python roadmap

python roadmap

[Michael 'Mickey' Lauer]

You may have seen that I just branched off mickey/python2.6. Python 2.6 has 
been in org.oe.dev for quite a while and I have did extensive tests with it. 
It's now ready for public consumption, however since it's a major version 
upgrade, I need to bump all packages that ship something to 
${libdir}/python2.x.

To get this in one commit, I'll do it in the branch. I'll also take this 
chance to go over all python related packages and run a cleanup round -- if 
there's something to clean up.

I expect this to be complete in one or two weeks. Please hold off any major 
changes to Python-recipes until then.

With regards to Python 3, I will probably add it in February. It's going to be 
a long time until the critical mass of external packages have upgraded to v3, 
I expect not before 2010. We can take it easy with BitBake as well, but we 
should start removing the stuff that Python 2.6 claims to be deprecated asap.

Cheers,

Mickey.

Re: python roadmap

[Michael 'Mickey' Lauer]

Thanks to some extrashifts this work has proceeded faster than scheduled.
I plan to merge mickey/python26 into master on friday. If you see something 
dangerous, yell.

-- 
:M:

Re: python roadmap

[Vitus Jensen]

Am Wed, 14 Jan 2009 15:47:26 +0100 schrieb Michael 'Mickey' Lauer:

> Thanks to some extrashifts this work has proceeded faster than
> scheduled. I plan to merge mickey/python26 into master on friday. If you
> see something dangerous, yell.

Yell ;-)

See http://bugs.openembedded.net/show_bug.cgi?id=5013

python26 fails to build for the big-endian machines I tried, namely 
nslu2be.  And as python25 was removed it is a real blocker for packages 
which DEPEND on pathon, which inturn is the case for php.  And php I'm 
trying to get a version working again.

do_compile tries to load a .so on the host which was build for armeb and 
complains about wrong byteorder.  But there is no complain when 
MACHINE=nslu2le in which case the .so shouldn't be loadable, too (it's 
for arm after all, not x86).  AND python25 used exactly the same 
do_compile method, and does not fail!

So I'm a little confused how it's supposed to work.  Could you please 
take a look at it?

Thanks,
  Vitus

-- 
Vitus Jensen, Hannover, Germany, Earth, Milky Way, Universe (current)

Re: python roadmap

[Vitus Jensen]

Am Thu, 05 Feb 2009 08:35:31 +0000 schrieb Vitus Jensen:

> Am Wed, 14 Jan 2009 15:47:26 +0100 schrieb Michael 'Mickey' Lauer:
> 
>> Thanks to some extrashifts this work has proceeded faster than
>> scheduled. I plan to merge mickey/python26 into master on friday. If
>> you see something dangerous, yell.
> 
> Yell ;-)
> 
> See http://bugs.openembedded.net/show_bug.cgi?id=5013
> 
> python26 fails to build for the big-endian machines I tried, namely
> nslu2be.  And as python25 was removed it is a real blocker for packages
> which DEPEND on pathon, which inturn is the case for php.  And php I'm
> trying to get a version working again.
...

Does anyone build packages for bigendian machines?

Vitus

-- 
Vitus Jensen, Hannover, Germany, Earth, Milky Way, Universe (current)

Re: python roadmap

[Koen Kooi]

On 17-02-09 07:35, Vitus Jensen wrote:
> Am Thu, 05 Feb 2009 08:35:31 +0000 schrieb Vitus Jensen:
>
>> Am Wed, 14 Jan 2009 15:47:26 +0100 schrieb Michael 'Mickey' Lauer:
>>
>>> Thanks to some extrashifts this work has proceeded faster than
>>> scheduled. I plan to merge mickey/python26 into master on friday. If
>>> you see something dangerous, yell.
>> Yell ;-)
>>
>> See http://bugs.openembedded.net/show_bug.cgi?id=5013
>>
>> python26 fails to build for the big-endian machines I tried, namely
>> nslu2be.  And as python25 was removed it is a real blocker for packages
>> which DEPEND on pathon, which inturn is the case for php.  And php I'm
>> trying to get a version working again.
> ...
>
> Does anyone build packages for bigendian machines?

yes

Re: python roadmap

[Vitus Jensen]

Am Tue, 17 Feb 2009 09:09:43 +0100 schrieb Koen Kooi:

