Re: [OE-core] [PATCH V2] gcc-runtime: Move content from gcclibdir into libdir
On Wed, Aug 21, 2019 at 9:11 PM Mittal, Anuj wrote: > > On Fri, 2019-08-16 at 08:08 -0700, Khem Raj wrote: > > On Fri, Aug 16, 2019 at 5:00 AM Martin Jansa > > wrote: > > > Hi, > > > > > > I have an app which includes omp.h from gomp, it used to find it > > > without adding any -I for that (with just -fopenmp to enable > > > openmp). > > > > > > Now the header file is included in RSS: > > > lib32-recipe-sysroot/usr/lib/arm-oemllib32-linux- > > > gnueabi/9.2.0/include/omp.h > > > but no longer found in default search dirs. > > > > > > Is this expected or should gcc be adjusted to search for it > > > automatically? > > > > > > > It should have adjusted automatically but these headers are searhed > > relative to its own install and not sysroot reelative > > so it might have to be addressed. Does it work with normal ( > > non-multilib ) case ? > > It's not working for me as well. It's easy to reproduce when using > find_package(OpenMP) with cmake. > > The search paths also don't list this new path and are looking for > headers in old one. Would the path in patch "0020-Ensure-target-gcc- > headers-can-be-included.patch" have to be adjusted as well to include > this new location? > there is patch posted already to fix this https://patchwork.openembedded.org/patch/164057/ > Thanks, > > Anuj -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH V2] gcc-runtime: Move content from gcclibdir into libdir
On Fri, 2019-08-16 at 08:08 -0700, Khem Raj wrote: > On Fri, Aug 16, 2019 at 5:00 AM Martin Jansa > wrote: > > Hi, > > > > I have an app which includes omp.h from gomp, it used to find it > > without adding any -I for that (with just -fopenmp to enable > > openmp). > > > > Now the header file is included in RSS: > > lib32-recipe-sysroot/usr/lib/arm-oemllib32-linux- > > gnueabi/9.2.0/include/omp.h > > but no longer found in default search dirs. > > > > Is this expected or should gcc be adjusted to search for it > > automatically? > > > > It should have adjusted automatically but these headers are searhed > relative to its own install and not sysroot reelative > so it might have to be addressed. Does it work with normal ( > non-multilib ) case ? It's not working for me as well. It's easy to reproduce when using find_package(OpenMP) with cmake. The search paths also don't list this new path and are looking for headers in old one. Would the path in patch "0020-Ensure-target-gcc- headers-can-be-included.patch" have to be adjusted as well to include this new location? Thanks, Anuj -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH V2] gcc-runtime: Move content from gcclibdir into libdir
On Fri, Aug 16, 2019 at 5:00 AM Martin Jansa wrote:
>
> Hi,
>
> I have an app which includes omp.h from gomp, it used to find it without
> adding any -I for that (with just -fopenmp to enable openmp).
>
> Now the header file is included in RSS:
> lib32-recipe-sysroot/usr/lib/arm-oemllib32-linux-gnueabi/9.2.0/include/omp.h
> but no longer found in default search dirs.
>
> Is this expected or should gcc be adjusted to search for it automatically?
>
It should have adjusted automatically but these headers are searhed
relative to its own install and not sysroot reelative
so it might have to be addressed. Does it work with normal (
non-multilib ) case ?
> Looking at the default search paths I see:
>
> ignoring nonexistent directory
> "BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/../../../../../arm-oemllib32-linux-gnueabi/include"
> ignoring nonexistent directory "/not/exist/usr/include/c++/9.2.0"
> ignoring nonexistent directory
> "/not/exist/usr/include/c++/9.2.0/arm-oemllib32-linux-gnueabi"
> ignoring nonexistent directory "/not/exist/usr/include/c++/9.2.0/backward"
> ignoring duplicate directory
> "BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/../../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include"
> ignoring nonexistent directory
> "/not/exist/usr/lib/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include"
> ignoring nonexistent directory "/not/exist/usr/local/include"
> ignoring duplicate directory
> "BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/../../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include-fixed"
> ignoring nonexistent directory
> "BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/../../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/../../../../../arm-oemllib32-linux-gnueabi/include"
> ignoring nonexistent directory "/not/exist/usr/include"
> #include "..." search starts here:
> #include <...> search starts here:
>
> BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include
>
> BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include-fixed
> End of search list.
