On 07/10/13 12:37 PM, Tim Bell wrote:
Hi Kelly
How did BUILD_INSTALL ever become true on an OpenJDK build?
From the 'generic_debug_build' rule in the top level 7update Makefile [1]
generic_debug_build:
$(MAKE) \
ALT_OUTPUTDIR=$(ABS_OUTPUTDIR)/$(REL_JDK_OUTPUTDIR) \
DEBUG_NAME=$(DEBUG_NAME) \
GENERATE_DOCS=false \
BUILD_INSTALL_BUNDLES=true \
CREATE_DEBUGINFO_BUNDLES=false \
$(BOOT_CYCLE_DEBUG_SETTINGS) \
generic_build_repo_series
After giving this more study, let me modify what I wrote a bit -
generic_debug_build sets "BUILD_INSTALL_BUNDLES=true", which causes the
'install-binaries-jdk-debug' rule in the same Makefile to also run.
This dives into install-rules.gmk [2] in a place where BUILD_INSTALL is
not checked.
Here is a revised fix and webrev:
http://cr.openjdk.java.net/~tbell/8012366/webrev.01/
Tim
[1] http://hg.openjdk.java.net/jdk7u/jdk7u/file/tip/Makefile
[2] http://hg.openjdk.java.net/jdk7u/jdk7u/file/tip/make/install-rules.gmk
-kto
On Jul 8, 2013, at 12:03 PM, Tim Bell wrote:
All-
My fix for 8007815 [1] did not handle the case of an OpenJDK build
that does not have the closed source trees. If a user tries to
build that far, the build will fail with "cd: can't cd to
./install/make/installer/binaries/linux"
This is a request for review and approval to fix 8012366 in 7u.
These changes will not throw a build error if the closed directories
do not exist. Makefile changes only; no product code has been
modified:
Link to the bug report:
http://bugs.sun.com/view_bug.do?bug_id=8012366
webrev of the Makefile changes:
http://cr.openjdk.java.net/~tbell/8012366/webrev.00/
Thanks in advance-
Tim
[1] There is no external link for 8007815, sorry. These changes
had to do with building and packaging in the install tree.
