On 11/12/2018 15.17, Bruce Ashfield wrote:
On Tue, Dec 11, 2018 at 2:49 AM He Zhe <zhe...@windriver.com> wrote:
On 2018/12/11 00:58, Bruce Ashfield wrote:
On Wed, Nov 21, 2018 at 9:06 AM <zhe...@windriver.com
<mailto:zhe...@windriver.com>> wrote:
From: He Zhe <zhe...@windriver.com <mailto:zhe...@windriver.com>>
This is a workaround for the following possible build failure.
*** Compiler lacks asm-goto support.. Stop.
When building linux-libc-headers we need to use binutils on build machine.
binutils v2.31 introduces a bug that could cause scripts/gcc-goto.sh to
fail
when running in an environment where /tmp is rarely used, e.g. in docker.
Signed-off-by: He Zhe <zhe...@windriver.com <mailto:zhe...@windriver.com>>
---
v2: Improve commit log
v3: Shorten long lines as patchwork suggested
v4: Shorten subject line...
...-fixed-temporary-file-instead-of-pipe-for.patch | 70
++++++++++++++++++++++
.../linux-libc-headers/linux-libc-headers_4.18.bb
<http://linux-libc-headers_4.18.bb> | 4 ++
2 files changed, 74 insertions(+)
create mode 100644
meta/recipes-kernel/linux-libc-headers/linux-libc-headers/0001-scripts-Use-fixed-temporary-file-instead-of-pipe-for.patch
diff --git
a/meta/recipes-kernel/linux-libc-headers/linux-libc-headers/0001-scripts-Use-fixed-temporary-file-instead-of-pipe-for.patch
b/meta/recipes-kernel/linux-libc-headers/linux-libc-headers/0001-scripts-Use-fixed-temporary-file-instead-of-pipe-for.patch
new file mode 100644
index 0000000..0d8fa80
--- /dev/null
+++
b/meta/recipes-kernel/linux-libc-headers/linux-libc-headers/0001-scripts-Use-fixed-temporary-file-instead-of-pipe-for.patch
@@ -0,0 +1,70 @@
+From 3bbea65e11918f8753e8006a2198b999cdb0af58 Mon Sep 17 00:00:00 2001
+From: He Zhe <zhe...@windriver.com <mailto:zhe...@windriver.com>>
+Date: Wed, 21 Nov 2018 15:12:43 +0800
+Subject: [PATCH] scripts: Use fixed temporary file instead of pipe for
+ here-doc
+
+There was a bug of "as" in binutils that when it checks if the input file
and
+output file are the same one, it would not check if they are on the same
block
+device. The check is introduced by the following commit in v2.31.
+
+https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=
+67f846b59b32f3d704c601669409c2584383fea9
<https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=+67f846b59b32f3d704c601669409c2584383fea9>
Can you double check the links to the upstream changes ? They didn't work
for me.
I guess that's due to the leading "+" prepended to the second line of the link.
It's part of the patch, not of the link.
That's not what I'm seeing. Even when I select and paste the link, I'm getting
a: "400 - Invalid hash parameter" back from the upstream git repo.
This link works for me:
https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=67f846b59b32f3d704c601669409c2584383fea9
Which is the same as the one in the patch without the + after the line
break.
// Martin
+
+The here-doc usage in this script creates temporary file in /tmp. When we
run in
+an environment where /tmp has rarely been used, the newly created
temporary file
+may have a very low inode number. If the inode number was 6 which is the
same as
+/dev/null, the as would wrongly think the input file and the output file
are the
+same and report the following error.
+
+*** Compiler lacks asm-goto support.. Stop.
+
+One observed case happened in docker where the /tmp could be so rarely
used that
+very low number inode may be allocated and triggers the error.
+
+The fix below for the bug only exists on the master branch of binutils so
far
+and has not been released from upstream. As the convict is introduced
since
+v2.31, only v2.31 is affected.
+
+https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=
+2a50366ded329bfb39d387253450c9d5302c3503
<https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=+2a50366ded329bfb39d387253450c9d5302c3503>
+
+When building linux-libc-headers we need to use "as" in binutils which
does not
+contain the fix for the moment. To work around the error, we create a
fixed
+temporary file to contain the program being tested.
+
+This patch also removes ">/dev/null 2>&1" so we will have more direct
error
+information in case something else wrong happened.
+
+Upstream-Status: Inappropriate [A work around for binutils v2.31]
It would be nice to have a way to test for the host binutils version, since
without it
we have no way to know when we can stop carrying this change.
Not that the change is all that complex or should cause any problems, it is just
very open ended without a check.
