This patch consists of:
(1) bind version bump
(2) removing some X-compiling build fixes merged upstream
(3) removing the bind-dlopen patch: not of general enough use
(4) remove the package/bind/bind9 file - it isn't used
(5) Use $(STAGING_DIR)/lib instead of $(STAGING_DIR)/usr/lib
(6) Fix chroot'ed bind handling by init script
svn 14777 seems to have broken cramfs/squashfs/tarroot.
The problem is that $(IMAGE) is the absolute path of the image including
directories, so the build tries to access
$(STAGING_DIR)/_fakeroot.</full/path/to/buildroot/directory> which doesn't
exist.
The fix is to define CRAMFS/SQUASHFS/TAR_TARGET as $(notdir
$(IMAGE).<extension>) instead.
When gcc 3.4.4 is configured for the toolchain, it subsequently fails to
compile udhcp-0.9.8/dhcpc.c complaining about an improper switch statement
syntax error:
dhcpc.c: In function `perform_renew':
dhcpc.c:136: error: label at end of compound statement
The term gdb-client is deceptive - although it parallels the gdb-server name,
it is actually a full fledged gdb version. I want to use it to debug uClibc
core files on my host system. Although I haven't got that to work yet, I wasted
time figuring out how to do build gdb for the host not realizing it already had
a target. With this documentation change, things would have been a lot clearer
to me.
libdirectfb* is copied to target, but when the package builds it also builds
libdirect* and libfusion*, which libdirectfb will call upon, but they are not
copied along with libdirectfb.
Building DirectFB compiles fine and installs fine, but when running on the
target the libraries look for DirectFB modules within a path that was defined
as $(PREFIX) during the build. Under buildroot PREFIX is set to
$(STAGING_DIR)/usr, which obviously doesn't exist on the target so it dies.
Instead, it should be built such that PREFIX=/usr then have prefix &
exec_prefix set to $(STAGING_DIR)/usr when it's time to install.