>
> On Tue, Aug 13, 2019 at 5:29 PM Khem Raj wrote:
>>
>> OE does not use the traditional /usr/lib/gcc prefix to store gcc-runtime
>> it basically is moved into libdir, however some newer files were
>> installed by newer versions of gcc especially libgomp ( omp.h openacc.h )
>> into gcclibdir, so we have content in both directories, this confuses
>> other tools which are trying to guess the gcc installation and its
>> runtime location, since now we have two directories, the tools either
>> choose one or other and we get inconsistent behavior, e.g. clang for
>> aarch64 uses /usr/lib but same clang for riscv64 chose /usr/lib/gcc
>>
>> This change ensures that OE ends up with single valid location for gcc
>> runtime files
>>
>> Move more common bits into common inc file
>>
>> Signed-off-by: Khem Raj
>> ---
>> v2: Divert packaging to use new path in whole recipe
>>
>> meta/recipes-devtools/gcc/gcc-runtime.inc| 18 +++---
>> meta/recipes-devtools/gcc/gcc-runtime_8.3.bb | 10 --
>> meta/recipes-devtools/gcc/gcc-runtime_9.1.bb | 10 --
>> 3 files changed, 15 insertions(+), 23 deletions(-)
>>
>> diff --git a/meta/recipes-devtools/gcc/gcc-runtime.inc
>> b/meta/recipes-devtools/gcc/gcc-runtime.inc
>> index a5c2600d7f..22c1d78dd1 100644
>> --- a/meta/recipes-devtools/gcc/gcc-runtime.inc
>> +++ b/meta/recipes-devtools/gcc/gcc-runtime.inc
>> @@ -17,6 +17,12 @@ EXTRA_OECONF_PATHS = "\
>> EXTRA_OECONF_append_linuxstdbase = " --enable-clocale=gnu"
>> EXTRA_OECONF_append = " --cache-file=${B}/config.cache"
>>
>> +# Disable ifuncs for libatomic on arm conflicts -march/-mcpu
>> +EXTRA_OECONF_append_arm = " libat_cv_have_ifunc=no "
>> +
>> +# Building with thumb enabled on armv6t fails
>> +ARM_INSTRUCTION_SET_armv6 = "arm"
>> +
>> RUNTIMELIBITM = "libitm"
>> RUNTIMELIBITM_arc = ""
>> RUNTIMELIBITM_mipsarch = ""
>> @@ -77,6 +83,11 @@ do_install () {
>> cd ${B}/${TARGET_SYS}/$d/
>> oe_runmake 'DESTDIR=${D}'
>> MULTIBUILDTOP=${B}/${TARGET_SYS}/$d/ install
>>
Re: [OE-core] [PATCH V2] gcc-runtime: Move content from gcclibdir into libdir
Hi,
I have an app which includes omp.h from gomp, it used to find it without
adding any -I for that (with just -fopenmp to enable openmp).
Now the header file is included in RSS:
lib32-recipe-sysroot/usr/lib/arm-oemllib32-linux-gnueabi/9.2.0/include/omp.h
but no longer found in default search dirs.
Is this expected or should gcc be adjusted to search for it automatically?
Looking at the default search paths I see:
ignoring nonexistent directory
"BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/../../../../../arm-oemllib32-linux-gnueabi/include"
ignoring nonexistent directory "/not/exist/usr/include/c++/9.2.0"
ignoring nonexistent directory
"/not/exist/usr/include/c++/9.2.0/arm-oemllib32-linux-gnueabi"
ignoring nonexistent directory "/not/exist/usr/include/c++/9.2.0/backward"
ignoring duplicate directory
"BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/../../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include"
ignoring nonexistent directory
"/not/exist/usr/lib/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include"
ignoring nonexistent directory "/not/exist/usr/local/include"
ignoring duplicate directory
"BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/../../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include-fixed"
ignoring nonexistent directory
"BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/../../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/../../../../../arm-oemllib32-linux-gnueabi/include"
ignoring nonexistent directory "/not/exist/usr/include"
#include "..." search starts here:
#include <...> search starts here:
BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include
BUILD/work/machine-oemllib32-linux-gnueabi/lib32-component/2.0.0-76-r7/recipe-sysroot-native/usr/bin/arm-oemllib32-linux-gnueabi/../../lib/arm-oemllib32-linux-gnueabi/gcc/arm-oemllib32-linux-gnueabi/9.2.0/include-fixed
End of search list.
