Skip to main content
aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLeo Ufimtsev2018-01-18 19:44:04 +0000
committerAlexander Kurtakov2018-01-26 14:19:08 +0000
commit65ca41b1f452e1f1e15f55c0bf09339f72e065eb (patch)
tree28c77d1bd705570692099e9a959db60b8f1eed0b /features/org.eclipse.equinox.executable.feature/library/gtk/build.sh
parente55e66e8760fe7f79c6becf09ad7f4519e02d2e5 (diff)
downloadrt.equinox.framework-65ca41b1f452e1f1e15f55c0bf09339f72e065eb.tar.gz
rt.equinox.framework-65ca41b1f452e1f1e15f55c0bf09339f72e065eb.tar.xz
rt.equinox.framework-65ca41b1f452e1f1e15f55c0bf09339f72e065eb.zip
Details: - This makes make_linux.mak independent, i.e the following works: make -f make_linux.mak clean all It does so by assuming default values if they are not set by build.sh. However, build.sh still overrides values for backwards compatibility. I.e, I expect it will work fine with build servers. - I migrated some linux-specific logic from build.sh to make_linux.mak * JAVA_HOME detection that I added earlier, * linux specific x86/x86_64 flags - Added a new target: dev_build_install This can be invoked like: export DEV_ECLIPSE="your/eclipse" make -f make_linux.mak clean dev_build_install It builds the launcher/lib and then you can test the output in your dev eclipse. (see target for details.). Test strategy: I've done some very through testing, I'm pretty confident it shouldn't break anything. I've accounted for x86/x86_64 differences. I've added a print statement that prints info about variables: $(info Input info: ....) This line generates the same values for when you run build.sh as well as when make_linux is invoked directly. Produced eclipse and eclipse_xxxx.so work well. Tested on Fedora 27, Wayland, gtk3.22, x86_64. Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=530012 Change-Id: I07ec8f453acd93d368e5e8f927fa65354da0f68b Signed-off-by: Leo Ufimtsev <lufimtse@redhat.com>
Diffstat (limited to 'features/org.eclipse.equinox.executable.feature/library/gtk/build.sh')
-rwxr-xr-xfeatures/org.eclipse.equinox.executable.feature/library/gtk/build.sh19
1 files changed, 0 insertions, 19 deletions
diff --git a/features/org.eclipse.equinox.executable.feature/library/gtk/build.sh b/features/org.eclipse.equinox.executable.feature/library/gtk/build.sh
index 691eb40ae..f29fc9bb7 100755
--- a/features/org.eclipse.equinox.executable.feature/library/gtk/build.sh
+++ b/features/org.eclipse.equinox.executable.feature/library/gtk/build.sh
@@ -86,23 +86,10 @@ case $defaultOS in
defaultOSArch="x86_64"
defaultJava=DEFAULT_JAVA_EXEC
[ -d /bluebird/teamswt/swt-builddir/JDKs/x86_64/jdk1.5.0 ] && defaultJavaHome="/bluebird/teamswt/swt-builddir/JDKs/x86_64/jdk1.5.0"
-
- # Automatically figure out where java is installed on the system if it was not found/given otherwise.
- if [ "$defaultJavaHome" = "" ]; then
- DYNAMIC_JAVA_HOME=`readlink -f /usr/bin/java | sed "s:jre/::" | sed "s:bin/java::"`
- if [ -a "${DYNAMIC_JAVA_HOME}include/jni.h" ]; then
- echo "Info: Java automatically set to: $DYNAMIC_JAVA_HOME "
- defaultJavaHome="$DYNAMIC_JAVA_HOME"
- else
- echo "Warning, java not found on system. Please specify JAVA_HOME manually"
- fi
- fi
- OUTPUT_DIR="$EXEC_DIR/bin/$defaultWS/$defaultOS/$defaultOSArch"
;;
i?86 | "x86")
defaultOSArch="x86"
[ -d /bluebird/teamswt/swt-builddir/JDKs/x86/ibm-java2-i386-50 ] && defaultJavaHome="/bluebird/teamswt/swt-builddir/JDKs/x86/ibm-java2-i386-50"
- OUTPUT_DIR="$EXEC_DIR/bin/$defaultWS/$defaultOS/$defaultOSArch"
;;
"ppc")
defaultOSArch="ppc"
@@ -265,12 +252,6 @@ elif [ "$defaultOS" = "solaris" ]; then
M_ARCH=-m64
export M_ARCH
fi
-elif [ "$defaultOS" = "linux" ]; then
- if [ "$defaultOSArch" = "x86_64" ]; then
- # Bug 517013: Avoid using memcpy() to remain compatible with older glibc
- M_CFLAGS="-fno-builtin-memcpy -fno-builtin-memmove"
- export M_CFLAGS
- fi
fi
LIBRARY_DIR="$EXEC_DIR/../org.eclipse.equinox.launcher.$defaultWS.$defaultOS.$defaultOSArch"

Back to the top