 140
FYI - the patch below against the latest ax_jni_include_dir.m4 works for me on Macos 11.1.
--- a/m4/ax_jni_include_dir.m4
+++ b/m4/ax_jni_include_dir.m4
@@ -73,13 +73,19 @@ fi
case "$host_os" in
darwin*) # Apple Java headers are inside the Xcode bundle.
- macos_version=$(sw_vers -productVersion | sed -n -e 's/^@<:@0-9@:>@
*.\(@<:@0-9@:>@*\).@<:@0-9@:>@*/\1/p')
- if @<:@ "$macos_version" -gt "7" @:>@; then
- _JTOPDIR="$(xcrun --show-sdk-path)/System/Library/Frameworks/JavaVM.framework"
- _JINC="$_JTOPDIR/Headers"
+ major_macos_version=$(sw_vers -productVersion | sed -n -e 's/^\(@<:@0-9@:>@*\).@<:@0-9@:>@*.@<:@0-9@:>@*/\1/p')
+ if @<:@ "$major_macos_version" -gt "10" @:>@; then
+ _JTOPDIR="$(/usr/libexec/java_home)"
+ _JINC="$_JTOPDIR/include"
else
- _JTOPDIR="/System/Library/Frameworks/JavaVM.framework"
- _JINC="$_JTOPDIR/Headers"
+ macos_version=$(sw_vers -productVersion | sed -n -e 's/^@<:@0-9@:>@*.\(@<:@0-9@:>@*\).@<:@0-9@:>@*/\1/p')
+ if @<:@ "$macos_version" -gt "7" @:>@; then
+ _JTOPDIR="$(xcrun --show-sdk-path)/System/Library/Frameworks/JavaVM.framework"
+ _JINC="$_JTOPDIR/Headers"
+ else
+ _JTOPDIR="/System/Library/Frameworks/JavaVM.framework"
+ _JINC="$_JTOPDIR/Headers"
+ fi
fi
;;
*) _JINC="$_JTOPDIR/include";;
Upvotes: 0
Reputation: 5542
Then there is the easy way: http://www.gnu.org/software/autoconf-archive/ax_jni_include_dir.html
Sometimes it is best to just use the standard recipies.
Upvotes: 7
Reputation: 1507
Checking for headers is easy; just use AC_CHECK_HEADER
. If it's in a weird place (i.e., one the compiler doesn't know about), it's entirely reasonable to expect users to set CPPFLAGS
.
The hard part is actually locating libjvm
. You typically don't want to link with this; but you may want to default to a location to dlopen
it from if JAVA_HOME
is not set at run time.
But I don't have a better solution than requiring that JAVA_HOME
be set at configure time. There's just too much variation in how this stuff is deployed across various OSes (even just Linux distributions). This is what I do:
AC_CHECK_HEADER([jni.h], [have_jni=yes])
AC_ARG_VAR([JAVA_HOME], [Java Runtime Environment (JRE) location])
AC_ARG_ENABLE([java-feature],
[AC_HELP_STRING([--disable-java-feature],
[disable Java feature])])
case $target_cpu in
x86_64) JVM_ARCH=amd64 ;;
i?86) JVM_ARCH=i386 ;;
*) JVM_ARCH=$target_cpu ;;
esac
AC_SUBST([JVM_ARCH])
AS_IF([test X$enable_java_feature != Xno],
[AS_IF([test X$have_jni != Xyes],
[AC_MSG_FAILURE([The Java Native Interface is required for Java feature.])])
AS_IF([test -z "$JAVA_HOME"],
[AC_MSG_WARN([JAVA_HOME has not been set. JAVA_HOME must be set at run time to locate libjvm.])],
[save_LDFLAGS=$LDFLAGS
LDFLAGS="-L$JAVA_HOME/lib/$JVM_ARCH/client -L$JAVA_HOME/lib/$JVM_ARCH/server $LDFLAGS"
AC_CHECK_LIB([jvm], [JNI_CreateJavaVM], [LIBS=$LIBS],
[AC_MSG_WARN([no libjvm found at JAVA_HOME])])
LDFLAGS=$save_LDFLAGS
])])
Upvotes: 5