Did you look into any conditional way to test the binutils in the recipe and
only
patch if required ?
I have not tried but we can prepend the do_patch of linux-libc-headers to
check the host binutils' version and modify the SRCURI there depending on
the result.
But the users may never want to upgrade their host binutils and thus need this
workaround forever. So even we know what the version is on the host and
warn them to upgrade, it seems we still don't know when we can drop the
workaround. I'm not sure the policy. Can we drop it when all supported
distributions have included the fix?
That is my suggestion (that we have a way to eventually drop the
patch), so either
we have to manually keep track of the host binutils versions and
revisit the patch
each release, we have some sort of check and conditional apply or we submit the
change to the upstream binutils -stable in the hopes that any distro
still tracking
v2.3x would bump to a 2.3x version with the change, and then we'd no longer have
to carry it anymore ... (but as you say in the commit it isn't a
problem in other
binutils, just this version, so there's likely no way to submit it
upstream and have
it do any good .. the problem is solved already).
That being said, I was more interested in hearing if you had looked into
conditional application, or if you had considered how we could
eventually drop the
patch, since I agree that as it currently stands .. we may have to
carry it for a long
time. We've had that discussion now, so we are covered.
Since the patch is simple enough, I'm not insisting on the complexity
of a dynamic
check, and I can't see another way to fix it.
So if you could double check the link to the upstream commit, I can
ack the patch
and it can go in as-is.
Cheers,
Bruce
Thanks,
Zhe
Bruce
+
+Signed-off-by: He Zhe <zhe...@windriver.com <mailto:zhe...@windriver.com>>
+---
+ scripts/gcc-goto.sh | 7 ++++++-
+ 1 file changed, 6 insertions(+), 1 deletion(-)
+
+diff --git a/scripts/gcc-goto.sh b/scripts/gcc-goto.sh
+index 083c526..0aaf1b4 100755
+--- a/scripts/gcc-goto.sh
++++ b/scripts/gcc-goto.sh
+@@ -3,7 +3,9 @@
+ # Test for gcc 'asm goto' support
+ # Copyright (C) 2010, Jason Baron <jba...@redhat.com
<mailto:jba...@redhat.com>>
+
+-cat << "END" | $@ -x c - -c -o /dev/null >/dev/null 2>&1 && echo "y"
++TMPFILE=`mktemp -p .`
++
++cat << "END" > ${TMPFILE}
+ int main(void)
+ {
+ #if defined(__arm__) || defined(__aarch64__)
+@@ -20,3 +22,6 @@ entry:
+ return 0;
+ }
+ END
++
++$@ -x c ${TMPFILE} -c -o /dev/null && echo "y"
++rm ${TMPFILE}
+--
+2.7.4
+
diff --git a/meta/recipes-kernel/linux-libc-headers/linux-libc-headers_4.18.bb
<http://linux-libc-headers_4.18.bb>
b/meta/recipes-kernel/linux-libc-headers/linux-libc-headers_4.18.bb
<http://linux-libc-headers_4.18.bb>
index eb7bee7..00420aa 100644
--- a/meta/recipes-kernel/linux-libc-headers/linux-libc-headers_4.18.bb
<http://linux-libc-headers_4.18.bb>
+++ b/meta/recipes-kernel/linux-libc-headers/linux-libc-headers_4.18.bb
<http://linux-libc-headers_4.18.bb>
@@ -9,5 +9,9 @@ SRC_URI_append_libc-musl = "\
file://0001-include-linux-stddef.h-in-swab.h-uapi-header.patch \
"
+SRC_URI_append = "\
+
file://0001-scripts-Use-fixed-temporary-file-instead-of-pipe-for.patch \
+"
+
SRC_URI[md5sum] = "bee5fe53ee1c3142b8f0c12c0d3348f9"
SRC_URI[sha256sum] =
"19d8bcf49ef530cd4e364a45b4a22fa70714b70349c8100e7308488e26f1eaf1"
--
2.7.4
--
_______________________________________________
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
<mailto:Openembedded-core@lists.openembedded.org>
http://lists.openembedded.org/mailman/listinfo/openembedded-core
--
- Thou shalt not follow the NULL pointer, for chaos and madness await thee at
its end
- "Use the force Harry" - Gandalf, Star Trek II
--
Kind regards,
Martin Hundebøll
Embedded Linux Consultant
+45 61 65 54 61
mar...@geanix.com
Geanix IVS
https://geanix.com
DK39600706
--
_______________________________________________
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.openembedded.org/mailman/listinfo/openembedded-core
