[meta-oe][PATCH 0/3] libhugetlbfs: fix build problems with ix86 and armv7ab

[meta-oe][PATCH 0/3] libhugetlbfs: fix build problems with ix86 and armv7ab

[Gary S. Robertson]

OE builds of libhugetlbfs were failing for armv7ab and i586 (qemux86)
architectures.  In addition, once patched so builds completed, execution of
some tests was failing for armv7ab.

Patched OE recipe and Makefile to add support for 32-bit ix86 architectures...
(patches 1 and 2).
Then patched linker control script to include armv7ab in the set of
architectures whose text section offset and page size were corrected by
the script...(patch 3).

Patches 2 and 3 have been submitted upstream to the libhugetlbfs maintainers
and acceptance has been indicated for both...
however patch 2 has yet to appear in the upstream repository,
while patch 3 has been verified as merged.

OE builds of libhugetlbfs with these patches added have been successfully
completed here for qemux86, qemux86_64, armv7a, armv7ab, armv8, and armv8b
platforms, and tests run successfully on all except armv8b - which failed
with unrelated boot time issues.

[meta-oe][PATCH 1/3] libhugetlbfs: Add ix86 to list of compatible host architectures

[Gary S. Robertson]

From: "Gary S. Robertson" <gary.robertson@linaro.org>

COMPATIBLE_HOST included support for x86_64 but not for
32-bit x86 architectures - added generic 32-bit x86 arch names

Upstream Status - not applicable - recipe change only

Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
---
 .../libhugetlbfs/libhugetlbfs_git.bb               |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
index c2262ea..737acb9 100644
--- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
@@ -23,7 +23,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
 
 S = "${WORKDIR}/git"
 
-COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
+COMPATIBLE_HOST = "(i386|i486|i586|i686|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
 
 LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
 LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
-- 
1.7.9.5

[meta-oe][PATCH 2/3] libhugetlbfs: Make cross-devel recognize all ix86 arches

[Gary S. Robertson]

From: "Gary S. Robertson" <gary.robertson@linaro.org>

In a non-native cross-development scenario, the Makefile
only recognized i386 or x86_64 PC architectures.
Extended this to also recognize i486, i586, and i686

Upstream Status: Accepted but not yet applied by libhugetlbfs project

Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
---
 .../0001-Makefile-Recognize-all-ix86-arches.patch  |   32 ++++++++++++++++++++
 .../libhugetlbfs/libhugetlbfs_git.bb               |    1 +
 2 files changed, 33 insertions(+)
 create mode 100644 meta-oe/recipes-benchmark/libhugetlbfs/files/0001-Makefile-Recognize-all-ix86-arches.patch

diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-Makefile-Recognize-all-ix86-arches.patch b/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-Makefile-Recognize-all-ix86-arches.patch
new file mode 100644
index 0000000..2718275
--- /dev/null
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-Makefile-Recognize-all-ix86-arches.patch
@@ -0,0 +1,32 @@
+From a0166583ba5f7b6a6d2de434f633126fb12c9d29 Mon Sep 17 00:00:00 2001
+From: "Gary S. Robertson" <gary.robertson@linaro.org>
+Date: Wed, 24 Sep 2014 15:27:31 -0500
+Subject: [PATCH] Makefile: Recognize all ix86 arches
+
+In a non-native build scenario, the makefile
+only recognized i386 or x86_64 arches.  Added support
+to recognize i486, i586, i686.
+
+Upstream Status: Accepted by libhugetlbfs project
+
+Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
+---
+ Makefile |    2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+diff --git a/Makefile b/Makefile
+index 91502e1..0bfaee8 100644
+--- a/Makefile
++++ b/Makefile
+@@ -71,7 +71,7 @@ ELF64 = aarch64elf
+ TMPLIB64 = lib64
+ CUSTOM_LDSCRIPTS = no
+ else
+-ifeq ($(ARCH),i386)
++ifneq (,$(filter i386 i486 i586 i686,$(ARCH)))
+ CC32 = $(CC)
+ ELF32 = elf_i386
+ TMPLIB32 = lib
+-- 
+1.7.9.5
+
diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
index 737acb9..5111935 100644
--- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
@@ -19,6 +19,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
     file://0001-aarch64-fix-page-size-not-properly-computed.patch \
     file://0001-replace-lib-lib64-hardcoded-values-by-LIBDIR32-LIBDI.patch \
     file://0001-Extend-arm32-support-to-include-BE-variants.patch \
+    file://0001-Makefile-Recognize-all-ix86-arches.patch \
 "
 
 S = "${WORKDIR}/git"
-- 
1.7.9.5

[meta-oe][PATCH 3/3] libhugetlbfs: Fix page size & text offset for arm arches

[Gary S. Robertson]

From: "Gary S. Robertson" <gary.robertson@linaro.org>

Fixed computation of page size and text segment offset for various arm
architectures - including both LE and BE variants of armv7 as well as
aarch64

