[Pkg-samba-maint] Processing of talloc_2.1.6-2_amd64.changes

[Jelmer Vernooij]

On 6 April 2016 20:10:15 BST, Andrew Bartlett <abartlet at samba.org> wrote:
>On Wed, 2016-04-06 at 10:38 +0000, Jelmer Vernooij wrote:
>> On Wed, Apr 06, 2016 at 09:07:36AM +0000, Debian FTP Masters wrote:
>> E.g. the symbol name issue hasn't been addressed, since we sitll
>> include the architecture. E.g building on ARM:
>> 
>> +++ dpkg-gensymbolscKvnsd	2016-04-06 10:03:36.195304221 +0000
>> @@ -1,16 +1,16 @@
>> -libpytalloc-util.cpython-35m-x86-64-linux-gnu.so.2 python3-talloc
>> #MINVER#
>> - PYTALLOC_UTIL.CPYTHON_35M_X86_64_LINUX_GNU_2.1.6 at PYTALLOC_UTIL.CPYT
>> HON_35M_X86_64_LINUX_GNU_2.1.6 2.1.6
>> - PYTALLOC_UTIL.PY3_2.1.5 at PYTALLOC_UTIL.PY3_2.1.5 2.1.6
>> - PYTALLOC_UTIL.PY3_2.1.6 at PYTALLOC_UTIL.PY3_2.1.6 2.1.6
>> - _pytalloc_get_mem_ctx at PYTALLOC_UTIL.PY3_2.1.6 2.1.6
>> - _pytalloc_get_ptr at PYTALLOC_UTIL.PY3_2.1.6 2.1.6
>> - _pytalloc_get_type at PYTALLOC_UTIL.PY3_2.1.6 2.1.6
>> - pytalloc_BaseObject_PyType_Ready at PYTALLOC_UTIL.PY3_2.1.6 2.1.6
>> - pytalloc_BaseObject_check at PYTALLOC_UTIL.PY3_2.1.6 2.1.6
>> - pytalloc_BaseObject_size at PYTALLOC_UTIL.PY3_2.1.6 2.1.6
>> - pytalloc_Check at PYTALLOC_UTIL.PY3_2.1.5 2.1.6
>> - pytalloc_GetBaseObjectType at PYTALLOC_UTIL.PY3_2.1.6 2.1.6
>> - pytalloc_GetObjectType at PYTALLOC_UTIL.PY3_2.1.5 2.1.6
>> - pytalloc_reference_ex at PYTALLOC_UTIL.PY3_2.1.5 2.1.6
>> - pytalloc_steal at PYTALLOC_UTIL.PY3_2.1.5 2.1.6
>> - pytalloc_steal_ex at PYTALLOC_UTIL.PY3_2.1.5 2.1.6
>> +libpytalloc-util.cpython-35m-i386-linux-gnu.so.2 python3-talloc
>> #MINVER#
>> + PYTALLOC_UTIL.CPYTHON_35M_I386_LINUX_GNU_2.1.6 at PYTALLOC_UTIL.CPYTHO
>> N_35M_I386_LINUX_GNU_2.1.6 2.1.6-2
>> + PYTALLOC_UTIL.PY3_2.1.5 at PYTALLOC_UTIL.PY3_2.1.5 2.1.6-2
>> + PYTALLOC_UTIL.PY3_2.1.6 at PYTALLOC_UTIL.PY3_2.1.6 2.1.6-2
>> + _pytalloc_get_mem_ctx at PYTALLOC_UTIL.PY3_2.1.6 2.1.6-2
>> + _pytalloc_get_ptr at PYTALLOC_UTIL.PY3_2.1.6 2.1.6-2
>> + _pytalloc_get_type at PYTALLOC_UTIL.PY3_2.1.6 2.1.6-2
>> + pytalloc_BaseObject_PyType_Ready at PYTALLOC_UTIL.PY3_2.1.6 2.1.6-2
>> + pytalloc_BaseObject_check at PYTALLOC_UTIL.PY3_2.1.6 2.1.6-2
>> + pytalloc_BaseObject_size at PYTALLOC_UTIL.PY3_2.1.6 2.1.6-2
>> + pytalloc_Check at PYTALLOC_UTIL.PY3_2.1.5 2.1.6-2
>> + pytalloc_GetBaseObjectType at PYTALLOC_UTIL.PY3_2.1.6 2.1.6-2
>> + pytalloc_GetObjectType at PYTALLOC_UTIL.PY3_2.1.5 2.1.6-2
>> + pytalloc_reference_ex at PYTALLOC_UTIL.PY3_2.1.5 2.1.6-2
>> + pytalloc_steal at PYTALLOC_UTIL.PY3_2.1.5 2.1.6-2
>> + pytalloc_steal_ex at PYTALLOC_UTIL.PY3_2.1.5 2.1.6-2
>> dh_makeshlibs: failing due to earlier errors
>
>So, the upstream ABI here does change per arch.  Is there a way to
>encode that in the symbols files?

>Otherwise, I guess we have no choice but to change (upstream) the ABI
>again before it becomes too widely used.  We do try and normalise out
>the cpython35 bit for the upstream ABI checker, but we only do it on
>the text files in ABI/if
We can disable symbol checking altogether. However, I think that would
be a bad idea. This is an upstream bug IMO. Let's include the patch
doko provided or get it fixed upstream before shipping these outside
experimental to avoid having to deal with ABI changes later.

Cheers,

Jelmer