On Tue, Aug 13, 2019 at 5:29 PM Khem Raj wrote:
> OE does not use the traditional /usr/lib/gcc prefix to store gcc-runtime
> it basically is moved into libdir, however some newer files were
> installed by newer versions of gcc especially libgomp ( omp.h openacc.h )
> into gcclibdir, so we have content in both directories, this confuses
> other tools which are trying to guess the gcc installation and its
> runtime location, since now we have two directories, the tools either
> choose one or other and we get inconsistent behavior, e.g. clang for
> aarch64 uses /usr/lib but same clang for riscv64 chose /usr/lib/gcc
>
> This change ensures that OE ends up with single valid location for gcc
> runtime files
>
> Move more common bits into common inc file
>
> Signed-off-by: Khem Raj
> ---
> v2: Divert packaging to use new path in whole recipe
>
> meta/recipes-devtools/gcc/gcc-runtime.inc| 18 +++---
> meta/recipes-devtools/gcc/gcc-runtime_8.3.bb | 10 --
> meta/recipes-devtools/gcc/gcc-runtime_9.1.bb | 10 --
> 3 files changed, 15 insertions(+), 23 deletions(-)
>
> diff --git a/meta/recipes-devtools/gcc/gcc-runtime.inc
> b/meta/recipes-devtools/gcc/gcc-runtime.inc
> index a5c2600d7f..22c1d78dd1 100644
> --- a/meta/recipes-devtools/gcc/gcc-runtime.inc
> +++ b/meta/recipes-devtools/gcc/gcc-runtime.inc
> @@ -17,6 +17,12 @@ EXTRA_OECONF_PATHS = "\
> EXTRA_OECONF_append_linuxstdbase = " --enable-clocale=gnu"
> EXTRA_OECONF_append = " --cache-file=${B}/config.cache"
>
> +# Disable ifuncs for libatomic on arm conflicts -march/-mcpu
> +EXTRA_OECONF_append_arm = " libat_cv_have_ifunc=no "
> +
> +# Building with thumb enabled on armv6t fails
> +ARM_INSTRUCTION_SET_armv6 = "arm"
> +
> RUNTIMELIBITM = "libitm"
> RUNTIMELIBITM_arc = ""
> RUNTIMELIBITM_mipsarch = ""
> @@ -77,6 +83,11 @@ do_install () {
> cd ${B}/${TARGET_SYS}/$d/
> oe_runmake 'DESTDIR=${D}'
> MULTIBUILDTOP=${B}/${TARGET_SYS}/$d/ install
> done
> + if [ -d ${D}${libdir}/gcc/${TARGET_SYS}/${BINV}/include ]; then
> + install -d ${D}${libdir}/${TARGET_SYS}/${BINV}/include
> + mv ${D}${libdir}/gcc/${TARGET_SYS}/${BINV}/include/*
> ${D}${libdir}/${TARGET_SYS}/${BINV}/include
> + rmdir --ignore-fail-on-non-empty -p
> ${D}${libdir}/gcc/${TARGET_SYS}/${BINV}/include
> + fi
>
[OE-core] [PATCH V2] gcc-runtime: Move content from gcclibdir into libdir
OE does not use the traditional /usr/lib/gcc prefix to store gcc-runtime
it basically is moved into libdir, however some newer files were
installed by newer versions of gcc especially libgomp ( omp.h openacc.h )
into gcclibdir, so we have content in both directories, this confuses
other tools which are trying to guess the gcc installation and its
runtime location, since now we have two directories, the tools either
choose one or other and we get inconsistent behavior, e.g. clang for
aarch64 uses /usr/lib but same clang for riscv64 chose /usr/lib/gcc
This change ensures that OE ends up with single valid location for gcc
runtime files
Move more common bits into common inc file
Signed-off-by: Khem Raj
---
v2: Divert packaging to use new path in whole recipe
meta/recipes-devtools/gcc/gcc-runtime.inc| 18 +++---
meta/recipes-devtools/gcc/gcc-runtime_8.3.bb | 10 --
meta/recipes-devtools/gcc/gcc-runtime_9.1.bb | 10 --
3 files changed, 15 insertions(+), 23 deletions(-)
diff --git a/meta/recipes-devtools/gcc/gcc-runtime.inc
b/meta/recipes-devtools/gcc/gcc-runtime.inc
index a5c2600d7f..22c1d78dd1 100644
--- a/meta/recipes-devtools/gcc/gcc-runtime.inc