> On 17-02-09 07:35, Vitus Jensen wrote:
>> Am Thu, 05 Feb 2009 08:35:31 +0000 schrieb Vitus Jensen:
>>
>>> Am Wed, 14 Jan 2009 15:47:26 +0100 schrieb Michael 'Mickey' Lauer:
>>>
>>>> Thanks to some extrashifts this work has proceeded faster than
>>>> scheduled. I plan to merge mickey/python26 into master on friday. If
>>>> you see something dangerous, yell.
>>> Yell ;-)
>>>
>>> See http://bugs.openembedded.net/show_bug.cgi?id=5013
>>>
>>> python26 fails to build for the big-endian machines I tried, namely
>>> nslu2be.  And as python25 was removed it is a real blocker for
>>> packages which DEPEND on pathon, which inturn is the case for php. 
>>> And php I'm trying to get a version working again.
>> ...
>>
>> Does anyone build packages for bigendian machines?
> 
> yes

OK.  Restarting from scratch...


-- 
Vitus Jensen, Hannover, Germany, Earth, Milky Way, Universe (current)

Re: python roadmap

[Vitus Jensen]

Am Tue, 17 Feb 2009 10:03:12 +0000 schrieb Vitus Jensen:

> Am Tue, 17 Feb 2009 09:09:43 +0100 schrieb Koen Kooi:
> 
>> On 17-02-09 07:35, Vitus Jensen wrote:
>>> Am Thu, 05 Feb 2009 08:35:31 +0000 schrieb Vitus Jensen:
>>>
>>>> Am Wed, 14 Jan 2009 15:47:26 +0100 schrieb Michael 'Mickey' Lauer:
>>>>
>>>>> Thanks to some extrashifts this work has proceeded faster than
>>>>> scheduled. I plan to merge mickey/python26 into master on friday. If
>>>>> you see something dangerous, yell.
>>>> Yell ;-)
>>>>
>>>> See http://bugs.openembedded.net/show_bug.cgi?id=5013
>>>>
>>>> python26 fails to build for the big-endian machines I tried, namely
>>>> nslu2be.  And as python25 was removed it is a real blocker for
>>>> packages which DEPEND on pathon, which inturn is the case for php.
>>>> And php I'm trying to get a version working again.
>>> ...
>>>
>>> Does anyone build packages for bigendian machines?
>> 
>> yes
> 
> OK.  Restarting from scratch...

... doesn't build after clearing TMPDIR.

I will reinstall my slug to nslu2le and build those packages.

-- 
Vitus Jensen, Hannover, Germany, Earth, Milky Way, Universe (current)

Re: python roadmap

[Vitus Jensen]

Am Thu, 05 Feb 2009 08:35:31 +0000 schrieb Vitus Jensen:

> Am Wed, 14 Jan 2009 15:47:26 +0100 schrieb Michael 'Mickey' Lauer:
> 
>> Thanks to some extrashifts this work has proceeded faster than
>> scheduled. I plan to merge mickey/python26 into master on friday. If
>> you see something dangerous, yell.
> 
> Yell ;-)
> 
> See http://bugs.openembedded.net/show_bug.cgi?id=5013
> 
> python26 fails to build for the big-endian machines I tried, namely
> nslu2be.  And as python25 was removed it is a real blocker for packages
> which DEPEND on pathon, which inturn is the case for php.  And php I'm
> trying to get a version working again.
> 
> do_compile tries to load a .so on the host which was build for armeb and
> complains about wrong byteorder.  But there is no complain when
> MACHINE=nslu2le in which case the .so shouldn't be loadable, too (it's
> for arm after all, not x86).  AND python25 used exactly the same
> do_compile method, and does not fail!

Yes yes, I know: python builds fine on all hosts ;-)

But here is a patch set which forces do_compile() to load shared
objects from STAGING_LIBDIR_NATIVE instead of work/python-2.6.1/
(which contains target code).  Additionally I removed Makefile
patching which created senseless paths
(like /oestuff/.../arm/python/oestuff/...arm/python).  Tested
on ixp4xxbe, nslu2be and nslu2le.


