[PATCH 0/1] fix python md5/sha checksum incorrect on qemu86-64 with corei7-64 as DEFAULTTUNE

[PATCH 0/1] fix python md5/sha checksum incorrect on qemu86-64 with corei7-64 as DEFAULTTUNE

[Hongxu Jia]

The following changes since commit 698c74c373110ed081a7586e21d4a27b8b44c89b:

  libsdl: remove redundant configure_tweak patch (2015-12-01 21:32:15 +0000)

are available in the git repository at:

  git://git.pokylinux.org/poky-contrib hongxu/fix-python
  http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=hongxu/fix-python

Hongxu Jia (1):
  python: gcc optimization level 2

 ...001-configure.ac-gcc-optimization-level-2.patch | 40 ++++++++++++++++++++++
 meta/recipes-devtools/python/python_2.7.9.bb       |  1 +
 2 files changed, 41 insertions(+)
 create mode 100644 meta/recipes-devtools/python/python/0001-configure.ac-gcc-optimization-level-2.patch

-- 
1.9.1

[PATCH 1/1] python: gcc optimization level 2

[Hongxu Jia]

The qemu didn't support '-march=corei7 -mtune=corei7' and
gcc optimization level 3 '-O3' very well, in which the
/usr/lib64/python2.7/test/test_md5.py and
/usr/lib64/python2.7/test/test_sha.py failed.

So we use gcc optimization level 2 to instead.

YOCTO [# 8747]

Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
---
 ...001-configure.ac-gcc-optimization-level-2.patch | 40 ++++++++++++++++++++++
 meta/recipes-devtools/python/python_2.7.9.bb       |  1 +
 2 files changed, 41 insertions(+)
 create mode 100644 meta/recipes-devtools/python/python/0001-configure.ac-gcc-optimization-level-2.patch

diff --git a/meta/recipes-devtools/python/python/0001-configure.ac-gcc-optimization-level-2.patch b/meta/recipes-devtools/python/python/0001-configure.ac-gcc-optimization-level-2.patch
new file mode 100644
index 0000000..f760403
--- /dev/null
+++ b/meta/recipes-devtools/python/python/0001-configure.ac-gcc-optimization-level-2.patch
@@ -0,0 +1,40 @@
+From 55520995fa6c72f6a93a7ccc8503167f359bc63d Mon Sep 17 00:00:00 2001
+From: Hongxu Jia <hongxu.jia@windriver.com>
+Date: Mon, 30 Nov 2015 06:25:08 -0500
+Subject: [PATCH] configure.ac: gcc optimization level 2
+
+The qemu didn't support '-march=corei7 -mtune=corei7' and
+gcc optimization level 3 '-O3' very well, in which the
+/usr/lib64/python2.7/test/test_md5.py and
+/usr/lib64/python2.7/test/test_sha.py failed.
+
+So we use gcc optimization level 2 to instead.
+
+Upstream-Status: Inappropriate [qemu corei7 specific]
+
+Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
+---
+ configure.ac | 4 ++--
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+diff --git a/configure.ac b/configure.ac
+index 87a4f02..d3ad673 100644
+--- a/configure.ac
++++ b/configure.ac
+@@ -1076,11 +1076,11 @@ then
+ 		# debug builds.
+ 		OPT="-g -O0 -Wall $STRICT_PROTO"
+ 	    else
+-		OPT="-g $WRAP -O3 -Wall $STRICT_PROTO"
++		OPT="-g $WRAP -O2 -Wall $STRICT_PROTO"
+ 	    fi
+ 	    ;;
+ 	*)
+-	    OPT="-O3 -Wall $STRICT_PROTO"
++	    OPT="-O2 -Wall $STRICT_PROTO"
+ 	    ;;
+ 	esac
+ 	case $ac_sys_system in
+-- 
+1.9.1
+
diff --git a/meta/recipes-devtools/python/python_2.7.9.bb b/meta/recipes-devtools/python/python_2.7.9.bb
index bc20417..12dd7fa 100644
--- a/meta/recipes-devtools/python/python_2.7.9.bb
+++ b/meta/recipes-devtools/python/python_2.7.9.bb
@@ -26,6 +26,7 @@ SRC_URI += "\
   file://parallel-makeinst-create-bindir.patch \
   file://use_sysroot_ncurses_instead_of_host.patch \
   file://avoid_parallel_make_races_on_pgen.patch \
+  file://0001-configure.ac-gcc-optimization-level-2.patch \
 "
 
 S = "${WORKDIR}/Python-${PV}"
-- 
1.9.1

Re: [PATCH 1/1] python: gcc optimization level 2

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 494 bytes --]

On 2 December 2015 at 05:06, Hongxu Jia <hongxu.jia@windriver.com> wrote:

> The qemu didn't support '-march=corei7 -mtune=corei7' and
> gcc optimization level 3 '-O3' very well, in which the
> /usr/lib64/python2.7/test/test_md5.py and
> /usr/lib64/python2.7/test/test_sha.py failed.
>
> So we use gcc optimization level 2 to instead.
>
> YOCTO [# 8747]
>

Has anyone decided if this is a bug in the compiler, or the Python code
making assumptions that don't hold at -O3?