Upstream Status: Accepted at libhugetlbfs project

Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
---
 ...rch64-fix-page-size-not-properly-computed.patch |   32 -----------
 ...s-arm-arches-fix-page-size-and-text-offse.patch |   57 ++++++++++++++++++++
 .../libhugetlbfs/libhugetlbfs_git.bb               |    2 +-
 3 files changed, 58 insertions(+), 33 deletions(-)
 delete mode 100644 meta-oe/recipes-benchmark/libhugetlbfs/files/0001-aarch64-fix-page-size-not-properly-computed.patch
 create mode 100644 meta-oe/recipes-benchmark/libhugetlbfs/files/0001-ld.hugetlbfs-arm-arches-fix-page-size-and-text-offse.patch

diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-aarch64-fix-page-size-not-properly-computed.patch b/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-aarch64-fix-page-size-not-properly-computed.patch
deleted file mode 100644
index da60ba0..0000000
--- a/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-aarch64-fix-page-size-not-properly-computed.patch
+++ /dev/null
@@ -1,32 +0,0 @@
-Subject: [PATCH] aarch64: fix page size not properly computed
-
-There's 2 issues fixed:
-* typo on MB variable (MB -> $MB)
-* some linker variants are missing (linux and big endian)
-  - aarch64elfb
-  - aarch64linux
-  - aarch64linuxb
-
-Signed-off-by: Fathi Boudra <fathi.boudra@linaro.org>
-
-Upstream-Status: Submitted
----
- ld.hugetlbfs | 2 +-
- 1 file changed, 1 insertion(+), 1 deletion(-)
-
-diff --git a/ld.hugetlbfs b/ld.hugetlbfs
-index df446dd..6553547 100755
---- a/ld.hugetlbfs
-+++ b/ld.hugetlbfs
-@@ -87,7 +87,7 @@ elf32ppclinux|elf64ppc)	HPAGE_SIZE=$((16*$MB)) SLICE_SIZE=$((256*$MB)) ;;
- elf64lppc)		HPAGE_SIZE=$((16*$MB)) SLICE_SIZE=$((256*$MB)) ;;
- elf_i386|elf_x86_64)	HPAGE_SIZE=$((4*$MB)) SLICE_SIZE=$HPAGE_SIZE ;;
- elf_s390|elf64_s390)	HPAGE_SIZE=$((1*$MB)) SLICE_SIZE=$HPAGE_SIZE ;;
--armelf_linux_eabi|aarch64elf)	HPAGE_SIZE=$((2*MB)) SLICE_SIZE=$HPAGE_SIZE ;;
-+armelf_linux_eabi|aarch64elf*|aarch64linux*)	HPAGE_SIZE=$((2*$MB)) SLICE_SIZE=$HPAGE_SIZE ;;
- esac
- 
- if [ "$HTLB_ALIGN" == "slice" ]; then
--- 
-1.9.2
-
diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-ld.hugetlbfs-arm-arches-fix-page-size-and-text-offse.patch b/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-ld.hugetlbfs-arm-arches-fix-page-size-and-text-offse.patch
new file mode 100644
index 0000000..4e5b7bb
--- /dev/null
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-ld.hugetlbfs-arm-arches-fix-page-size-and-text-offse.patch
@@ -0,0 +1,57 @@
+From 1ab6f7f9b34fc745451140fc21db6763dffc5785 Mon Sep 17 00:00:00 2001
+From: "Gary S. Robertson" <gary.robertson@linaro.org>
+Date: Thu, 25 Sep 2014 14:57:06 -0500
+Subject: [PATCH] ld.hugetlbfs: arm arches - fix page size and text offset
+ setup
+
+There's 3 issues fixed:
+* typo on MB variable (MB -> $MB)
+* some linker variants are missing (linux and big endian)
+  - armelfb_linux_eabi
+  - aarch64elfb
+  - aarch64linux
+  - aarch64linuxb
+* text segment offset was not specified for armv7 BE
+
+NOTE: This patch replaces and makes obsolete the following previously
+      submitted patch:
+
+      aarch64: fix page size not properly computed
+      by: Fathi Boudra <fathi.boudra@linaro.org>
+
+Upstream Status: Accepted at libhugetlbfs project
+
+Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
+---
+ ld.hugetlbfs |    8 ++++----
+ 1 file changed, 4 insertions(+), 4 deletions(-)
+
+diff --git a/ld.hugetlbfs b/ld.hugetlbfs
+index df446dd..4417442 100755
+--- a/ld.hugetlbfs
++++ b/ld.hugetlbfs
+@@ -87,7 +87,7 @@ elf32ppclinux|elf64ppc)	HPAGE_SIZE=$((16*$MB)) SLICE_SIZE=$((256*$MB)) ;;
+ elf64lppc)		HPAGE_SIZE=$((16*$MB)) SLICE_SIZE=$((256*$MB)) ;;
+ elf_i386|elf_x86_64)	HPAGE_SIZE=$((4*$MB)) SLICE_SIZE=$HPAGE_SIZE ;;
+ elf_s390|elf64_s390)	HPAGE_SIZE=$((1*$MB)) SLICE_SIZE=$HPAGE_SIZE ;;
+-armelf_linux_eabi|aarch64elf)	HPAGE_SIZE=$((2*MB)) SLICE_SIZE=$HPAGE_SIZE ;;
++armelf*_linux_eabi|aarch64elf*|aarch64linux*)	HPAGE_SIZE=$((2*$MB)) SLICE_SIZE=$HPAGE_SIZE ;;
+ esac
+ 
+ if [ "$HTLB_ALIGN" == "slice" ]; then
+@@ -96,9 +96,9 @@ if [ "$HTLB_ALIGN" == "slice" ]; then
+ 
+ 	# targeting the ARM platform one needs to explicitly set the text segment offset
+ 	# otherwise it will be NULL.
+-	if [ "$EMU" == "armelf_linux_eabi" ]; then
+-		HTLBOPTS="$HTLBOPTS -Ttext-segment=$SLICE_SIZE"
+-	fi
++	case "$EMU" in
++	armelf*_linux_eabi)	HTLBOPTS="$HTLBOPTS -Ttext-segment=$SLICE_SIZE" ;;
++	esac
+ fi
+ 
+ ${LD} "${args[@]}" ${HTLBOPTS}
+-- 
+1.7.9.5
+
diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
index 5111935..72d5c66 100644
--- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
@@ -16,7 +16,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
     file://tests-Makefile-install-static-4G-edge-testcases.patch \
     file://0001-run_test.py-not-use-hard-coded-path-.-obj-hugeadm.patch \
     file://0001-aarch64-fix-cross-compilation.patch \
-    file://0001-aarch64-fix-page-size-not-properly-computed.patch \
+    file://0001-ld.hugetlbfs-arm-arches-fix-page-size-and-text-offse.patch \
     file://0001-replace-lib-lib64-hardcoded-values-by-LIBDIR32-LIBDI.patch \
     file://0001-Extend-arm32-support-to-include-BE-variants.patch \
     file://0001-Makefile-Recognize-all-ix86-arches.patch \
-- 
1.7.9.5

