Yocto: BUILD_LDFLAGS set to build system libraries, not target - yocto

Question

I've got a makefile recipe that builds a shared library (an alsa plugin). If I build the library outside if yocto everything works correctly and alsa will link to the library.
However if I build it with yocto, even though the log is error free, when I try and run alsa, I get an error "Cannot open shared library". The library is installed in location referenced by the error message and it's permissions are correct.
From within the recipe if I print out what BUILD_LDFLAGS is set to I notice the it's pointing to the x86_64-linux (build system) libraries instead of the 'MACHINE' libraries (example: -L//.build-yocto/tmp/sysroots/x86_64-linux/lib"
My questions are:
Is the BUILD_LDFLAGS the source of my problem?
If so how do I remedy it?
If not BUILD_LDFLAGS, any idea what is the problem.
Here is a copy of my recipe bb file:
SUMMARY = "..."
LICENSE = "CLOSED"
#Package release number
PR = "r0"
###################################################################
#The following lines tell yocto where to get the source code from
# This section is for git. Comment out ALL this section if
# you DO NOT want to pull from a git repo (local or remote).
# If pulling from git uncomment and modify paths.
###################################################################
#Uncomment following line to pull from REMOTE git repo
#SRC_URI = "git://gitpath;protocol=ssh;branch=master"
#Uncomment following line and modify path to pull from LOCAL git repo clone
##SRC_URI = "git:///localgitpath;protocol=file;branch=master"
#Change hash to match the commit you want yocto to use
##SRCREV="xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
##S = "${WORKDIR}/git/"
# End of git section
###################################################################
#The following lines tell yocto where to use a local file system
# for the source. Uncomment all lines and modify paths
###################################################################
SRC_URI = ""
inherit externalsrc
EXTERNALSRC = "/home/<my_path>"
EXTERNALSRC_BUILD = "/home/<my_path>"
# End of local file system section
##################################################################
#END of where to get source code
##################################################################
#Ignore vendor ldflags checking and use ours
INSANE_SKIP_${PN} = "ldflags"
#Don't strip debug symbols
INHIBIT_PACKAGE_STRIP = "1"
INHIBIT_SYSROOT_STRIP = "1"
SOLIBS = ".so"
#Tell yocto that the .so files are real and not sym-links.
FILES_SOLIBSDEV = ""
#/usr/lib/alsa-lib
FILES_${PN} += "${libdir}/alsa-lib"
#/usr/<PATH>
FILES_${PN} += "${prefix}/<PATH>"
DEPENDS += "alsa-lib"
EXTRA_OEMAKE += "'CC=${CC}' 'RANLIB=${RANLIB}' 'AR=${AR}' 'CFLAGS=${CFLAGS} -I${S}/include' 'BUILDDIR=${S}' 'DESTDIR=${D}'"
TARGET_CFLAGS += "-DPIC -fPIC -Wall -Wextra -O2 -g -I./include -I<path> -I-I<path2> -I<path3> -lasound"
TARGET_LDFLAGS += "-shared -lasound"
do_configure() {
oe_runmake -f Makefile.yocto clean
}
do_compile() {
# unset LDFLAGS TARGET_LDFLAGS BUILD_LDFLAGS
echo " Werkdir ${WORKDIR}"
echo " Compiler ${CC}"
echo " BUILD_LDFLAGS ${BUILD_LDFLAGS}"
echo " LDFLAGS ${LDFLAGS}"
echo " TARGET_LDFLAGS ${TARGET_LDFLAGS}"
oe_runmake -f Makefile.yocto all 'CC=${CC}'
}
do_install() {
install -d ${D}${libdir}
install -d ${D}${libdir}/alsa-lib
install -d ${D}${bindir}
install -d ${D}${prefix}
install -d ${D}${prefix}/<PATH>
install -m 0644 <path_n>lib1.so ${D}${libdir}
install -m 0644 <path_n>lib2.so.so ${D}${libdir}
install -m 0644 <path_n>lib3.so.so ${D}${libdir}
install -m 0644 <path_n>lib4.so.so ${D}${libdir}
install -m 0644 <path_n>lib1pcm_plugin.so ${D}${libdir}/alsa-lib
install -m 0755 <path_n>app1 ${D}${bindir}
install -m 0755 <path_n>app2 ${D}${bindir}
install -m 0755 <path_n>app3 ${D}${bindir}
install -m 0755 <path_n>app4 ${D}${bindir}
install -m 0755 <path_n>app5 ${D}${bindir}
install -m 0755 <path_n>app6 ${D}${bindir}
install -m 0755 <path_n>app7 ${D}${bindir}
install -m 0755 <path_n>app8 ${D}${bindir}
}
Makefile:
# Makefile template for shared library
#Yocto will pass in the CC flag so this is commented out. Otherwise the correct compiler won't be used
#CC = gcc # C compiler
#These are here to allow a build outside of Yocto (testing the build). Yocto's CFLAGS
#and LDFLAGS will override these.
CFLAGS += -fPIC -Wall -Wextra -O2 -g -I<path1> -I<path2> -I<path2> # C flags
LDFLAGS = -shared # linking flags
RM = rm -f # rm command
TARGET_LIB = libasoundplugin.so # target lib
LIB1=lib1
PATH1=<path1>
LIB2=lib2
PATH2=<path2>
INCLUDE_FLAGS = -L$(PATH1) -l$(LIB1I) \
-L$(PATH2) -l$(LIB2) \
-lasound
SRCS = source.c
OBJS = $(SRCS:.c=.o)
.PHONY: all
all: ${TARGET_LIB}
$(TARGET_LIB): $(OBJS)
$(CC) ${LDFLAGS} ${INCLUDE_FLAGS} -o $# $^
$(SRCS:.c=.d):%.d:%.c
$(CC) $(CFLAGS) -MM $< >$#
include $(SRCS:.c=.d)
.PHONY: clean
clean:
-${RM} ${TARGET_LIB} ${OBJS} $(SRCS:.c=.d)
Thanks!

Answer 1

Is the BUILD_LDFLAGS the source of my problem?
No. BUILD_LDFLAGS is used to build native packages, while you build a target package. In your case TARGET_LDFLAGS variable will be used as LDFLAGS source.
Here is a copy of my recipe bb file
Where does it come from? Did you write it from a template?
As I can see there is a recipe for alsa plugins, maybe you can add your plugin to this recipe?
You can also take as example the recipe for a2jmidid, it is close to what you are trying to do as I understand. It does nothing special, except for setting dependencies, getting sources, adding certain ld flags and passing resulting files to the package.
Regarding the Makefile (I suppose it is also your recipe):
First of all there is no need to have a separate Makefile for yocto.
#Yocto will pass in the CC flag so this is commented out. Otherwise the correct compiler won't be used
#CC = gcc # C compiler
You are passing CC from your .bb recipe (even twice: via EXTRA_OEMAKE and in do_compile). There is no need to comment out CC variable, as definitions inside Makefile have lower priority compared to definitions passed as command line arguments.
Why do you need INCLUDE_FLAGS variable? I don't understand what does it do (except for passing to gcc some flags for the third time).
Ultimately, you can go to the workdir, open file temp/log.do_compile and you will see which commands were executed to compile your plugin. Compare it to the settings you use to compile it manually, and you will see what is the difference between successful and unsuccessful builds.