Ross

[-- Attachment #2: Type: text/html, Size: 940 bytes --]

Re: [PATCH 1/1] python: gcc optimization level 2

[Hongxu Jia]

[-- Attachment #1: Type: text/plain, Size: 851 bytes --]

On 12/02/2015 05:19 PM, Burton, Ross wrote:
>
> On 2 December 2015 at 05:06, Hongxu Jia <hongxu.jia@windriver.com 
> <mailto:hongxu.jia@windriver.com>> wrote:
>
>     The qemu didn't support '-march=corei7 -mtune=corei7' and
>     gcc optimization level 3 '-O3' very well, in which the
>     /usr/lib64/python2.7/test/test_md5.py and
>     /usr/lib64/python2.7/test/test_sha.py failed.
>
>     So we use gcc optimization level 2 to instead.
>
>     YOCTO [# 8747]
>
>
> Has anyone decided if this is a bug in the compiler, or the Python 
> code making assumptions that don't hold at -O3?
>

The root cause is the qemu didn't support gcc optimization level 3 very 
well.

This fix is a workaround, and I file another defect to trace the root cause.
https://bugzilla.yoctoproject.org/show_bug.cgi?id=8748

//Hongxu

> Ross


[-- Attachment #2: Type: text/html, Size: 2418 bytes --]

Re: [PATCH 1/1] python: gcc optimization level 2

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 442 bytes --]

On 2 December 2015 at 09:51, Hongxu Jia <hongxu.jia@windriver.com> wrote:

> The root cause is the qemu didn't support gcc optimization level 3 very
> well.
>

Oh, so the problem only appears in qemu if you enable tuning that is known
to be broken in qemu?  I don't think we should apply the Python workaround
to oe-core as everyone will suffer for a configuration that is known to be
broken (and isn't enabled in oe-core).

Ross

[-- Attachment #2: Type: text/html, Size: 858 bytes --]

Re: [PATCH 1/1] python: gcc optimization level 2

[Hongxu Jia]

[-- Attachment #1: Type: text/plain, Size: 998 bytes --]

On 12/02/2015 05:57 PM, Burton, Ross wrote:
>
> On 2 December 2015 at 09:51, Hongxu Jia <hongxu.jia@windriver.com 
> <mailto:hongxu.jia@windriver.com>> wrote:
>
>     The root cause is the qemu didn't support gcc optimization level 3
>     very well.
>
>
> Oh, so the problem only appears in qemu if you enable tuning that is 
> known to be broken in qemu?

Yes, since I tested qemux86/qemux86-64 with default tune, and everything 
is ok.

> I don't think we should apply the Python workaround to oe-core as 
> everyone will suffer for a configuration that is known to be broken 
> (and isn't enabled in oe-core).

Trace the issue costs a lot of time,  the original failure was invoking
'smart update' in qemu, and found out the python md5/sha checksum
output incorrect, and then found out qemu86-64 with corei7-64 did
not support gcc optimization level 3 '-O3' very well [YOCTO #8748]

BTW, boot the same image with virtualbox, everything is ok

//Hongxu

>
> Ross


[-- Attachment #2: Type: text/html, Size: 2658 bytes --]

Re: [PATCH 1/1] python: gcc optimization level 2

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 460 bytes --]

On 2 December 2015 at 10:28, Hongxu Jia <hongxu.jia@windriver.com> wrote:

> Trace the issue costs a lot of time,  the original failure was invoking
> 'smart update' in qemu, and found out the python md5/sha checksum
> output incorrect, and then found out qemu86-64 with corei7-64 did
> not support gcc optimization level 3 '-O3' very well [YOCTO #8748]
>

Sure.  So you've found out that Python is broken, but who knows what else
is broken?

Ross

[-- Attachment #2: Type: text/html, Size: 909 bytes --]

Re: [PATCH 1/1] python: gcc optimization level 2

[Khem Raj]

[-- Attachment #1.1: Type: text/plain, Size: 1447 bytes --]


> On Dec 2, 2015, at 2:28 AM, Hongxu Jia <hongxu.jia@windriver.com> wrote:
> 
> On 12/02/2015 05:57 PM, Burton, Ross wrote:
>> 
>> On 2 December 2015 at 09:51, Hongxu Jia <hongxu.jia@windriver.com <mailto:hongxu.jia@windriver.com>> wrote:
>> The root cause is the qemu didn't support gcc optimization level 3 very well.
>> 
>> Oh, so the problem only appears in qemu if you enable tuning that is known to be broken in qemu?
> 
> Yes, since I tested qemux86/qemux86-64 with default tune, and everything is ok.
> 
>> I don't think we should apply the Python workaround to oe-core as everyone will suffer for a configuration that is known to be broken (and isn't enabled in oe-core).
> 
> Trace the issue costs a lot of time,  the original failure was invoking
> 'smart update' in qemu, and found out the python md5/sha checksum
> output incorrect, and then found out qemu86-64 with corei7-64 did
> not support gcc optimization level 3 '-O3' very well [YOCTO #8748]
> 
> BTW, boot the same image with virtualbox, everything is ok
> 

This is a Qemu bug, So it should be persued in that direction. Secondly, this patch should be applied only
for emulator BSPs and not across the board

> //Hongxu
> 
>> 
>> Ross
> 
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core


[-- Attachment #1.2: Type: text/html, Size: 4049 bytes --]

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