Re: [meta-oe][PATCH 1/3] libhugetlbfs: Add ix86 to list of compatible host architectures

[Gary Robertson]

I am wondering if this patch was overlooked, as I have seen no commentary
regarding it, and the two other patches which accompanied it have been
merged into the meta-oe libhugetlbfs recipe already.
I added explicit arch designators for the typically-used 32-bit intel
architectures so the recipe could be built for those platforms and not just
for 64-bit intel platforms.

Thanks & have a happy holiday season!

Gary Robertson

On Mon, Nov 24, 2014 at 4:59 PM, Gary S. Robertson <
gary.robertson@linaro.org> wrote:
>
> From: "Gary S. Robertson" <gary.robertson@linaro.org>
>
> COMPATIBLE_HOST included support for x86_64 but not for
> 32-bit x86 architectures - added generic 32-bit x86 arch names
>
> Upstream Status - not applicable - recipe change only
>
> Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
> ---
>  .../libhugetlbfs/libhugetlbfs_git.bb               |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> index c2262ea..737acb9 100644
> --- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> +++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> @@ -23,7 +23,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
>
>  S = "${WORKDIR}/git"
>
> -COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> +COMPATIBLE_HOST =
> "(i386|i486|i586|i686|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
>
>  LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
>  LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
> --
> 1.7.9.5
>
>

Re: [meta-oe][PATCH 1/3] libhugetlbfs: Add ix86 to list of compatible host architectures

[Martin Jansa]

I've intentionally skipped it, because it failed to build.

Unfortunately I've reported the issue in reply to different patch which
wasn't the cause of the error:
http://lists.openembedded.org/pipermail/openembedded-devel/2014-December/099239.html

On Fri, Dec 19, 2014 at 10:29 PM, Gary Robertson <gary.robertson@linaro.org>
wrote:
>
> I am wondering if this patch was overlooked, as I have seen no commentary
> regarding it, and the two other patches which accompanied it have been
> merged into the meta-oe libhugetlbfs recipe already.
> I added explicit arch designators for the typically-used 32-bit intel
> architectures so the recipe could be built for those platforms and not just
> for 64-bit intel platforms.
>
> Thanks & have a happy holiday season!
>
> Gary Robertson
>
> On Mon, Nov 24, 2014 at 4:59 PM, Gary S. Robertson <
> gary.robertson@linaro.org> wrote:
> >
> > From: "Gary S. Robertson" <gary.robertson@linaro.org>
> >
> > COMPATIBLE_HOST included support for x86_64 but not for
> > 32-bit x86 architectures - added generic 32-bit x86 arch names
> >
> > Upstream Status - not applicable - recipe change only
> >
> > Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
> > ---
> >  .../libhugetlbfs/libhugetlbfs_git.bb               |    2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > index c2262ea..737acb9 100644
> > --- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > +++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > @@ -23,7 +23,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
> >
> >  S = "${WORKDIR}/git"
> >
> > -COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> > +COMPATIBLE_HOST =
> > "(i386|i486|i586|i686|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> >
> >  LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
> >  LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
> > --
> > 1.7.9.5
> >
> >
> --
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel
>

Re: [meta-oe][PATCH 1/3] libhugetlbfs: Add ix86 to list of compatible host architectures

[Gary Robertson]

Hmm...  Perhaps I goofed with the way I submitted the patches -
I'm afraid I'm a relative noob at patch submission, and still learning best
practices.

The recipe won't build for qemux86 without also applying the Makefile patch
in
'[meta-oe][PATCH 2/3] libhugetlbfs: Make cross-devel recognize all ix86
arches'
(BTW, that Makefile patch has now also been merged into libhugetlbfs.)