diff --git a/packages/python/python-2.6.1/07-ld_library_path.patch b/packages/python/python-2.6.1/07-ld_library_path.patch
new file mode 100644
index 0000000..bdc47d7
--- /dev/null
+++ b/packages/python/python-2.6.1/07-ld_library_path.patch
@@ -0,0 +1,16 @@
+# RUNSHARED uses cwd as LD_LIBRARY_PATH which contains target executable
+# code.  LDLIBRARYDIR can be passed from the receipe to point to native
+# (host executable) code.
+
+diff -Naur Python-2.6.1.orig/configure.in Python-2.6.1/configure.in
+--- Python-2.6.1.orig/configure.in	2009-02-26 00:24:41.000000000 +0100
++++ Python-2.6.1/configure.in	2009-02-26 00:34:32.000000000 +0100
+@@ -707,7 +707,7 @@
+     Linux*|GNU*|NetBSD*|FreeBSD*|DragonFly*)
+ 	  LDLIBRARY='libpython$(VERSION).so'
+ 	  BLDLIBRARY='-L. -lpython$(VERSION)'
+-	  RUNSHARED=LD_LIBRARY_PATH=`pwd`:${LD_LIBRARY_PATH}
++	  RUNSHARED=LD_LIBRARY_PATH='${LDLIBRARYDIR}':${LD_LIBRARY_PATH}
+ 	  case $ac_sys_system in
+ 	      FreeBSD*)
+ 		SOVERSION=`echo $SOVERSION|cut -d "." -f 1`
diff --git a/packages/python/python_2.6.1.bb b/packages/python/python_2.6.1.bb
index 57c64df..4ec9eae 100644
--- a/packages/python/python_2.6.1.bb
+++ b/packages/python/python_2.6.1.bb
@@ -12,6 +12,7 @@ SRC_URI = "\
   file://04-default-is-optimized.patch;patch=1 \
   file://05-enable-ctypes-cross-build.patch;patch=1 \
   file://06-libffi-enable-default-mips.patch;patch=1 \
+  file://07-ld_library_path.patch;patch=1 \
   file://99-ignore-optimization-flag.patch;patch=1 \
   \
 # not yet pushed forward
@@ -41,17 +42,13 @@ do_compile_prepend() {
 	install -d ${STAGING_INCDIR}/python${PYTHON_MAJMIN}/
 	install -d ${STAGING_LIBDIR}/python${PYTHON_MAJMIN}/config/
 	install -m 0644 pyconfig.h ${STAGING_INCDIR}/python${PYTHON_MAJMIN}/
-	install -m 0644 Makefile Makefile.orig
-	install -m 0644 Makefile Makefile.backup
-	sed -e 's,${includedir},${STAGING_INCDIR},' < Makefile.backup > Makefile
-	install -m 0644 Makefile Makefile.backup
-	sed -e 's,${libdir},${STAGING_LIBDIR},' < Makefile.backup > Makefile
 	install -m 0644 Makefile ${STAGING_LIBDIR}/python${PYTHON_MAJMIN}/config/
 }
 
 do_compile() {
 	oe_runmake HOSTPGEN=${STAGING_BINDIR_NATIVE}/pgen \
 		HOSTPYTHON=${STAGING_BINDIR_NATIVE}/python \
+		LDLIBRARYDIR=${STAGING_LIBDIR_NATIVE} \
 		STAGING_LIBDIR=${STAGING_LIBDIR} \
 		STAGING_INCDIR=${STAGING_INCDIR} \
 		BUILD_SYS=${BUILD_SYS} HOST_SYS=${HOST_SYS} \
@@ -61,6 +58,7 @@ do_compile() {
 
 	oe_runmake HOSTPGEN=${STAGING_BINDIR_NATIVE}/pgen \
 		HOSTPYTHON=${STAGING_BINDIR_NATIVE}/python \
+		LDLIBRARYDIR=${STAGING_LIBDIR_NATIVE} \
 		STAGING_LIBDIR=${STAGING_LIBDIR} \
 		STAGING_INCDIR=${STAGING_INCDIR} \
 		BUILD_SYS=${BUILD_SYS} HOST_SYS=${HOST_SYS} \
@@ -73,10 +71,9 @@ do_stage() {
 }
 
 do_install() {
-	install -m 0644 Makefile.orig Makefile
-	
 	oe_runmake HOSTPGEN=${STAGING_BINDIR_NATIVE}/pgen \
 		HOSTPYTHON=${STAGING_BINDIR_NATIVE}/python \
+		LDLIBRARYDIR=${STAGING_LIBDIR_NATIVE} \
 		STAGING_LIBDIR=${STAGING_LIBDIR} \
 		STAGING_INCDIR=${STAGING_INCDIR} \
 		BUILD_SYS=${BUILD_SYS} HOST_SYS=${HOST_SYS} \


-- 
Vitus Jensen, Hannover, Germany, Earth, Milky Way, Universe (current)