SPECS: python-xmlrpc.spec - I can be python-aware: arch-independen...

[Radoslaw Zielinski]

Paweł Sakowski <saq at pld-linux.org> [06-07-2005 15:46]:
> On Wed, 2005-07-06 at 13:53 +0200, Radoslaw Zielinski wrote:
>>>> and *.py{c,o} shouldn't differ between archs.  They do.
>>> They don't.
>> The one I've checked before writing that mail did (python-ipaddr). ;-]
> Read my lips: if they do differ, that's only in the "timestamp" field.
> diff shows such files as different, but closer inspection proves that
> there are no functional differences. Further: the serialized bytecode is
> strictly speaking identical.

Fine.

>> So, usually, they don't differ, but not always.  Question: do we allow
>> noarch packages, containing files, which differ between architectures?
>> IIRC: no, but I may be wrong.
> It's normal that different builders produce non-identical rpm files out
> of noarch specs. It's most common for %{BUILDTIME} and files' mtime to
> vary. With py[co] contents it's the same situation.

For RPMS, it's normal, but it's not normal to produce different *files*
in noarch packages.  The situation we're talking about is the only one
I know.

%{BUILDTIME} (or other RPM header's fields) and ctime/mtime are
meaningless in this context.

> The important thing is that a packages built on various builders are
> functionally equivalent, that is, can be freely interchanged between all
> archs.

Agreed.  But IIRC someone (havner?) wanted all noarch packages to have
identical content.

I don't really care, just want to know if the packages containing *.py[co]
can be noarch, in case I'd touch the Python mess again.  From ankry's
"No." I understand they can't.

-- 
Radosław Zieliński <radek at pld-linux.org>