So if the patches were applied and tested separately in numerical sequence,
I would expect the qemux86 build to fail with only the first patch applied.

However, I generated these patches specifically to enable me to build
libhugetlbfs for qemux86, and with both patches in place I have been
building successfully for both i586 (qemux86) and x86_64 intel
architectures.

Just for confirmation I rebuilt tonight for qemux86, and had no problems.
So could I ask you to give this patch another try with the other patches
already in place?

Thanks again - and thanks also for the fast reply... looking at the OE lists
I know there is lots of activity to be dealt with.

Gay Robertson

On Fri, Dec 19, 2014 at 5:07 PM, Martin Jansa <martin.jansa@gmail.com>
wrote:
>
> I've intentionally skipped it, because it failed to build.
>
> Unfortunately I've reported the issue in reply to different patch which
> wasn't the cause of the error:
>
> http://lists.openembedded.org/pipermail/openembedded-devel/2014-December/099239.html
>
> On Fri, Dec 19, 2014 at 10:29 PM, Gary Robertson <
> gary.robertson@linaro.org>
> wrote:
> >
> > I am wondering if this patch was overlooked, as I have seen no commentary
> > regarding it, and the two other patches which accompanied it have been
> > merged into the meta-oe libhugetlbfs recipe already.
> > I added explicit arch designators for the typically-used 32-bit intel
> > architectures so the recipe could be built for those platforms and not
> just
> > for 64-bit intel platforms.
> >
> > Thanks & have a happy holiday season!
> >
> > Gary Robertson
> >
> > On Mon, Nov 24, 2014 at 4:59 PM, Gary S. Robertson <
> > gary.robertson@linaro.org> wrote:
> > >
> > > From: "Gary S. Robertson" <gary.robertson@linaro.org>
> > >
> > > COMPATIBLE_HOST included support for x86_64 but not for
> > > 32-bit x86 architectures - added generic 32-bit x86 arch names
> > >
> > > Upstream Status - not applicable - recipe change only
> > >
> > > Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
> > > ---
> > >  .../libhugetlbfs/libhugetlbfs_git.bb               |    2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/
> libhugetlbfs_git.bb
> > > b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > > index c2262ea..737acb9 100644
> > > --- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > > +++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > > @@ -23,7 +23,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code
> \
> > >
> > >  S = "${WORKDIR}/git"
> > >
> > > -COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> > > +COMPATIBLE_HOST =
> > > "(i386|i486|i586|i686|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> > >
> > >  LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
> > >  LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
> > > --
> > > 1.7.9.5
> > >
> > >
> > --
> > _______________________________________________
> > Openembedded-devel mailing list
> > Openembedded-devel@lists.openembedded.org
> > http://lists.openembedded.org/mailman/listinfo/openembedded-devel
> >
> --
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel
>

Re: [meta-oe][PATCH 1/3] libhugetlbfs: Add ix86 to list of compatible host architectures

[Martin Jansa]

> So could I ask you to give this patch another try with the other patches
already in place?

I'll do that if I don't forget, but I'm pretty sure that first it was
failing with both your patches applied together and then again with all 3
pending libhugetlbfs together.

You have submitted it correctly and it was applied correctly, I always
trigger the build with as many cherry-picked changes as possible, only
after discovering the qemux86 build issue I've tried to remove this 1/3
patch from master-next to confirm that this is the root cause for the build
issue and then I've applied the rest, because it was building fine again.

Regards,



On Sat, Dec 20, 2014 at 6:56 AM, Gary Robertson <gary.robertson@linaro.org>
wrote:

> Hmm...  Perhaps I goofed with the way I submitted the patches -
> I'm afraid I'm a relative noob at patch submission, and still learning best
> practices.
>
> The recipe won't build for qemux86 without also applying the Makefile patch
> in
> '[meta-oe][PATCH 2/3] libhugetlbfs: Make cross-devel recognize all ix86
> arches'
> (BTW, that Makefile patch has now also been merged into libhugetlbfs.)
>
> So if the patches were applied and tested separately in numerical sequence,
> I would expect the qemux86 build to fail with only the first patch applied.
>
> However, I generated these patches specifically to enable me to build
> libhugetlbfs for qemux86, and with both patches in place I have been
> building successfully for both i586 (qemux86) and x86_64 intel
> architectures.
>
> Just for confirmation I rebuilt tonight for qemux86, and had no problems.
> So could I ask you to give this patch another try with the other patches
> already in place?
>
> Thanks again - and thanks also for the fast reply... looking at the OE
> lists
> I know there is lots of activity to be dealt with.
>
> Gay Robertson
>
> On Fri, Dec 19, 2014 at 5:07 PM, Martin Jansa <martin.jansa@gmail.com>
> wrote:
> >
> > I've intentionally skipped it, because it failed to build.
> >
> > Unfortunately I've reported the issue in reply to different patch which
> > wasn't the cause of the error:
> >
> >
> http://lists.openembedded.org/pipermail/openembedded-devel/2014-December/099239.html
> >
> > On Fri, Dec 19, 2014 at 10:29 PM, Gary Robertson <
> > gary.robertson@linaro.org>
> > wrote:
> > >
> > > I am wondering if this patch was overlooked, as I have seen no
> commentary
> > > regarding it, and the two other patches which accompanied it have been
> > > merged into the meta-oe libhugetlbfs recipe already.
> > > I added explicit arch designators for the typically-used 32-bit intel
> > > architectures so the recipe could be built for those platforms and not
> > just
> > > for 64-bit intel platforms.
> > >
> > > Thanks & have a happy holiday season!
> > >
> > > Gary Robertson
> > >
> > > On Mon, Nov 24, 2014 at 4:59 PM, Gary S. Robertson <
> > > gary.robertson@linaro.org> wrote:
> > > >
> > > > From: "Gary S. Robertson" <gary.robertson@linaro.org>
> > > >
> > > > COMPATIBLE_HOST included support for x86_64 but not for
> > > > 32-bit x86 architectures - added generic 32-bit x86 arch names
> > > >
> > > > Upstream Status - not applicable - recipe change only
> > > >
> > > > Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
> > > > ---
> > > >  .../libhugetlbfs/libhugetlbfs_git.bb               |    2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/
> > libhugetlbfs_git.bb
> > > > b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > > > index c2262ea..737acb9 100644
> > > > --- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > > > +++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> > > > @@ -23,7 +23,7 @@ SRC_URI = "git://
> git.code.sf.net/p/libhugetlbfs/code
> > \
> > > >
> > > >  S = "${WORKDIR}/git"
> > > >
> > > > -COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> > > > +COMPATIBLE_HOST =
> > > > "(i386|i486|i586|i686|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> > > >
> > > >  LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
> > > >  LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
> > > > --
> > > > 1.7.9.5
> > > >
> > > >
> > > --
> > > _______________________________________________
> > > Openembedded-devel mailing list
> > > Openembedded-devel@lists.openembedded.org
> > > http://lists.openembedded.org/mailman/listinfo/openembedded-devel
> > >
> > --
> > _______________________________________________
> > Openembedded-devel mailing list
> > Openembedded-devel@lists.openembedded.org
> > http://lists.openembedded.org/mailman/listinfo/openembedded-devel
> >
> --
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel
>

[meta-oe][PATCH V2] libhugetlbfs: add i586 to compatible hosts

[Gary S. Robertson]

I am revisiting this patch after a long interlude of doing other work...
please accept my apologies for resurrecting such an ancient thread.

After speaking with Koen Kool at Linaro Connect Hong Kong 2015,
it dawned on me that the only 32-bit ix86 architecture I had actually tested
was i586, and that the other arch designators might indeed somehow cause
build breakage in an OE build encompassing those architectures.

Consequently I have revised the patch such that it extends the list of
compatible hosts by adding ONLY the i586 architecture which I have tested
successfully.  I hope you will have similar success and can apply the
amended patch.

Regards,

Gary Robertson

[PATCH V2] libhugetlbfs: Add i586 to list of compatible hosts

[Gary S. Robertson]

COMPATIBLE_HOST included support for x86_64 but not for
32-bit x86 architectures - added i586 arch name to cover qemux86, etc.

Upstream Status - not applicable - recipe change only

Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
---
 meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
index 4ce6a19..6fe5dce 100644
--- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
@@ -24,7 +24,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
 
 S = "${WORKDIR}/git"
 
-COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
+COMPATIBLE_HOST = "(i586|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
 
 LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
 LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
-- 
1.9.1

[meta-oe][PATCH V3] libhugetlbfs: add i586 to compatible hosts

[Gary S. Robertson]

Some months back in your last comment regarding this proposed patch,
you indicated that my original version of this patch had broken
one or more builds when applied.

That version added i386, i486, i586, and i686 ito the list of compatible
host architectures for libhugetlbfs in an effort to cover all 32-bit
ix86 variants.

After speaking with Koen Kool at Linaro Connect Hong Kong 2015,
it dawned on me that the only 32-bit ix86 architecture I had actually tested
was i586, and that the other arch designators might indeed somehow cause
build breakage in an OE build encompassing those architectures.

Consequently I have revised the patch such that it extends the list of
compatible hosts by adding ONLY the i586 architecture which I have tested
successfully.  I hope you will have similar success and can apply the
amended patch.

Finally your response to another recent patch submission I made caused me
to realize that the 'Upstream-Status' line in this commit was malformed -
so I am resubmitting the commit with a more compliant upstream status.

Regards,

Gary Robertson

[PATCH] libhugetlbfs: Add i586 to list of compatible host architectures

[Gary S. Robertson]

COMPATIBLE_HOST included support for x86_64 but not for
32-bit x86 architectures - added i586 arch name to cover qemux86, etc.

Upstream-Status: Inappropriate - OE recipe change only

Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
---
 meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
index 4ce6a19..6fe5dce 100644
--- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
@@ -24,7 +24,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
 
 S = "${WORKDIR}/git"
 
-COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
+COMPATIBLE_HOST = "(i586|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
 
 LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
 LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
-- 
1.9.1

Re: [PATCH] libhugetlbfs: Add i586 to list of compatible host architectures

[Otavio Salvador]

On Wed, Apr 1, 2015 at 7:23 PM, Gary S. Robertson
<gary.robertson@linaro.org> wrote:
> COMPATIBLE_HOST included support for x86_64 but not for
> 32-bit x86 architectures - added i586 arch name to cover qemux86, etc.
>
> Upstream-Status: Inappropriate - OE recipe change only

This field is used in patch files, not commit log of the Git changes. Drop this.

> Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
> ---
>  meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> index 4ce6a19..6fe5dce 100644
> --- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> +++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> @@ -24,7 +24,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
>
>  S = "${WORKDIR}/git"
>
> -COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> +COMPATIBLE_HOST = "(i586|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
>
>  LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
>  LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
> --
> 1.9.1
>
> --
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel



-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: [PATCH] libhugetlbfs: Add i586 to list of compatible host architectures

[Otavio Salvador]

On Wed, Apr 1, 2015 at 9:31 PM, Otavio Salvador <otavio@ossystems.com.br> wrote:
> On Wed, Apr 1, 2015 at 7:23 PM, Gary S. Robertson
> <gary.robertson@linaro.org> wrote:
>> COMPATIBLE_HOST included support for x86_64 but not for
>> 32-bit x86 architectures - added i586 arch name to cover qemux86, etc.
>>
>> Upstream-Status: Inappropriate - OE recipe change only
>
> This field is used in patch files, not commit log of the Git changes. Drop this.

Also, respect version of patch. I think your next send is v4, isn't it?

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

[meta-oe][PATCH V4] libhugetlbfs: add i586 to compatible hosts

[Gary S. Robertson]

Thanks also for pointing out that I neglected to update the patch version
in the commit message itself as opposed to the cover email subject line.

I believe I have corrected both oversights with this submission and will
try to avoid these mistakes in any future submissions to OE.
Hopefully with the formatting issues resolved this patch can finally be
validated and merged.

Thank you all for your patience.

Regards,

Gary Robertson

[PATCH V4] libhugetlbfs: Add i586 to list of compatible host architectures

[Gary S. Robertson]

COMPATIBLE_HOST included support for x86_64 but not for
32-bit x86 architectures - added i586 arch name to cover qemux86, etc.

Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
---
 meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
index 4ce6a19..6fe5dce 100644
--- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
@@ -24,7 +24,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
 
 S = "${WORKDIR}/git"
 
-COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
+COMPATIBLE_HOST = "(i586|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
 
 LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
 LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
-- 
1.9.1

Re: [PATCH V4] libhugetlbfs: Add i586 to list of compatible host architectures

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 1554 bytes --]

On Thu, Apr 02, 2015 at 11:04:03AM -0500, Gary S. Robertson wrote:
> COMPATIBLE_HOST included support for x86_64 but not for
> 32-bit x86 architectures - added i586 arch name to cover qemux86, etc.

But it still fails for qemux86 :(

http://errors.yoctoproject.org/Errors/Details/9749/

Please fix it ASAP, because I've already merged this change before
noticing this new build error.

Thanks,

> Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
> ---
>  meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> index 4ce6a19..6fe5dce 100644
> --- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> +++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> @@ -24,7 +24,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
>  
>  S = "${WORKDIR}/git"
>  
> -COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> +COMPATIBLE_HOST = "(i586|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
>  
>  LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
>  LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
> -- 
> 1.9.1
> 
> -- 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 188 bytes --]

[meta-oe][PATCH] libhugetlbfs: x86 - don't link against host libs

[Gary S. Robertson]

The x86 build linker scripts were still searching hard-coded paths on the
build host for libraries.  As it happens, the libraries needed for
linking the failing executable are present on my build host, so the build
succeeds here despite the erroneous linker paths.  Consequently I had
stupidly assumed the resulting QA warning was a fairly benign error
until your test build in a pristine environment demonstrated otherwise.

I'm unclear as to the exact behavior shown by your log of the failed build,
but since it is a linker error I am for now assuming it is precipitated
somehow by the flawed library search paths.  In any case with the
hard-coded host paths commented out I get a clean build without the QA
warning and see no similar logs in my local builds - so I hope this will
also fix the build failures on your end.

Regards,

Gary Robertson

[PATCH] libhugetlbfs: x86 - don't link against host libs

[Gary S. Robertson]

An earlier patch from Chunrong Guo <B40290@freescale.com> fixed the
linker command scripts in ldscripts to comment out hard-coded paths
for the linker search directories.  This allows the command line
to specifiy the proper paths without being overridden.
However, the 32-bit ix86 scripts evidently were added after this
earlier patch was created and were not corrected by it - so here
we simply duplicate the technique in the earlier patch and extend it
to correct the linker scripts for 32-bit ix86 builds.

Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
---
 ...86-avoid-host-library-search-in-cross-dev.patch | 48 ++++++++++++++++++++++
 .../libhugetlbfs/libhugetlbfs_git.bb               |  1 +
 2 files changed, 49 insertions(+)
 create mode 100644 meta-oe/recipes-benchmark/libhugetlbfs/files/0001-x86-avoid-host-library-search-in-cross-dev.patch

diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-x86-avoid-host-library-search-in-cross-dev.patch b/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-x86-avoid-host-library-search-in-cross-dev.patch
new file mode 100644
index 0000000..959574e
--- /dev/null
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/files/0001-x86-avoid-host-library-search-in-cross-dev.patch
@@ -0,0 +1,48 @@
+From b028ae9f43111677056f2661c800096f1994b687 Mon Sep 17 00:00:00 2001
+From: "Gary S. Robertson" <gary.robertson@linaro.org>
+Date: Mon, 13 Apr 2015 14:41:15 -0500
+Subject: [PATCH] libhugetlbfs: x86 avoid host library search in cross-dev
+
+Upstream-Status: Inappropriate [oe-core specific]
+
+The ldscripts for 32-bit ix86 architectures were not corrected by the earlier
+patch from Chunrong Guo <B40290@freescale.com>... presumably they were added
+after that patch was created.  Here we extend the same technique to include
+the 32-bit ix86 linker scripts, so the command line variables can set the
+linker search directories rather than using hard-coded paths in the ldscripts.
+
+Signed-off-by: Gary S. Robertson <gary.robertson@linaro.org>
+---
+ ldscripts/elf_i386.xB   | 2 +-
+ ldscripts/elf_i386.xBDT | 2 +-
+ 2 files changed, 2 insertions(+), 2 deletions(-)
+
+diff --git a/ldscripts/elf_i386.xB b/ldscripts/elf_i386.xB
+index 43fe51c..b852781 100644
+--- a/ldscripts/elf_i386.xB
++++ b/ldscripts/elf_i386.xB
+@@ -3,7 +3,7 @@ OUTPUT_FORMAT("elf32-i386", "elf32-i386",
+ 	      "elf32-i386")
+ OUTPUT_ARCH(i386)
+ ENTRY(_start)
+-SEARCH_DIR("/usr/i486-linux-gnu/lib"); SEARCH_DIR("/usr/local/lib"); SEARCH_DIR("/lib"); SEARCH_DIR("/usr/lib");
++/*SEARCH_DIR("/usr/i486-linux-gnu/lib"); SEARCH_DIR("/usr/local/lib"); SEARCH_DIR("/lib"); SEARCH_DIR("/usr/lib");*/
+ INPUT(-lhugetlbfs);
+ /* Do we need any of these for elf?
+    __DYNAMIC = 0;    */
+diff --git a/ldscripts/elf_i386.xBDT b/ldscripts/elf_i386.xBDT
+index d72aebe..b25e225 100644
+--- a/ldscripts/elf_i386.xBDT
++++ b/ldscripts/elf_i386.xBDT
+@@ -3,7 +3,7 @@ OUTPUT_FORMAT("elf32-i386", "elf32-i386",
+ 	      "elf32-i386")
+ OUTPUT_ARCH(i386)
+ ENTRY(_start)
+-SEARCH_DIR("/usr/i486-linux-gnu/lib"); SEARCH_DIR("/usr/local/lib"); SEARCH_DIR("/lib"); SEARCH_DIR("/usr/lib");
++/*SEARCH_DIR("/usr/i486-linux-gnu/lib"); SEARCH_DIR("/usr/local/lib"); SEARCH_DIR("/lib"); SEARCH_DIR("/usr/lib");*/
+ INPUT(-lhugetlbfs);
+ /* Do we need any of these for elf?
+    __DYNAMIC = 0;    */
+-- 
+1.9.1
+
diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
index 6fe5dce..25f1672 100644
--- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
@@ -20,6 +20,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
     file://0001-replace-lib-lib64-hardcoded-values-by-LIBDIR32-LIBDI.patch \
     file://0001-Extend-arm32-support-to-include-BE-variants.patch \
     file://0001-Makefile-Recognize-all-ix86-arches.patch \
+    file://0001-x86-avoid-host-library-search-in-cross-dev.patch \
 "
 
 S = "${WORKDIR}/git"
-- 
1.9.1

Re: [meta-oe][PATCH] libhugetlbfs: x86 - don't link against host libs

[Martin Jansa]

On Mon, Apr 13, 2015 at 11:18:59PM -0500, Gary S. Robertson wrote:
> The x86 build linker scripts were still searching hard-coded paths on the
> build host for libraries.  As it happens, the libraries needed for
> linking the failing executable are present on my build host, so the build
> succeeds here despite the erroneous linker paths.  Consequently I had
> stupidly assumed the resulting QA warning was a fairly benign error
> until your test build in a pristine environment demonstrated otherwise.
> 
> I'm unclear as to the exact behavior shown by your log of the failed build,
> but since it is a linker error I am for now assuming it is precipitated
> somehow by the flawed library search paths.  In any case with the
> hard-coded host paths commented out I get a clean build without the QA
> warning and see no similar logs in my local builds - so I hope this will
> also fix the build failures on your end.

