Re: [PATCH] python.inc: add `--disable-ipv6` to `EXTRA_OECONF`

[Khem Raj <raj.khem@gmail.com>]

On 1/14/2011 2:12 PM, Khem Raj wrote:
> On Fri, Jan 14, 2011 at 1:44 PM, Koen Kooi<k.kooi@student.utwente.nl>  wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 14-01-11 21:16, Khem Raj wrote:
>>> On Fri, Jan 14, 2011 at 11:20 AM, Koen Kooi<k.kooi@student.utwente.nl>  wrote:
>>> On 14-01-11 19:58, Koen Kooi wrote:
>>>>>> On 14-01-11 10:10, Paul Menzel wrote:
>>>>>>>  From b2e249702a0f6b5e36a4b095f5a4c495aa5ec14f Mon Sep 17 00:00:00 2001
>>>>>>> From: Paul Menzel<paulepanter@users.sourceforge.net>
>>>>>>> Date: Fri, 14 Jan 2011 09:48:08 +0100
>>>>>>> Subject: [PATCH] python.inc: add `--disable-ipv6` to `EXTRA_OECONF`
>>>>>>
>>>>>>> Since commit d729d14d [1] task `configure` for `python_2.6.5` and `python_2.7.1` fails with the following error message when compiling with `angstrom_2008.1` for `MACHINE = "beagleboard"`.
>>>>>>
>>>>>>>       checking for major... yes
>>>>>>>       checking for getaddrinfo... yes
>>>>>>>       checking getaddrinfo bug... buggy
>>>>>>>       Fatal: You must get working getaddrinfo() function.
>>>>>>>                    or you can specify "--disable-ipv6".
>>>>>>>       ERROR: Function do_configure failed
>>>>>>
>>>>>>> Disable IPv6 for now in Python unconditionally until a better solution is found.
>>>>>>
>>>>>> I don't think disabling ipv6 is the way to go, I'm comparing the eglibc
>>>>>> before and after d729d14d to see if that reveals a clue. If it doesn't,
>>>>>> let's put in your patch for the time being.
>>>
>>> .... and the difference is that options-group.config is empty before d72
>>> and fully populated after it. OPTION_EGLIBC_ADVANCED_INET6 is set to Y
>>> in both cases it seems.
>>>
>>>> I was expecting it to converge to this. Do u see that libraries
>>>> libc.so before and after have different content too ?
>>>> Could be that the option config mechanism is getting it wrong somehow
>>>> may be I should only populate options-group.config
>>>> with disabled features
>>
>> Populating it only with disabled features would match the old behaviour
>> and should make python build again. I doesn't give us the root cause,
>> but at least it would build :)
>
> yeah it would be a bug in option config if it could not deal with an
> option being set to 'y' twice
>
>>

OK it has nothing to do with eglibc change. Infact its due to

commit a88aca1d7dfa3a08957dd49cb61bac850f197106
Author: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
Date:   Wed Jan 12 20:34:55 2011 +0100

     autotools.bbclass: pass distro_imposed_configure_flags



This makes --enable-ipv6 passed to configure of all recipes
which use autotools. Now in configure there is a runtime test
to determine buggygetaddrinfo and as we are cross compiling this test 
goes wrong and if ipv6 is enabled then its reported as fatal

if test "$ipv6" = "yes"; then
echo 'Fatal: You must get working getaddrinfo() function.'
echo '       or you can specify "--disable-ipv6"'.


One solution we have is that we cache this variable

I will post a patch in this direction soon.

Thanks
-Khem