+++ b/meta/recipes-devtools/gcc/gcc-runtime.inc
@@ -17,6 +17,12 @@ EXTRA_OECONF_PATHS = "\
EXTRA_OECONF_append_linuxstdbase = " --enable-clocale=gnu"
EXTRA_OECONF_append = " --cache-file=${B}/config.cache"
+# Disable ifuncs for libatomic on arm conflicts -march/-mcpu
+EXTRA_OECONF_append_arm = " libat_cv_have_ifunc=no "
+
+# Building with thumb enabled on armv6t fails
+ARM_INSTRUCTION_SET_armv6 = "arm"
+
RUNTIMELIBITM = "libitm"
RUNTIMELIBITM_arc = ""
RUNTIMELIBITM_mipsarch = ""
@@ -77,6 +83,11 @@ do_install () {
cd ${B}/${TARGET_SYS}/$d/
oe_runmake 'DESTDIR=${D}' MULTIBUILDTOP=${B}/${TARGET_SYS}/$d/
install
done
+ if [ -d ${D}${libdir}/gcc/${TARGET_SYS}/${BINV}/include ]; then
+ install -d ${D}${libdir}/${TARGET_SYS}/${BINV}/include
+ mv ${D}${libdir}/gcc/${TARGET_SYS}/${BINV}/include/*
${D}${libdir}/${TARGET_SYS}/${BINV}/include
+ rmdir --ignore-fail-on-non-empty -p
${D}${libdir}/gcc/${TARGET_SYS}/${BINV}/include
+ fi
rm -rf ${D}${infodir}/libgomp.info ${D}${infodir}/dir
rm -rf ${D}${infodir}/libitm.info ${D}${infodir}/dir
rm -rf ${D}${infodir}/libquadmath.info ${D}${infodir}/dir
@@ -205,7 +216,7 @@ FILES_libssp-dev = "\
${libdir}/libssp*.so \
${libdir}/libssp*_nonshared.a \
${libdir}/libssp*.la \
-${libdir}/gcc/${TARGET_SYS}/${BINV}/include/ssp \
+${libdir}/${TARGET_SYS}/${BINV}/include/ssp \
"
SUMMARY_libssp-dev = "GNU stack smashing protection library - development
files"
FILES_libssp-staticdev = "${libdir}/libssp*.a"
@@ -214,7 +225,7 @@ SUMMARY_libssp-staticdev = "GNU stack smashing protection
library - static devel
FILES_libquadmath = "${libdir}/libquadmath*.so.*"
SUMMARY_libquadmath = "GNU quad-precision math library"
FILES_libquadmath-dev = "\
-${libdir}/gcc/${TARGET_SYS}/${BINV}/include/quadmath* \
+${libdir}/${TARGET_SYS}/${BINV}/include/quadmath* \
${libdir}/libquadmath*.so \
${libdir}/libquadmath.la \
"
@@ -239,7 +250,8 @@ FILES_libgomp-dev = "\
${libdir}/libgomp*${SOLIBSDEV} \
${libdir}/libgomp*.la \
${libdir}/libgomp.spec \
-${libdir}/gcc/${TARGET_SYS}/${BINV}/include/omp.h \
+${libdir}/${TARGET_SYS}/${BINV}/include/omp.h \
+${libdir}/${TARGET_SYS}/${BINV}/include/openacc.h \
"
SUMMARY_libgomp-dev = "GNU OpenMP parallel programming library - development
files"
FILES_libgomp-staticdev = "${libdir}/libgomp*.a"
diff --git a/meta/recipes-devtools/gcc/gcc-runtime_8.3.bb
b/meta/recipes-devtools/gcc/gcc-runtime_8.3.bb
index a1c7a76d0b..dd430b57eb 100644
--- a/meta/recipes-devtools/gcc/gcc-runtime_8.3.bb
+++ b/meta/recipes-devtools/gcc/gcc-runtime_8.3.bb
@@ -1,12 +1,2 @@
require recipes-devtools/gcc/gcc-${PV}.inc
require gcc-runtime.inc
-
-# Disable ifuncs for libatomic on arm conflicts -march/-mcpu
-EXTRA_OECONF_append_arm = " libat_cv_have_ifunc=no "
-
-FILES_libgomp-dev += "\
-${libdir}/gcc/${TARGET_SYS}/${BINV}/include/openacc.h \
-"
-
-# Building with thumb enabled on armv6t fails
-ARM_INSTRUCTION_SET_armv6 = "arm"
diff --git a/meta/recipes-devtools/gcc/gcc-runtime_9.1.bb
b/meta/recipes-devtools/gcc/gcc-runtime_9.1.bb
index a1c7a76d0b..dd430b57eb 100644
--- a/meta/recipes-devtools/gcc/gcc-runtime_9.1.bb
+++ b/meta/recipes-devtools/gcc/gcc-runtime_9.1.bb
@@ -1,12 +1,2 @@
require recipes-devtools/gcc/gcc-${PV}.inc
require gcc-runtime.inc
-
-# Disable ifuncs for libatomic on arm conflicts -march/-mcpu
-EXTRA_OECONF_append_arm = " libat_cv_have_ifunc=no "
-
-FILES_libgomp-dev += "\
-${libdir}/gcc/${TARGET_SYS}/${BINV}/include/openacc.h \
-"
-
-# Building with thumb enabled on armv6t fails
-ARM_INSTRUCTION_SET_armv6 = "arm"
--
2.22.0
--
___