Thanks for quick response, but this doesn't seem to be enough:

| /home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/x86_64-linux/usr/libexec/i586-oe-linux/gcc/i586-oe-linux/4.9.2/ld: Attempt to open /home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86/lib/libgcc_s.so succeeded
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -I.. -O2 -Wall -g -o obj32/zero_filesize_segment.o -c zero_filesize_segment.c
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed -Wl,-z,noexecstack -ldl -L../obj32 -o obj32/zero_filesize_segment -Lobj32 zero_filesize_segment.ld obj32/zero_filesize_segment.o obj32/testutils.o -Wl,--no-as-needed -lpthread -ldl -lhugetlbfs_privutils || cp bad-toolchain.sh obj32/zero_filesize_segment
| /home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/x86_64-linux/usr/libexec/i586-oe-linux/gcc/i586-oe-linux/4.9.2/ld: error: zero_filesize_segment.ld: SECTIONS seen after other input files; try -T/--script
| /home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/x86_64-linux/usr/libexec/i586-oe-linux/gcc/i586-oe-linux/4.9.2/ld: internal error in input_sections_begin, at /home/jenkins/oe/world/shr-core/tmp-glibc/work/x86_64-oe-linux/binutils-cross-i586/2.25-r0/git/gold/merge.h:269
| collect2: error: ld returned 1 exit status
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -I.. -O2 -Wall -g -o obj32/linkhuge.o -c linkhuge.c
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed -Wl,-z,noexecstack -ldl -L../obj32 -o obj32/linkhuge obj32/linkhuge.o obj32/testutils.o -Wl,--no-as-needed -lpthread -ldl -lhugetlbfs_privutils
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -I.. -O2 -Wall -g -o obj32/linkhuge_nofd.o -c linkhuge_nofd.c
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed -Wl,-z,noexecstack -ldl -L../obj32 -o obj32/linkhuge_nofd obj32/linkhuge_nofd.o obj32/testutils.o -Wl,--no-as-needed -lpthread -ldl -lhugetlbfs_privutils
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -I.. -O2 -Wall -g -o obj32/linkshare.o -c linkshare.c
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed -Wl,-z,noexecstack -ldl -L../obj32 -o obj32/linkshare obj32/linkshare.o obj32/testutils.o -Wl,--no-as-needed -lpthread -ldl -lhugetlbfs_privutils
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -I.. -O2 -Wall -g -o obj32/linkhuge_rw.o -c linkhuge_rw.c
| i586-oe-linux-gcc  -m32 -march=i586 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed --sysroot=/home/jenkins/oe/world/shr-core/tmp-glibc/sysroots/qemux86 -B./obj32 -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed -Wl,-z,noexecstack -ldl -L../obj32 -o obj32/linkhuge_rw -Wl,--no-as-needed -lpthread -ldl -lhugetlbfs_privutils -Wl,--hugetlbfs-align obj32/linkhuge_rw.o obj32/testutils.o
| i586-oe-linux-ld: internal error in do_write, at /home/jenkins/oe/world/shr-core/tmp-glibc/work/x86_64-oe-linux/binutils-cross-i586/2.25-r0/git/gold/output.cc:464
| collect2: error: ld returned 1 exit status

Please try to build qemux86 with gold (ld-is-gold in DISTRO_FEATURES) to see if you reproduce this issue.

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[meta-oe][PATCH] Revert "libhugetlbfs: Add i586 to list of compatible host architectures"

[Martin Jansa]

It's still failing in world builds even when
"libhugetlbfs: x86 - don't link against host libs" is applied as well.

It was reported month ago:
https://www.mail-archive.com/openembedded-devel@lists.openembedded.org/msg41565.html

This reverts commit 3cc7f186be46a0de049d1fab70ddb84111be6e29.
---
 meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
index 6fe5dce..4ce6a19 100644
--- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
+++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
@@ -24,7 +24,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
 
 S = "${WORKDIR}/git"
 
-COMPATIBLE_HOST = "(i586|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
+COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
 
 LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
 LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
-- 
2.4.0

Re: [meta-oe][PATCH] Revert "libhugetlbfs: Add i586 to list of compatible host architectures"

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 1854 bytes --]


> On May 13, 2015, at 5:48 AM, Martin Jansa <martin.jansa@gmail.com> wrote:
> 
> It's still failing in world builds even when
> "libhugetlbfs: x86 - don't link against host libs" is applied as well.
> 
> It was reported month ago:
> https://www.mail-archive.com/openembedded-devel@lists.openembedded.org/msg41565.html
> 

From this message it seems to me that linker script used for x86 is not digested well by gold and there
might be incompatibilities and gold certainly should not segfault. so its worth filing a bug report against gold as well.
Then there should be another one for libhugetlbfs to ensure that linker script is portable. But then in OE we can use

LDFLAGS += “-fuse-ld=bfd”

in recipe and that should get us going.

Avoid the revert

> This reverts commit 3cc7f186be46a0de049d1fab70ddb84111be6e29.
> ---
> meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> index 6fe5dce..4ce6a19 100644
> --- a/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> +++ b/meta-oe/recipes-benchmark/libhugetlbfs/libhugetlbfs_git.bb
> @@ -24,7 +24,7 @@ SRC_URI = "git://git.code.sf.net/p/libhugetlbfs/code \
> 
> S = "${WORKDIR}/git"
> 
> -COMPATIBLE_HOST = "(i586|x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> +COMPATIBLE_HOST = "(x86_64|powerpc|powerpc64|aarch64|arm).*-linux*"
> 
> LIBARGS = "LIB32=${baselib} LIB64=${baselib}"
> LIBHUGETLBFS_ARCH = "${TARGET_ARCH}"
> --
> 2.4.0
> 
> --
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel


[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 211 bytes --]